Skip to content

Latest commit

 

History

History
7730 lines (5364 loc) · 316 KB

varnish_book.rst

File metadata and controls

7730 lines (5364 loc) · 316 KB
Raw

Contents

Abstract

The Varnish Book is the training material for Varnish Plus courses. This book teaches such concepts to understand the theory behind Varnish Cache 4. Covered are the Varnish finite state machine, design principles, HTTP, cache invalidation and more. With these foundations, the book builds practical knowledge on Varnish Configuration Language (VCL), Varnish Test Code (VTC) and Varnish utility programs such as varnishlog, varnishstat and varnishtest. Examples and exercises develop the needed skills to administrate and extend the functionality of Varnish. Also included are appendices that explain how to develop Varnish Modules (VMODs) and how to use selected modules of Varnish Plus.

Preface

  • Course for Varnish Plus
  • Learn specific features depending the course and your needs
  • Necessary Background
  • How to Use the Book
  • Acknowledgments

After finishing this course, you will be able to install and configure the Varnish server, and write effective VCL code. The Varnish Book is designed for attendees of Varnish Plus courses. Most of the presented material in this book applies to both, the open source Varnish Cache and the commercial edition Varnish Cache Plus. Therefore, you can also refer to the Varnish Cache documentation at https://www.varnish-cache.org/docs/4.0/.

Varnish Plus is a commercial suite by Varnish Software that offers products for scalability, customization, monitoring, and expert support services. The engine of Varnish Plus is Varnish Cache Plus, which is the enhanced commercial edition of Varnish Cache. Varnish Cache Plus should not be confused with Varnish Plus, a product offering by Varnish Software. Varnish Cache Plus is one of the software components available for Varnish Plus customers.

For simplicity, the book refers to Varnish Cache or Varnish Cache Plus as Varnish when there is no difference between them. There is more information about differences between Varnish Cache and Varnish Cache Plus in the Varnish Cache and Varnish Plus chapter.

The goal of this book is to make you confident when using Varnish. Varnish instructors focus on your area, needs or interest. Varnish courses are usually flexible enough to make room for it.

The instructor will cover selected material for the course you take. The System Administration (Admin) course provides attendees with the necessary knowledge to troubleshoot and tune common parameters of a Varnish server. The Web Developer (Webdev) course teaches how to adapt web applications so that they work with Varnish, which guarantees a fast experience for visitors of any website. Besides that, other courses may also be taught with this book.

Necessary Background

The Admin course requires that you:

  • have expertise in a shell on a Linux/UNIX machine, including editing text files and starting daemons,
  • understand HTTP cache headers,
  • understand regular-expressions, and
  • be able to install the software listed below.

The Webdev course requires that you:

  • have expertise in a shell on a Linux/UNIX machine, including editing text files and starting daemons,
  • understand HTTP cache headers,
  • understand regular-expressions, and
  • be able to install the software listed below.

You do not need background in theory or application behind Varnish to complete this course. However, it is assumed that you have experience and expertise in basic UNIX commands, and that you can install the following software:

  • Varnish Cache 4.x or Varnish Cache Plus 4.x,
  • Apache/2.4 or later,
  • HTTPie 0.8.0 or later,
  • PHP 5.4 or later, and
  • curl – command line tool for transferring data with URL syntax

More specific required skills depend on the course you take. The book starts with the installation of Varnish and navigation of some of the common configuration files. This part is perhaps the most UNIX-centric part of the course.

How to Use the Book

  • Most of the material in this book applies to both: Varnish Cache and Varnish Cache Plus. Parts that apply only to Varnish Cache Plus are clearly stated.
  • Varnish caching mechanisms are different than in other caching technologies. Open your mind and try to think different when using Varnish.
  • The instructor guides you through the book.
  • Use the manual pages and help options.
  • See Appendix E: Varnish Three Letter Acronyms for a list of acronyms.

The Varnish Book is designed to be used as training material under the Varnish Plus course taught by a certified instructor. Under the course, the instructor guides you and selects the relevant sections to learn. However, you can also use this book as self-instructional material.

There are almost always many ways to do an exercise. The solutions provided in this book or by your instructor are not necessarily better than your own.

Varnish installs several reference manuals that are accessible through the manual page command man. You can issue the command man -k varnish to list the manual pages that mention Varnish in their short description. In addition, the vsl man page that explains the Varnish Shared memory Logging (VSL). This man page does not come out when issuing man -k varnish, because it does not contain the word varnish in its short description.

The command man varnishd, for example, retrieves the manual page of the Varnish HTTP accelerator daemon. Also, some commands have a help option to print the usage of the command. For example, varnishlog -h prints the usage and options of the command with a short description of them.

In addition, you should refer to the documentation of Varnish Cache and Varnish Cache Plus. This documentation provides you extended details on the topics covered in this book and more. To access to this documentation, please visit https://www.varnish-software.com/resources.

The Varnish installation described in this book uses Ubuntu Linux 14.04 LTS (trusty), therefore most of the commands instructed in this book are for this Linux distribution. We point out some differences on how to configure Varnish for other Linux distributions, but you should reference your Linux distribution's documentation for more details.

The book is written with different formatting conventions. Varnish Configuration Language (VCL) code uses the mono-spaced font type inside boxes:

vcl 4.0;

backend default {
    .host = "127.0.0.1";
    .port = "8080";
}

sub vcl_recv {
    # Do request header transformations here.
    if (req.url ~ "/admin") {
        return(pass);
    }
}

The first occurrence of a new term is usually its definition, and appears in italics. File names are indicated like this: /path/to/yourfile. Important notes, tips and warnings are also inside boxes, but they use the normal body text font type.

Resources, and Errata

This book is meant to be understandable to everyone who takes a Varnish Plus course and has the required skills. If you find something unclear, do not be shy and blame yourself, ask your instructor for help. You can also contact the Varnish open source community at https://varnish-cache.org. To book training, please look at https://varnish-software.com/academy.

Additional examples from different Varnish versions are available at https://www.varnish-cache.org/trac/wiki/VCLExamples. These examples are maintained by the community.

For those interested in development, the developers arrange weekly bug washes were recent tickets and development is discussed. This usually takes place on Mondays around 13:00 CET on the IRC channel #varnish-hacking on irc.linpro.net.

Errata, updates and general improvements of this book are available at its repository https://github.com/varnish/Varnish-Book.

Acknowledgments

In addition to the authors, the following deserve special thanks (in no particular order):

  • Rubén Romero
  • Dag Haavi Finstad
  • Martin Blix Grydeland
  • Reza Naghibi
  • Federico G. Schwindt
  • Dridi Boukelmoune
  • Lasse Karstensen
  • Per Buer
  • Sevan Janiyan
  • Kacper Wysocki
  • Magnus Hagander
  • Arianna Aondio
  • Poul-Henning Kamp
  • Everyone who has participated on the training courses

Introduction

  • What is Varnish?
  • Benefits of Varnish
  • Open source / Free software
  • Varnish Software: The company
  • What is Varnish Plus?
  • Varnish: more than a cache server
  • History of Varnish
  • Varnish Governance Board (VGB)

What is Varnish?

Reverse Proxy

Figure :counter:`figure`: Reverse Proxy

Varnish is a reverse HTTP proxy, sometimes referred to as an HTTP accelerator or a web accelerator. A reverse proxy is a proxy server that appears to clients as an ordinary server. Varnish stores (caches) files or fragments of files in memory that are used to reduce the response time and network bandwidth consumption on future, equivalent requests. Varnish is designed for modern hardware, modern operating systems and modern work loads.

Varnish is flexible because you can configure it and write your own caching policies in its Varnish Configuration Language (VCL). VCL is a domain specific language based on C, therefore it compiles and executes lightning fast. Varnish has shown itself to work well both on large (and expensive) servers and tiny appliances.

Varnish is more than a reverse HTTP proxy that caches content to speed up your server. Depending on the installation, Varnish can also be used as:

  • web application firewall,
  • DDoS attacks defender,
  • load balancer,
  • integration point,
  • single sign-on gateway,
  • authentication and authorization policy mechanism,
  • quick fix for unstable backends, and
  • HTTP router.

Varnish Cache and Varnish Plus

Varnish Cache is an open source project, and free software. The development process is public and everyone can submit patches, or just take a peek at the code if there is some uncertainty on how does Varnish Cache work. There is a community of volunteers who help each other and newcomers. The BSD-like license used by Varnish Cache does not place significant restriction on re-use of the code, which makes it possible to integrate Varnish Cache in virtually any solution.

Varnish Cache is developed and tested on GNU/Linux and FreeBSD. The code-base is kept as self-contained as possible to avoid introducing out-side bugs and unneeded complexity. Therefore, Varnish uses very few external libraries.

Varnish Software is the company behind Varnish Cache. Varnish Software and the Varnish community maintain a package repository of Varnish Cache for several common GNU/Linux distributions.

Varnish Software also provides a commercial suite called Varnish Plus with software products for scalability, customization, monitoring and expert support services. The engine of the Varnish Plus commercial suite is the enhanced commercial edition of Varnish Cache. This edition is proprietary and it is called Varnish Cache Plus.

Table 1 shows the components covered in this book and their availability for Varnish Cache users and Varnish Plus customers. The covered components of Varnish Plus are described in the Varnish Plus Software Components chapter. For more information about the complete Varnish Plus offer, please visit https://www.varnish-software.com/what-is-varnish-plus.

At the moment of writing this book, Varnish Cache supports the operating systems and Linux distributions listed in Table 2.

Varnish Cache and Varnish Plus support only 64-bit systems.

Note

Varnish Cache Plus should not be confused with Varnish Plus, a product offering by Varnish Software. Varnish Cache Plus is one of the software components available for Varnish Plus customers.

Varnish Cache and Varnish Software Timeline

  • 2005: Ideas! Verdens Gang (www.vg.no, Norway's biggest newspaper) were looking for alternative cache solutions.
  • 2006: Work began. Redpill Linpro was in charge of project management, infrastructure and supporting development. Poul-Henning Kamp did the majority of the actual development.
  • 2006: Varnish 1.0 released.
  • 2008: Varnish 2.0 released.
  • 2009: The first Varnish User Group Meeting is held in London. Roughly a dozen people participate from all around the world.
  • 2010: Varnish Software is born as a spin-off to Redpill Linpro AS.
  • 2011: Varnish 3.0 released.
  • 2012: The fifth Varnish User Group Meeting is held in Paris. Roughly 70 people participate on the User-day and around 30 on the developer-day!
  • 2012: The Varnish Book is published.
  • 2013: Varnish Software chosen as a 2013 Red Herring Top 100 Europe company.
  • 2013: BOSSIE award winner.
  • 2013: Varnish Software receives World Summit on Innovation & Entrepreneurship Global Hot 100 award.
  • 2014: Varnish Plus launched.
  • 2014: Varnish 4.0 released.
  • 2015: Varnish API released.
  • 2015: Gartner names Varnish Software as a 2015 ‘Cool Vendor’ in Web-Scale Platforms.
  • 2015: Varnish Plus supports SSL/TLS.

VG, a large Norwegian newspaper, initiated the Varnish project in cooperation with Linpro. The lead developer of the Varnish project, Poul-Henning Kamp, is an experienced FreeBSD kernel hacker. Poul-Henning Kamp continues to bring his wisdom to Varnish in most areas where it counts.

From 2006 throughout 2008, most of the development was sponsored by VG, API, Escenic and Aftenposten, with project management, infrastructure and extra man-power provided by Redpill Linpro. At the time, Redpill Linpro had roughly 140 employees mostly centered around consulting services.

Today Varnish Software is able to fund the core development with income from service agreements, in addition to offering development of specific features on a case-by-case basis. The interest in Varnish continues to increase. An informal study based on the list of most popular web sites in Norway indicates that about 75% or more of the web traffic that originates in Norway is served through Varnish.

Varnish development is governed by the Varnish Governance Board (VGB), which thus far has not needed to intervene. The VGB consists of an architect, a community representative and a representative from Varnish Software.

As of November 2015, the VGB positions are filled by Poul-Henning Kamp (Architect), Rogier Mulhuijzen (Community) and Lasse Karstensen (Varnish Software). On a day-to-day basis, there is little need to interfere with the general flow of development.

What Is New in Varnish 4

  • Version statement vcl 4.0;
  • req.request is now req.method
  • vcl_fetch is now vcl_backend_response
  • Directors have been moved to the vmod_directors
  • Hash directors as a client directors
  • vcl_error is now vcl_backend_error
  • error() is now synth(), and you must explicitly return it: return (synth(999, "Response"));
  • Synthetic responses in vcl_synth
  • Setting headers on synthetic response bodies made in vcl_synth are now done on resp.http instead of obj.http.
  • obj.* in vcl_error replaced by beresp.* in vcl_backend_error
  • hit_for_pass objects are created using beresp.uncacheable
  • req.* not available in vcl_backend_response
  • bereq.* in vcl_backend_response
  • vcl_* prefix reserved for builtin subroutines
  • req.backend.healthy replaced by std.healthy(req.backend_hint)
  • client.port and server.port replaced by std.port(client.ip) and std.port(server.ip)
  • Cache invalidation with purges is now done via return(purge) in vcl_recv
  • obj.* is now read-only
  • obj.last_use is retired
  • vcl_recv must now return hash instead of lookup
  • vcl_hash must now return lookup instead of hash
  • vcl_pass must now return fetch instead of pass
  • restart in the backend is now retry, this is now called return(retry), and jumps back up to vcl_backend_fetch
  • default VCL is now called builtin VCL
  • The builtin VCL now honors Cache-Control: no-cache (and friends) to indicate uncacheable content from the backend
  • remove keyword replaced by unset
  • X-Forwarded-For is now set before vcl_recv
  • session_linger has been renamed to timeout_linger and it is in seconds now (previously was milliseconds)
  • sess_timeout is renamed to timeout_idle
  • Increasing sess_workspace is not longer necessary, you may need to increase either workspace_backend or workspace_client
  • thread_pool_purge_delay is renamed to thread_pool_destroy_delay and it is in seconds now
  • thread_pool_add_delay and thread_pool_fail_delay are in seconds now
  • New parameter vcc_allow_inline_c to disable inline C in your VCL
  • New query language to filter logs: -m option replaced by -q

The above list tries to summarize the most important changes from Varnish Cache 3 to Varnish Cache 4. For more information, please visit: https://www.varnish-cache.org/docs/trunk/whats-new/upgrading.html

If you want to migrate your VCL code from Varnish 3 to Varnish 4, you may be interested in looking at the varnish3to4 script. See the VCL Migrator from Varnish 3 to Varnish 4 section for more information.

Design Principles

Varnish is designed to:

  • Solve real problems
  • Run on modern hardware (64-bit multi-core architectures)
  • Work with the kernel, not against it
  • Translate Varnish Configuration Language (VCL) to C programming language
  • Be extendible via Varnish Modules (VMODs)
  • Reduce lock-contention via its workspace-oriented shared memory model

The focus of Varnish has always been performance and flexibility. Varnish is designed for hardware that you buy today, not the hardware you bought 15 years ago. This is a trade-off to gain a simpler design and focus resources on modern hardware. Varnish is designed to run on 64-bit architectures and scales almost proportional to the number of CPU cores you have available. Though CPU-power is rarely a problem.

32-bit systems, in comparison to 64-bit systems, allow you to allocate less amount of virtual memory space and less number of threads. The theoretical maximum space depends on the operating system (OS) kernel, but 32-bit systems usually are bounded to 4GB. You may get, however, about 3GB because the OS reserves some space for the kernel.

Varnish uses a workspace-oriented memory-model instead of allocating the exact amount of space it needs at run-time. Varnish does not manage its allocated memory, but it delegates this task to the OS because the kernel can normally do this task better than a user-space program.

Event filters and notifications facilities such as epoll and kqueue are advanced features of the OS that are designed for high-performance services like Varnish. By using these, Varnish can move a lot of the complexity into the OS kernel which is also better positioned to decide which threads are ready to execute and when.

In addition, Varnish uses the Varnish Configuration Language (VCL) that is translated to C programming language code. This code is compiled with a standard C compiler and then dynamically linked directly into Varnish at run-time. This has several advantages. The most practical is the freedom you get as system administrator.

You can use VCL to decide how you want to interact with Varnish, instead of having a developer trying to predict every possible caching scenario. The fact that VCL is translated to C code, gives Varnish a very high performance. You can also by-pass the process of code translation and write raw C code, this is called in-line C in VCL. In short: VCL allows you to specify exactly how to use and combine the features of Varnish.

Varnish allows integration of Varnish Modules also known as VMODs. VMODs are typically coded in VCL and C programming language. A VMOD is a shared library, which can be called from VCL code.

The standard (std) VMOD, for instance, is a VMOD included in Varnish Cache. These modules let you extend the functionality of VCL by pulling in custom-written features. Some examples include non-standard header manipulation, access to memcached or complex normalization of headers. Appendix D: VMOD Development explains in more details how VMODs work and how to develop yours.

The Varnish Shared memory Log (VSL) allows Varnish to log large amounts of information at almost no cost by having other applications parse the data and extract the useful bits. This design and other mechanisms decrease lock-contention in the heavily threaded environment of Varnish.

To summarize: Varnish is designed to run on modern hardware under real work-loads and to solve real problems. Varnish does not cater to the "I want to make Varnish run on my 486 just because"-crowd. If it does work on your 486, then that's fine, but that's not where you will see our focus. Nor will you see us sacrifice performance or simplicity for the sake of niche use-cases that can easily be solved by other means -- like using a 64-bit OS.

How objects are stored

  • Object: local store of HTTP response message
  • Objects in Varnish are stored in memory and addressed by hash keys
  • You can control the hashing
  • Multiple objects can have the same hash key

Objects are local stores of response messages as defined in https://tools.ietf.org/html/rfc7234. .. hash Varnish maps this objects with a hash key. Objects are stored in memory and references to these objects are kept in a hash tree.

A rather unique feature of Varnish is that it allows you to control the input of the hashing algorithm. The key is by default made out of the HTTP Host header and the URL, which is sufficient and recommended for typical cases. However, you are able to create the key from something else. For example, you can use cookies or the user-agent of a client request to create a hash key.

HTTP specifies that multiple objects can be served from the same URL, depending on the preferences of the client. For instance, content in gzip format is sent only to clients that indicate gzip support. Varnish stores various objects under one key. Upon a client request, Varnish selects the object that matches the client preferences.

Object Lifetime

Figure :counter:`figure`: Object Lifetime

Figure 2 shows the lifetime of cached objects. A cached object has an origin timestamp t_origin and three duration attributes: 1) TTL, 2) grace, and 3) keep. t_origin is the time when an object was created in the backend. An object lives in cache until TTL + grace + keep elapses. After that time, the object is removed by the Varnish daemon.

In a timeline, objects within the time-to-live TTL are considered fresh objects. Stale objects are those within the time period TTL and grace. Objects within t_origin and keep are used when applying conditions with the HTTP header field If-Modified-Since.

The VCL – vcl_backend_fetch and VCL – vcl_backend_response sections explain how Varnish handles backend responses and how these duration attributes affect subsequent actions.

Getting Started

In this chapter, you will:

  • learn about the Varnish distribution,
  • install Varnish,
  • learn about varnishtest, and
  • cover basic Varnish configuration.

Most of the commands you will type in this course require root privileges. You can get temporary root privileges by typing sudo <command>, or permanent root privileges by typing sudo -i.

In Varnish terminology, a backend is the origin server. In other words, it is whatever server Varnish talks to fetch content. This can be any sort of service as long as it understands HTTP. Most of the time, Varnish talks to a web server or an application frontend server. In this book, we use backend, origin server, web server or application frontend server depending the context.

Varnish Distribution

Utility programs part of the Varnish distribution:

  • varnishd
  • varnishtest
  • varnishadm
  • varnishlog
  • varnishstat
  • and more

The Varnish distribution includes several utility programs that you will use in this course. You will learn how to use these programs as you progress, but it is useful to have a brief introduction about them before we start.

The central block of Varnish is the Varnish daemon varnishd. This daemon accepts HTTP requests from clients, sends requests to a backend, caches the returned objects and replies to the client request. varnishd is further explained in the Varnish Architecture section.

varnishtest is a script driven program used to test your Varnish installation. varnishtest is very powerful because it allows you to create client mock-ups, fetch content from mock-up or real backends, interact with your actual Varnish configuration, and assert the expected behavior. varnishtest is also very useful to learn more about the behavior of Varnish. Therefore, varnishtest is used throughout the book as main testbed.

varnishadm controls a running Varnish instance. The varnishadm utility establishes a command line interface (CLI) connection to varnishd. This utility is the only one that may affect a running instance of Varnish. You can use varnishadm to:

  • start and stop varnishd,
  • change configuration parameters,
  • reload the Varnish Configuration Language (VCL),
  • view the most up-to-date documentation for parameters, and
  • more.

The Management Interface varnishadm section explains in more detail this utility.

The Varnish log provides large amounts of information, thus it is usually necessary to filter it. For example, "show me only what matches X". varnishlog does precisely that. You will learn more about varnishlog in the Examining Varnish Server's Output chapter.

varnishstat is used to access global counters. It provides overall statistics, e.g the number of total requests, number of objects, and more. varnishstat is particularly useful when using it together with varnishlog to analyze your Varnish installation. The varnishstat section explains in detail this utility.

In addition, there are other utility programs such as varnishncsa, varnishtop and varnishhist. Appendix B: Varnish Programs explains them.

Exercise: Install Varnish

  • Install Varnish
  • Use packages provided by
    • varnish-software.com for Varnish Cache Plus
    • varnish-cache.org for Varnish Cache
  • When you are done, verify your Varnish version, run varnishd -V
You may skip this exercise if already have a well configured environment to test Varnish. In case you get stuck, you may look at the proposed solution.

[1] The file does not exist by default. Copy it from /lib/systemd/system/ and edit it.

[2] There is no configuration file. Use the command chkconfig varnishlog/varnishncsa on/off instead.

[3] There is no configuration file. Use the command systemctl start/stop/enable/disable/ varnishlog/varnishncsa instead.

The configuration file is used to give parameters and command line arguments to the Varnish daemon. This file also specifies the location of the VCL file. Modifications to this file require to run service varnish restart for the changes to take effect.

The location of the Varnish configuration file depends on the operating system and whether it uses the init system of SysV, or systemd. Table 3 shows the locations for each system installation.

To install packages on Ubuntu and Debian, use the command apt-get install <package>, e.g., apt-get install varnish. For Red Hat, use yum install <package>.

Solution: Install Varnish

All the following commands must be executed with root permissions. First, make sure you have apt-transport-https and curl:

$ apt-get install apt-transport-https
$ apt-get install curl

To use the varnish-software.com repository and install Varnish Cache Plus 4 on Ubuntu 14.04 trusty:

$ curl https://<username>:<password>@repo.varnish-software.com/GPG-key.txt \
  | apt-key add -

To use the varnish-cache.org repository and install Varnish Cache 4 on Ubuntu 14.04 trusty:

$ curl https://repo.varnish-cache.org/ubuntu/GPG-key.txt | apt-key add -
$ echo "deb https://repo.varnish-cache.org/ubuntu/ trusty varnish-4.0" >> \
  /etc/apt/sources.list.d/varnish-cache.list

If you are installing Varnish Cache Plus, add the repositories for VMODs in /etc/apt/sources.list.d/varnish-4.0-plus.list:

# Remember to replace DISTRO and RELEASE with what applies to your system.
# distro=(debian|ubuntu), RELEASE=(precise|trusty|wheezy|jessie)

# Varnish Cache Plus 4.0 and VMODs
$ deb https://<username>:<password>@repo.varnish-software.com/DISTRO RELEASE \
$ varnish-4.0-plus

# non-free contains VAC, VCS, Varnish Tuner and proprietary VMODs.
$ deb https://<username>:<password>@repo.varnish-software.com/DISTRO RELEASE \
$ non-free

Re-synchronize the package index files of your repository:

$ apt-get update

Install Varnish Cache Plus:

$ apt-get install varnish-plus

Install Varnish-Cache:

$ apt-get install varnish

Finally, verify the version you have installed:

$ varnishd -V

All software related to Varnish Cache Plus including VMODs are available in RedHat and Debian package repositories. These repositories are available on http://repo.varnish-software.com/, using your customer specific username and password.

Varnish is already distributed in many package repositories, but those packages might contain an outdated Varnish version. Therefore, we recommend you to use the packages provided by varnish-software.com for Varnish Cache Plus or varnish-cache.org for Varnish Cache. Please be advised that we only provide packages for LTS releases, not all the intermediate releases. However, these packages might still work fine on newer releases.

To use Varnish Cache Plus repositories on RHEL 6, put the following in /etc/yum.repos.d/varnish-4.0-plus.repo:

[varnish-4.0-plus]
name=Varnish Cache Plus
baseurl=https://<username>:<password>@repo.varnish-software.com/redhat
/varnish-4.0-plus/el$releasever
enabled=1
gpgcheck=0

[varnish-admin-console]
name=Varnish Administration Console
baseurl=
https://<username>:<password>@repo.varnish-software.com/redhat
/vac/el$releasever
enabled=1
gpgcheck=0

[varnishtuner]
name=Varnish Tuner
baseurl=
https://<username>:<password>@repo.varnish-software.com/redhat
/varnishtuner/el$releasever
enabled=1
gpgcheck=0

varnishtest

  • Script driven program used to test the configuration of Varnish, run regression tests, and develop VMODs
  • Useful for system administrators, web developers, and VMODs developers

Varnish is distributed with many utility programs. varnishtest is a script driven program that allows you create client mock-ups, simulate transactions, fetch content from mock-up or real backends, interact with your actual Varnish configuration and assert expected behaviors.

You can use varnishtest when configuring your Varnish installation, i.e., writing VCL code, or developing VMODs. In fact, we recommend you first to write Varnish Case Tests (VTCs) as part of your design. varnishtest is also useful to reproduce bugs when filing a bug report.

This section develops basic knowledge about varnishtest, and you will learn more about it throughout the book. Another way to learn how to create VTCs is by reading and running the .vtc files included in Varnish Cache under bin/varnishtest/tests/. Further documentation of varnishtest is found in its man page, bin/varnishtest/tests/README and https://www.varnish-cache.org/docs/trunk/reference/varnishtest.html.

varnishtest has its own language: the Varnish Test Case (VTC) language. This language is fairly simple to understand as we shall see next.

The Varnish Test Case (VTC) Language

  • Test against simulated or real backends
  • Starts real instance of varnishd
  • Simulates clients
  • Asserts using expect

b00001.vtc

varnishtest does not follow the unit testing framework (up/test/assert/tear down) nor behavior-driven development (given/when/then). Depending on your use case, there might be test preparations, executions and assertions all over the place. VTC is not compiled but simply interpreted on the fly.

There is a naming convention for VTC files. Files starting with b as the example above contain basic functionality tests. The naming scheme is in Varnish-Cache/bin/varnishtest/tests/README or https://raw.githubusercontent.com/varnish/Varnish-Cache/master/bin/varnishtest/tests/README.

All VTC programs start by naming the test:

varnishtest "Varnish as Proxy"

In this example we declare a simulated origin server:

server s1 {
   rxreq
   txresp
} -start

All server declarations must start with s. In the code above, s1 receives a request rxreq, and transmits a response txresp. -start boots s1 and makes available the macros ${s1_addr} and ${s1_port} with the IP address and port of your simulated backend. You may also start a declaration at a later point in your code, for example server s1 -start.

To declare an instance of your real Varnish server:

varnish v1 -arg "-b ${s1_addr}:${s1_port}" -start

varnish v1 declares an instance of your real Varnish server, i.e., varnishd. The names for Varnish servers must start with v. This instance is controlled by the manager process, and -start forks a child, which is the actual cacher process. You will learn about the manager and cacher in The Parent Process: The Manager and The Child Process: The Cacher sections.

There are many ways to configure varnishd. On way is by passing arguments with -arg as in -arg "-b ${s1_addr}:${s1_port}". -b is a varnishd option to define the backend. In this case, we use the IP address and port of the simulated backend s1, but you can also use a real backend. Therefore, varnishtest can be used as integration tool when testing your real backend.

There are other ways to define backends. The most common one is perhaps by defining them in your VCL code, as we shall see in the next section.

To simulate a client:

client c1 {
   txreq
   rxresp

   expect resp.http.via ~ "varnish"
} -run

Simulated clients in varnishtest start with c. In this example, c1 transmits one request and receives one response.

Since Varnish is a proxy, we expect to receive the response from the backend via Varnish. Therefore, c1 expects varnish in the via HTTP header field. We use tilde ~ as match operator of regular expressions because the exact text in resp.http.via depends on the Varnish version you have installed.

Finally, you start client c1 with the -run command.

Semaphores in Varnish Tests

  • Varnish is a multi-threaded program
  • Use -wait or sema as synchronization mechanism
  • -run: -start -wait

You might have noticed that we used -start for v1, but -run for c1. The difference between these commands is that -run executes -start -wait.

Varnish is a multi-threaded program. Therefore, each instance in varnishtest, i.e., s1, v1 and c1, is executed by a different thread. Sometimes, you will need some sort of synchronization mechanism to ensure you avoid race conditions or other non-intuitive behaviors. For those cases, you can use the -wait command.

-wait tells the executor of varnishtest to wait for a given instance to complete before proceeding to the next instruction in your VTC program. To illustrate this, see the difference between:

varnishtest "Synchronized"

server s1 {
        rxreq
        txresp
}

server s1 -start

client c1 -connect ${s1_sock} {
        txreq
        rxresp
}

# -run: -start - wait
client c1 -run

server s1 -wait

and:

varnishtest "Unsynchronized"

server s1 {
        rxreq
        txresp
}

server s1 -start -wait

client c1 -connect ${s1_sock} {
        txreq
        rxresp
}

client c1 -run

The second test fails in comparison to the first one, because varnishtest times out while waiting for s1 to receive a request and transmit a response. Therefore, you typically start Varnish servers with the -start command, but start clients with the -run command.

Another important synchronization mechanism is sema. There are many tests under Varnish-Cache/bin/varnishtest/tests that use sema. Please refer to them in order to learn how to use sema.

Note

You will learn more about the Threading Model of Varnish in its own section.

Note

Note that we do not instantiate a Varnish server in the examples, but connect the client directly to the server. For that purpose we use ${s1_sock}. This macro translates to the IP address and port of s1.

Running Your Varnish Tests

$varnishtest b00001.vtc
#     top  TEST b00001.vtc passed (1.458)

To run your test, you simply issue the command above. By default, varnishtest outputs the summary of passed tests, and a verbose output for failed tests only. If you want to always get a verbose output, run varnishtest with the -v option.

A passed test means that you have the most basic Varnish configuration correct in the testbed varnishtest. In the next section we explain how to configure Varnish in the way you normally would do after your tests have passed or when the varnishtest testbed is not enough for your needs.

There is much more to explain about varnishtest, but before that, you must learn more about the fundamentals of Varnish. We will introduce new concepts and make a more advanced use of varnishtest as we progress in the book.

Configure Varnish

  • Configure Varnish to use real backends

Varnish DAEMON_OPTS:

-a ${VARNISH_LISTEN_ADDRESS}:${VARNISH_LISTEN_PORT}
-T ${VARNISH_ADMIN_LISTEN_ADDRESS}:${VARNISH_ADMIN_LISTEN_PORT}

See Table 3 and locate the Varnish configuration file for your installation. Open and edit that file to listen on port 80 and have a management interface on port 1234. This is configured with the variable DAEMON_OPTS, and options -a and -T respectively.

In order for changes in the configuration file to take effect, varnishd must be restarted. The safest way to restart Varnish is by using service varnish restart.

The default VCL file location is /etc/varnish/default.vcl. You can change this location by editing the configuration file. The VCL file contains your VCL and backend definitions.

In this book, we use Apache as backend. Before continuing, make sure you have Apache installed and configured to listen in port 8080. See Appendix F: Apache as Backend if you do not know how to do it.

Edit /etc/varnish/default.vcl to use Apache as backend:

backend default {
  .host = "127.0.0.1";
  .port = "8080";
}

To reload your VCL file, run service varnish reload. This command does not restart varnishd, it only reloads the VCL code. The result of your configuration is resumed in Table 4.

* These files are for a SysV Ubuntu/Debian configuration.

You can get an overview over services listening on TCP ports by issuing the command netstat -nlpt. Within the result, you should see something like:

tcp     0     0 0.0.0.0:80         0.0.0.0:*     LISTEN     9223/varnishd
tcp     0     0 127.0.0.1:1234     0.0.0.0:*     LISTEN     9221/varnishd

Note

Varnish 4.0 does not start when enforcing Security-Enhanced Linux (SELinux). A quick workaround, although a bad practice, is to disable SELinux. If you prefer otherwise, then set the boolean varnishd_connect_any variable to 1. You can do that by executing the command sudo setsebool varnishd_connect_any 1. Also, be aware that SELinux defines the ports 6081 and 6082 for varnishd.

Tip

Issue the command man vcl to see all available options to define a backend.

Tip

You can also configure Varnish via the Varnish Administration Console (VAC).

ui/img/vac_config.png

Figure :counter:`figure`: GUI to configure Varnish via the Varnish Administration Console (VAC).

Test Varnish Using Apache as Backend

  • Run http -p Hh localhost

  • Your output should look as:

    # http -p Hh localhost
GET / HTTP/1.1
Accept: */*
Accept-Encoding: gzip, deflate, compress
Host: localhost
User-Agent: HTTPie/0.8.0

HTTP/1.1 200 OK
Accept-Ranges: bytes
Age: 0
Connection: keep-alive
Content-Encoding: gzip
Content-Length: 3256
Content-Type: text/html
Date: Wed, 18 Mar 2015 13:55:28 GMT
ETag: "2cf6-5118f93ad6885-gzip"
Last-Modified: Wed, 18 Mar 2015 12:53:59 GMT
Server: Apache/2.4.7 (Ubuntu)
Vary: Accept-Encoding
Via: 1.1 varnish-plus-v4
X-Varnish: 32770

You can test your Varnish installation by issuing the command http -p Hh localhost. If you see the HTTP response header field Via containing varnish, then your installation is correct.

The X-Varnish HTTP header field contains the Varnish Transaction ID (VXID) of the client request and if applicable, the VXID of the backend transaction that stored in cache the object delivered. X-Varnish is useful to find the correct log entries in the Varnish log. For a cache hit, X-Varnish contains both the ID of the current request and the ID of the request that populated the cache. You will learn more about VXIDs in the Transactions section.

Exercise: Test Apache as Backend with varnishtest

  • Use VTC to test the Server and Via HTTP header fields.
In this exercise you have to define a backend pointing to your Apache server and use assertions with expect.
Solution: Test Apache as Backend with varnishtest

vtc/b00002.vtc

The Management Interface varnishadm

You can use the varnishadm utility to:

  • start and stop the cacher (aka child) process
  • change configuration parameters without restarting Varnish
  • reload the Varnish Configuration Language (VCL) without restarting Varnish
  • view the most up-to-date documentation for parameters
  • varnishadm help and man varnishadm

Varnish offers a management command line interface (CLI) to control a running Varnish instance. This interface implements a list of management commands in the varnishadm utility program. varnishadm establishes a connection to the Varnish daemon varnishd.

To connect to the management interface, issue the command varnishadm. If there are many Varnish instances running in one machine, specify the instance with the -n option. Keep the following in mind when using the management interface:

  1. Changes take effect on the running Varnish daemon instance without need to restart it.
  2. Changes are not persistent across restarts of Varnish. If you change a parameter and you want the change to persist after you restart Varnish, you need to store your changes in the configuration file of the boot script. The location of the configuration file is is in Table 3

varnishadm uses a non-encrypted key stored in a secret file to authenticate and connect to a Varnish daemon. You can now provide access to the interface on a per user basis by adjusting the read permission on the secret file. The location of the secret file is /etc/varnish/secret by default, but you can use the -S option to specify other location. The content of the file is a shared secret, which is a string generated under Varnish installation.

The management interface authenticates with a challenge-response mechanism. Therefore, the shared secret is never transmitted, but a challenge and the response to the challenge. This authentication mechanism offers a reasonably good access control, but it does not protect the data transmitted over the connection. Therefore, it is very important to avoid eavesdroppers like in the man-in-the-middle attack. The simplest way to avoid eavesdroppers is to configure the management interface listening address of varnishd to listen only on localhost (127.0.0.1). You configure this address with the -T option of the varnishd command.

Tip

Varnish provides many on-line reference manuals. To learn more about varnishadm, issue man varnishadm. To check the Varnish CLI manual page, issue man varnish-cli.

Tip

You can also access the varnishadm via the Varnish Administration Console (VAC). To do that, you just have to navigate to the CONFIGURE tab and click on the Varnish server you want to administrate. Then, varnishadm is ready to use in a terminal emulator right in your web browser.

ui/img/vac_varnishadm.png

Figure :counter:`figure`: Access to varnishadm by clicking on the Varnish server that you want to administrate.

ui/img/vac_cli.png

Figure :counter:`figure`: Terminal emulator in your web browser.

More About Varnish Configuration

The location of the configuration file is in Table 3.

Using the service commands is recommended, safe and fast.

Command line options and tunable parameters are used to: 1) define how Varnish should work with operating system and hardware, and 2) set default values. Configuration in VCL defines how Varnish should interact with web servers and clients.

Almost every aspect of Varnish can be reconfigured without restarting Varnish. Notable exceptions are: 1) allocated memory size for caching, 2) cache file location, 3) ownership (for user and group privileges) of the Varnish daemon, and 4) the hashing algorithm.

Some parameters changes require to restart Varnish to take effect. For example, the modification of the listening port. Other changes might not take effect immediately, but restart is not required. Changes to cache time-to-live (TTL), for instance, take effect only after the current cached objects expire. In this example, the value of the TTL parameter is only applicable to caches fetched after the TTL modification.

param.show <parameter> outputs a description of parameter. The description includes when and how modifications takes effect, and the default and current value of the parameter.

There are other ways to reload VCL and make parameter-changes take effect, mostly using the varnishadm tool. However, using the service varnish reload and service varnish restart commands is a good habit.

Note

If you want to know how the service varnish-commands work, look at the script that runs behind the scenes. The script is in /etc/init.d/varnish.

Warning

The varnish script-configuration (located under /etc/default/ or /etc/sysconfig/) is directly sourced as a shell script. Pay close attention to any backslashes (\) and quotation marks that might move around as you edit the DAEMON_OPTS environmental variable.

Command Line Configuration

Relevant options for the course are:

-a <[hostname]:port>
 listening address and port for client requests
-f <filename> VCL file
-p <parameter=value>
 set tunable parameters
-S <secretfile>
 shared secret file for authorizing access to the management interface
-T <hostname:port>
 listening address and port for the management interface
-s <storagetype,options>
 where and how to store objects

All the options that you can pass to the varnishd binary are documented in the varnishd(1) manual page (man varnishd). You may want to take a moment to skim over the options mentioned above.

For Varnish to start, you must specify a backend. You can specify a backend by two means: 1) declare it in a VCL file, or 2) use the -b to declare a backend when starting varnishd.

Though they are not strictly required, you almost always want to specify a -s to select a storage backend, -a to make sure Varnish listens for clients on the port you expect and -T to enable a management interface, often referred to as a telnet interface.

For both -T and -a, you do not need to specify an IP, but can use :80 to tell Varnish to listen to port 80 on all IPs available. Make sure you do not forget the colon, as -a 80 tells Varnish to listen to the IP with the decimal-representation "80", which is almost certainly not what you want. This is a result of the underlying function that accepts this kind of syntax.

You can specify -p for parameters multiple times. The workflow for tuning Varnish parameters usually is that you first try the parameter on a running Varnish through the management interface to find the value you want. Then, you store the parameter and value in a configuration file. This file is read every time you start Varnish.

The -S option specifies a file which contains a secret to be used for authentication. This can be used to authenticate with varnishadm -S as long as varnishadm can read the same secret file -- or rather the same content: The content of the file can be copied to another machine to allow varnishadm to access the management interface remotely.

Note

Varnish requires at least one backend, which is normally specified in the VCL file. The VCL file is passed to varnishd with the -f <filename.vcl> option. However, it is possible to start Varnish without a VCL file. In this case, the backend is passed directly to varnishd with the -b <hostname:port> option. -f and -b are mutually exclusive.

Tip

Type man varnishd to see all options of the Varnish daemon.

Defining a Backend in VCL

/etc/varnish/default.vcl

The above example defines a backend named default, where the name default is not special. Varnish uses the first backend you specify as default. You can specify many backends at the same time, but for now, we will only specify one to get started.

Tip

You can also add and edit your VCL code via the Varnish Administration Console (VAC). This interface also allows you to administrate your VCL files.

ui/img/vac_vcl.png

Figure :counter:`figure`: GUI of Varnish Administration Console (VAC) with command line interface to edit your VCL code.

Exercise: Use the administration interface to learn, review and set Varnish parameters

  1. Use param.show and param.set from varnishadm to see and set the value of default_ttl and default_grace
  2. Refer to Figure 2 to see the object's timeline
You will learn more about how to tune parameters in the Tunable Parameters section.

Setting Parameters in varnishtest

vtc/b00003.vtc

Parameters can also be set in varnishtest. To execute commands via the CLI, you have three options: -cli "command", -cliok "command" and -clierr "status" "command". -cli executes a command without checking the return status. -clickok executes a command and expects it to return OK 200 status. -clickerr executes a command and checks whether the expected return status matches.

Note

You have to instruct varnishtest to assert the expected behavior as much as you can. For example, varnish v1 -cli "param.set default_ttl -1" does not fail because -cli does not assert the return status.

Note

The macro ${bad_ip} translates to 192.0.2.255. This IP address is for test use only, and it is used here because we do not need a backend to set parameters in Varnish. However, we must always declare at least one backend when varnishd is to be started.

Note

Note that we do not start v1, because in this example, we do not need to start the cacher process.

Exercise: Fetch Data Through Varnish

  • Execute http -p hH http://localhost/ on the command line
  • Pay attention to the Age response header field
  • Compare and discuss the results from multiple executions

-p hH specifies HTTPie to print only request and response headers, but not the content. The typical HTTP response is "200 OK" or "404 File not found". Feel free to try removing some of the options and observe the effect. For more information about the HTTPie command, type man http.

Testing Varnish with a web browser can be confusing, because web browsers have their own cache. Therefore, it is useful to double-check web browsers requests with HTTPie or varnishtest. For more information about the Age response header field refer to the Age subsection.

Fetch Data with varnishtest

vtc/b00004.vtc

You can use the delay command in varnishtest. The unit of the command are seconds and it also accepts float numbers. For more information about the Age response header field refer to the Age subsection.

The Age value depends on the time to live (TTL) value of the cached object. We will learn more about it in The Initial Value of beresp.ttl section.

Examining Varnish Server's Output

In this chapter you will learn about:

  • log records,
  • statistics out from global counters and the Varnish log,
  • the log layout,
  • transactions,
  • the query language, and
  • notable counters.

Varnish logs information of requests, caches and responses to The Varnish Shared memory Log (VSL). Logs are available through Varnish tools with a short delay, but usually not noticeable. The VSL is overwritten when filled-up in circular order.

The memory log overwriting has two effects. On the one hand, there is no historic data, but on the other hand, there is an abundance of information accessible at a very high speed. Still, you can instruct Varnish to store logs in files.

Varnish generates very large amounts of data, therefore it does not write logs to disk by default, but only to memory. However, if you need to enable logging to disk, as when debugging a crashing Varnish installation, you set VARNISHNCSA_ENABLED=1 or VARNISHNCSA_ENABLED=1 in /etc/default/varnishlog or /etc/default/varnishncsa respectively. Table 3 shows the location of the configuration file based on different platforms.

Varnish provides specific tools to parse the content of logs: varnishlog, varnishncsa, varnishstat, and varnishstat among others. varnishlog and varnishstat are the two most common used tools.

Tip

All utility programs have installed reference manuals. Use the man command to retrieve their manual pages.

Log Data Tools

Tools to display detailed log records:

  • varnishlog is used to access request-specific data. It provides information about specific clients and requests.
  • varnishncsa displays Varnish access logs in NCSA Common log format.
  • varnishtest allows you to display log records and counters from your tests.

Statistical tools:

  • varnishstat is used to access global counters.
  • varnishtop reads the Varnish log and presents a continuously updated list of the most commonly occurring log entries.
  • varnishhist reads the Varnish log and presents a continuously updated histogram showing the distribution of the last N requests by their processing.

If you have multiple Varnish instances running in the same machine, you need to specify -n <name> both when starting Varnish and when using the tools. This option is used to specify the instance of varnishd, or the location of the shared memory log. All tools (including varnishadm) can also take a -n option.

In this course, we focus on the two most important tools: varnishlog and varnishstat. Unlike all other tools, varnishstat does not read entries from the Varnish log, but from global counters. You can find more details about the other Varnish tools varnishncsa, varnishtop and varnishhist in Appendix B: Varnish Programs.

Log Layout

Figure :counter:`figure`: Log Layout Timeline

Varnish logs transactions chronologically as Figure 7 shows. The varnishlog is one of the most used tools and offers mechanisms to reorder transactions grouped by TCP session, frontend- or backend worker. We talk more on transactions in the next subsection.

The various arguments of varnishlog are mostly designed to help you find exactly what you want, and filter out the noise. On production traffic, the amount of log data that Varnish produces is staggering, and filtering is a requirement for using varnishlog effectively. Next section explains transactions and how to reorder them.

varnishtest starts a real varnishd process for each test, therefore it also logs in VSL. When your Varnish test fails or you run varnishtest in verbose mode, you can see the vsl entry for each Varnish log record. You can also use the logexpect to assert the expected behavior in your tests.

Transactions

$ varnishlog -d

Figure :counter:`figure`: VXIDs in Log Layout

A transaction is one work item in Varnish and it is a set of log lines that belong together, e.g., a client request or a backend request. Varnish Transaction IDs (VXIDs) are applied to lots of different kinds of work items. A unique VXID is assigned to each type of transaction. The 0 VXID is reserved for everything that Varnish does but not part of a specific transaction. You can follow the VXID when you analyze the log through varnishlog or varnishtest.

Transaction types are:

  • session: TCP session
  • request: Transaction handled by the frontend- or backend worker

Varnish logs are grouped by VXID by default. For example, when viewing a log for a simple cache miss, you see logs in the order they end. That is: 1) backend request (BeReq), 2) client request (Request) and 3) session (Session).

Each transaction has a reason, for examples:

  • Client request
  • ESI request
  • restart
  • fetch

Example of Transactions in varnishtest

$varnishtest -v b00001.vtc

(...)
**** v1    0.3 vsl|          0 CLI             - Rd vcl.load "boot" (...)
(...)
**** v1    0.4 vsl|       1000 Begin           c sess 0 HTTP/1
(...)
**** v1    0.4 vsl|       1002 Begin           b bereq 1001 fetch
(...)
**** v1    0.4 vsl|       1002 End             b
**** v1    0.4 vsl|       1001 Begin           c req 1000 rxreq
(...)
**** v1    0.4 vsl|       1001 End             c
(...)
**** v1    0.4 vsl|       1000 End             c
(...)
**** v1    0.5 vsl|          0 CLI             - EOF on CLI connection (...)

Above is a snippet of how Varnish logs are displayed in varnishtest. varnishtest does not group logs by default as varnishlog does. Still, varnishtest allows you to group the transactions for assertions with the command logexpect.

varnishtest starts client transactions in 1000. Note the VXID 0 for Varnish specific records.

Transaction Groups

  • -g <session | request | vxid | raw> groups together transactions in varnishlog and varnishtest

  • Transaction groups are hierarchical

  • Levels are equal to relationships (parents and children):

    Level 1: Client request (cache miss)
  Level 2: Backend request
  Level 2: ESI subrequest (cache miss)
    Level 3: Backend request
    Level 3: Backend request (VCL restart)
    Level 3: ESI subrequest (cache miss)
      Level 4: Backend request
  Level 2: ESI subrequest (cache hit)

When grouping transactions, there is a hierarchy structure showing which transaction initiated what. In client request grouping mode, the various work items are logged together with their originating client request. For example, a client request that triggers a backend request might trigger two more ESI subrequests, which in turn might trigger yet another ESI subrequest.

All these requests together with the client response are arranged in the order they are initiated. This arrangement is easier to grasp than when grouping by VXID. The Content Composition section shows how to analyze the log for Edge Side Includes (ESI) transactions.

When a subrequest occurs in the log, the subrequest tells us about the relationship to its parent request through the Link statement. This statement contains the VXID of the parent request. varnishlog indents its output based on the level of the request, making it easier to see the level of the current request.

Example of Transaction Grouping with varnishlog

$ varnishlog -g request -i Begin,Link -d

Figure :counter:`figure`: Example of Transaction Grouping with varnishlog

Figure 9 shows a client request in a cache miss scenario. In the figure, varnishlog returns records grouped by request. For simplicity, we use the -i option to include only the Begin and Link tags.

For more information about the format and content of all Varnish shared memory logging (VSL) tags, see the VSL man page by typing man vsl. The workflow of Varnish is detailed in the VCL Basics chapter.

To reproduce the example, issue http -p hH http://localhost/, and then the varnishlog command as above. The -d option processes all recorded entries in Varnish log. To learn more about the available varnishlog options, enter varnishlog -h or see the varnishlog man page.

varnishlog accepts all options that are syntactically correct. The output, however, might be different from your first interpretation. Therefore, you should make sure that your results make sense.

Options -b and -c display only transactions coming from the backend and client communication respectively. You can verify the meaning of your results by double checking the filters, and separating your results with the -b and -c options.

Note

The logexpect command from varnishtest accepts the same arguments as varnishlog.

logexpect

  • Assert log records in varnishtest
  • Uses varnishlog API

logexpect is a program that uses the varnishlog API. Therefore, it is able to group and query the Varnishlog just as varnishlog does. In addition, logexpect allows you to assert what you are expecting to appear in VSL.

Note logexpect l1 -wait at the end of the script. Without it, the test would finish successfully without concluding the assert in l1, because varnishtest would not wait for it. -wait instructs the executor of varnishtest to wait until l1 is done.

Below is the synopsis of arguments and options of logexpect:

-v <varnish-instance>
-d <0|1> (head/tail mode)
-g <grouping-mode>
-q <query>

vsl arguments (vsl_arg.c)
-b                   Only display backend records
-c                   Only display client records
-C                   Caseless regular expressions
-i <taglist>         Include tags
-I <[taglist:]regex> Include by regex
-L <limit>           Incomplete transaction limit
-T <seconds>         Transaction end timeout

logexpect lN -v <id> [-g <grouping>] [-d 0|1] [-q query] [vsl arguments] {
   expect <skip> <vxid> <tag> <regex>
}

skip: [uint|*]               Max number of record to skip
vxid: [uint|*|=]             vxid to match
tag:  [tagname|*|=]          Tag to match against
regex:                       regular expression to match against (optional)
*:                           Match anything
=:                           Match value of last successfully matched record

Query Language

  • Operates on transaction groups.
  • Query expression is true if it matches one or more records, false otherwise.
  • Supports:
    • string matching, e.g.: RespProtocol eq "HTTP/1.1"
    • regex, e.g.: ReqMethod ~ "GET|POST"
    • integer and float matching, e.g.: RespStatus == 200
    • boolean operators, e.g.: RespStatus >= 500 and RespStatus < 600
    • parenthesis hierarchy
    • negation using not

Examples of Varnish log queries:

varnishlog -q 'RespStatus < 500'
varnishlog -g request -q 'ReqURL eq "/"'
varnishlog -g request -q 'Backend ~ default'

The -q option allows you to add a query to varnishlog. Think of it as a sort of select filter for varnishlog. It works together with the grouping so that if the query matches some part of any of the work items in the transaction group, the whole group matches and gets displayed.

Query expressions can be combined using boolean functions. In addition, there are many output control options, such as -i taglist. These options are output filters, thus they do not affect transaction matching. Output controls are applied last.

A query expression consists of record selection criteria, and optionally an operator and a value to match against the selected records:

<record selection criteria> <operator> <operand>

The <record selection criteria> determines what kind of records from the transaction group the expression applies to. The syntax is:

{level}taglist:record-prefix[field]

For example:

  • Response time exceeds 1⁄2 second Timestamp:Process[2] > 0.5
  • Client requests connection closed ReqHeader:connection ~ close
  • ESI miss (-g request) {3+}Begin ~ Bereq

Taglists are not case-sensitive, but we recommend you to follow the same format as declared in man vsl.

The grouping and the query log processing all happens in the varnishlog API. This means that other programs using this API automatically get grouping and query language. For example, logexpect as we shall see next.

Tip

man vsl-query shows you more details about query expressions. man vsl lists all taglists and their syntax.

Exercise: Filter Varnish Log Records

  • Use varnishlog to print transactions for Service Unavailable (RespStatus == 503) responses
  • Use varnishtest to provoke a Service Unavailable response and assert it by reading VSL with logexpect

varnishstat

Uptime mgt:   1+23:38:08                                                                Hitrate n:       10      100      438
Uptime child: 1+23:38:08                                                                   avg(n):   0.9967   0.5686   0.3870

NAME                                               CURRENT       CHANGE      AVERAGE       AVG_10      AVG_100     AVG_1000
MAIN.uptime                                           171488         1.00         1.00         1.00         1.00         1.00
MAIN.sess_conn                                          1055         7.98          .           8.35         4.49         2.11
MAIN.client_req                                         1055         7.98          .           8.35         4.49         2.11
MAIN.cache_hit                                          1052         7.98          .           8.35         4.49         2.10
MAIN.cache_miss                                            3         0.00          .           0.00         0.00         0.00
MAIN.backend_conn                                          4         0.00          .           0.00         0.00         0.01
MAIN.backend_toolate                                       3         0.00          .           0.00         0.00         0.01
MAIN.backend_recycle                                       4         0.00          .           0.00         0.00         0.01
MAIN.fetch_length                                          4         0.00          .           0.00         0.00         0.01
MAIN.pools                                                 2         0.00          .           2.00         2.00         2.00
MAIN.threads                                             200         0.00          .         200.00       200.00       200.00
MAIN.threads_created                                     200         0.00          .           0.00         0.00         0.00
MAIN.n_object                                              1         0.00          .           1.00         0.85         0.81
MAIN.n_objectcore                                          3         0.00          .           3.00         2.85         2.81
MAIN.n_objecthead                                          4         0.00          .           4.00         3.89         3.83
MAIN.n_backend                                             1         0.00          .           1.00         1.00         1.00
MAIN.n_expired                                             2         0.00          .           2.00         1.76         1.33
MAIN.s_sess                                             1055         7.98          .           8.35         4.49         2.11
MAIN.s_req                                              1055         7.98          .           8.35         4.49         2.11
MAIN.s_fetch                                               3         0.00          .           0.00         0.00         0.00
MAIN.s_req_hdrbytes                                   122380       926.07          .         968.24       520.74       244.35
MAIN.s_resp_hdrbytes                                  376249      2854.04         2.00      2982.17      1602.59       751.87
MAIN.s_resp_bodybytes                                3435094     25993.71        20.00     27177.59     14616.67      6858.74
MAIN.backend_req                                           4         0.00          .           0.00         0.00         0.01
MAIN.n_vcl                                                 1         0.00          .           0.00         0.00         0.00
MAIN.bans                                                  1         0.00          .           1.00         1.00         1.00
MAIN.n_gunzip                                              4         0.00          .           0.00         0.00         0.01
MGT.uptime                                            171488         1.00         1.00         1.00         1.00         1.00
SMA.s0.c_req                                               8         0.00          .           0.00         0.01         0.01
SMA.s0.c_bytes                                         15968         0.00          .           0.01        18.98        27.33
SMA.s0.c_freed                                         11976         0.00          .           0.00        12.17        18.56
SMA.s0.g_alloc                                             2         0.00          .           2.00         1.70         1.62
SMA.s0.g_bytes                                          3992         0.00          .        3991.87      3398.82      3235.53
SMA.s0.g_space                                     268431464         0.00          .   268431464.13 268432057.18 268432220.47
VBE.default(127.0.0.1,,8080).bereq_hdrbytes              630         0.00          .           0.00         0.70         1.13
VBE.default(127.0.0.1,,8080).beresp_hdrbytes            1128         0.00          .           0.00         1.34         1.93
VBE.default(127.0.0.1,,8080).beresp_bodybytes          13024         0.00          .           0.01        15.48        22.29

MAIN.cache_hit                                                                                                         INFO
Cache hits:
Count of cache hits. A cache hit indicates that an object has been delivered to a client without fetching it from a
backend server.

varnishstat looks only at counters. These counters are easily found in VSL, and are typically polled at reasonable interval to give the impression of real-time updates. Counters, unlike the rest of the log, are not directly mapped to a single request, but represent how many times a specific action has occurred since Varnish started.

varnishstat gives a good representation of the general health of Varnish. Unlike all other tools, varnishstat does not read log entries, but counters that Varnish updates in real-time. It can be used to determine your request rate, memory usage, thread usage, number of failed backend connections, and more. varnishstat gives you information just about anything that is not related to a specific request.

There are over a hundred different counters available. To increase the usefulness of varnishstat, only counters with a value different from 0 is shown by default.

varnishstat can be used interactively, or it can display the current values of all the counters with the -1 option. Both methods allow you to specify specific counters using -f field1 -f field2 .. to limit the list.

In interactive mode, varnishstat has three areas. The top area shows process uptime and hitrate information. The center area shows a list of counter values. The bottom area shows the description of the currently selected counter.

Hitrate n and avg(n) are related, where n is the number intervals. avg(n) measures the cache hit rate within n intervals. The default interval time is one second. You can configure the interval time with the -w option.

Since there is no historical data of counters changes, varnishstat has to compute the average while it is running. Therefore, when you start varnishstat, Hitrate values start at 1, then they increase to 10, 100 and 1000. In the example above, the interval is one second. The hitrate average avg(n) show data for the last 10, 100, and 438 seconds. The average hitrate is 0.9967 (or 99.67%) for the last 10 seconds, 0.5686 for the last 100 seconds and 0.3870 for the last 438 seconds.

In the above example Varnish has served 1055 requests and is currently serving roughly 7.98 requests per second. Some counters do not have "per interval" data, but are gauges with values that increase and decrease. Gauges normally start with a g_ prefix.

Tip

You can also see many parameters in real-time graphs with the Varnish Administration Console (VAC).

ui/img/vac_hit_meter.png

Figure :counter:`figure`: Cache Hit Meter

ui/img/vac_hitrate.png

Figure :counter:`figure`: Hit vs Miss vs Hit for Pass

ui/img/vac_hitrate_req.png

Figure :counter:`figure`: Req/sec, and Hit/sec

ui/img/vac_realtime_counters.png

Figure :counter:`figure`: Real Time Counters

Tip

If you need to collect statistics from more than a single Varnish server, Varnish Custom Statistics (VCS) allows you to do that. In addition, VCS allows you to define your metrics to collect and analyze aggregated statistics, for example:

  • A/B testing
  • Measuring click-through rate
  • Track slow pages and cache misses
  • Analyze what is "hot" right now in a news website
  • Track changes in currency conversions in e-commerce
  • Track changes in Stock Keeping Units (SKUs) behavior in e-commerce
  • Track number of unique consumers of HLS/HDS/DASH video streams

Notable Counters

Varnish provides a large number of counters for information, and debugging purposes. Table 8 presents counters that are typically important. Other counters may be relevant only for Varnish developers when providing support.

Counters also provide feedback to Varnish developers on how Varnish works in production environments. This feedback in turn allows Varnish to be developed according to its real usage. Issue varnishstat -1 to list all counters with their current values.

Note

If you have many backends, consider to increase the size of the shared memory log. For that, see the option -l in the man page of varnishd.

Tip

Remember that Varnish provides many reference manuals. To see all Varnish counter field definitions, issue man varnish-counters.

Exercise: Try varnishstat and varnishlog together

  • Run varnishstat and varnishlog while performing a few requests.
  • See, analyze and understand how counters and parameters change in varnishstat and varnishlog.

Exercise: Assert Counters in varnishtest

  • Write a Varnish test to check the counters for cache misses, cache hits, and number of cached objects.
  • Use cache_miss, cache_hit, and n_object counters respectively.

Solution: Assert Counters in varnishtest

vtc/b00005.vtc

Tuning

This chapter is for the system administration course only

This section covers:

  • Architecture
  • Best practices
  • Parameters

Perhaps the most important aspect of tuning Varnish is writing effective VCL code. For now, however, we will focus on tuning Varnish for your hardware, operating system and network. To be able to do that, knowledge of Varnish architecture is helpful.

It is important to know the internal architecture of Varnish for two reasons. First, the architecture is chiefly responsible for the performance, and second, it influences how you integrate Varnish in your own architecture.

There are several aspects of the design that were unique to Varnish when it was originally implemented. Truly good solutions, regardless of reusing ancient ideas or coming up with something radically different, is the aim of Varnish.

Varnish Architecture

Figure :counter:`figure`: Varnish Architecture

Figure 14 shows a block diagram of the Varnish architecture. The diagram shows the data flow between the principal parts of Varnish.

The main block is the Manager process, which is contained in the varnishd binary program. The task of the Manager process is to delegate tasks, including caching, to child processes. The Manager process ensures that there is always a process for each task. The main driver for these design decisions is security, which is explain at Security barriers in Varnish https://www.varnish-cache.org/docs/trunk/phk/barriers.html.

The Manager's command line interface (CLI) is accessible through: 1) varnishadm as explained in The Management Interface varnishadm section, 2) the Varnish Agent vagent2, or 2) the Varnish Administration Console (VAC) (via vagent2).

The Varnish Agent vagent2 is an open source HTTP REST interface that exposes varnishd services to allow remote control and monitoring. vagent2 offers a web UI as shown in Figure 15, but you can write your own UI since vagent2 is an open interface. Some features of the vagent2 are:

  • VCL uploading, downloading, persisting (storing to disk).
  • parameter viewing, storing (not persisting yet)
  • show/clear of panic messages
  • start/stop/status of varnishd
  • banning
  • varnishstat in JSON format
ui/img/vagent2.png

Figure :counter:`figure`: Varnish Agent's HTML interface; designed to showcase the various features of the Varnish Agent.

For more information about vagent2 and installation instructions, please visit https://github.com/varnish/vagent2.

Varnish Software has a commercial offering of a fully functional web UI called Varnish Administration Console (VAC). For more information about VAC, refer to the Varnish Administration Console (VAC) section.

The Parent Process: The Manager

The Manager process is owned by the root user, and its main functions are:

  • apply configuration changes (from VCL files and parameters)
  • delegate tasks to child processes: the Cacher and the VCL to C Compiler (VCC)
  • monitor Varnish
  • provide a Varnish command line interface (CLI)
  • initialize the child process: the Cacher

The Manager checks every few seconds whether the Cacher is still there. If the Manager does not get a reply within a given interval defined in ping_interval, the Manager kills the Cacher and starts it up again. This automatic restart also happens if the Cacher exits unexpectedly, for example, from a segmentation fault or assert error. You can ping manually the cacher by executing varnishadm ping.

Automatic restart of child processes is a resilience property of Varnish. This property ensures that even if Varnish contains a critical bug that crashes the child, the child starts up again usually within a few seconds. You can toggle this property using the auto_restart parameter.

Note

Even if you do not perceive a lengthy service downtime, you should check whether the Varnish child is being restarted. This is important, because child restarts introduce extra loading time as varnishd is constantly emptying its cache. Automatic restarts are logged into /var/log/syslog.

To verify that the child process is not being restarted, you can also check its lifetime with the MAIN.uptime counter in varnishstat.

Varnish Software and the Varnish community at large occasionally get requests for assistance in performance tuning Varnish that turn out to be crash-issues.

The Child Process: The Cacher

Since the Cacher listens on public IP addresses and known ports, it is exposed to evil clients. Therefore, for security reasons, this child process is owned by an unprivileged user, and it has no backwards communication to its parent, the Manager.

The main functions of the Cacher are:

  • listen to client requests
  • manage worker threads
  • store caches
  • log traffic
  • update counters for statistics

Varnish uses workspaces to reduce the contention between each thread when they need to acquire or modify memory. There are multiple workspaces, but the most important one is the session workspace, which is used to manipulate session data. An example is changing www.example.com to example.com before it is entered into the cache, to reduce the number of duplicates.

It is important to remember that even if you have 5 MB of session workspace and are using 1000 threads, the actual memory usage is not 5 GB. The virtual memory usage will indeed be 5GB, but unless you actually use the memory, this is not a problem. Your memory controller and operating system will keep track of what you actually use.

To communicate with the rest of the system, the child process uses the VSL accessible from the file system. This means that if a thread needs to log something, all it has to do is to grab a lock, write to a memory area and then free the lock. In addition to that, each worker thread has a cache for log-data to reduce lock contention. We will discuss more about the Threading Model later in this chapter.

The log file is usually about 80MB, and split in two. The first part is counters, the second part is request data. To view the actual data, a number of tools exist that parses the VSL.

Since the log-data is not meant to be written to disk in its raw form, Varnish can afford to be very verbose. You then use one of the log-parsing tools to extract the piece of information you want -- either to store it permanently or to monitor Varnish in real-time.

If something goes wrong in the Cacher, it logs a detailed panic message to syslog. For testing, you can induce panic to varnishd by issuing the command varnishadm debug.panic.worker or by pressing the Induce Panic button in the Varnish Agent web interface.

VCL Compilation

  • Command to print VCL code compiled to C language and exit:

    $varnishd -C -f <vcl_filename>

  • Useful to check whether your VCL code compiles correctly.

Configuring the caching policies of Varnish is done in the Varnish Configuration Language (VCL). Your VCL is then translated by the VCC process to C, which is compiled by a normal C compiler – typically gcc, and linked into the running Varnish instance. Since the VCL compilation is done outside of the child process, there is no risk of affecting the running Varnish by accidentally loading an ill-formatted VCL.

As a result of this, changing configuration while running Varnish is very cheap. Policies of the new VCL takes effect immediately. However, objects cached with an older configuration may persist until they have no more old references or the new configuration acts on them.

A compiled VCL file is kept around until you restart Varnish completely, or until you issue vcl.discard from the management interface. You can only discard compiled VCL files after all references to them are gone. You can see the amount of VCL references by reading the parameter vcl.list.

Storage Backends

Varnish supports different methods to allocate space for the cache. You can choose one method with the -s option of varnishd.

  • malloc
  • file
  • persistent (deprecated)
  • Varnish Massive Storage Engine (MSE) in Varnish Plus only

Note

As a rule of thumb use: malloc if it fits in memory, or file otherwise. Expect around 1kB of overhead per object cached.

They approach the same basic problem from different angles. With the -s malloc method, Varnish will request the entire size of the cache with a malloc() (memory allocation) library call. The operating system divides the cache between memory and disk by swapping out what it can't fit in memory.

Another possibility is to use the -s file storage backend. This option creates a file on a filesystem to contain the entire cache. Then, the operating system maps the entire file into memory if possible.

The -s file storage method does not retain data when you stop or restart Varnish! For this purpose, Varnish provides a persistence option -s persistent. The usage of this option, however, is strongly discouraged mainly because of the consistency issues that arise with it.

The Varnish Massive Storage Engine (MSE) is an improved storage backend for Varnish Plus only. Its main improvements are decreased disk IO load and lower storage fragmentation. MSE is designed and tested with storage sizes up to 10 TB.

MSE uses a hybrid of two cache algorithms, least recently used (LRU) and least frequently used (LFU), to manage memory. Benchmarks show that this algorithm outperforms the malloc and file strategies. MSE also implements a mechanism to eliminate internal fragmentation. For more details about its design, implementation and benchmark results, please refer to https://www.varnish-software.com/blog/introducing-varnish-massive-storage-engine.

When choosing storage backend, use malloc if your cache will be contained entirely or mostly in memory. If your cache will exceed the available physical memory, you have two options: file or MSE. We recommend you to use MSE because it performs much better than file storage backend.

It is important to keep in mind that the size you specify with the -s option is the size for the actual cache. Varnish has an overhead on top of this for keeping track of the cache, so the actual memory footprint of Varnish will exceed what the '-s' argument specifies if the cache is full. The current estimate (subject to change on individual Varnish-versions) is that about 1kB of overhead needed for each object. For 1 million objects, that means 1GB extra memory usage.

In addition to the per-object overhead, there is also a fairly static overhead which you can calculate by starting Varnish without any objects. Typically around 100MB.

The Varnish Shared memory Log (VSL)

  • Avoid I/O operations
  • Mount the shared memory log as tmpfs
  • VSL is not persistent

The Varnish Shared memory Log (VSL), sometimes called shm-log or SHMLOG, is used to log most data. VSL operates on a circular buffer. Therefore, there is no a start or an end of it, but you can issue varnishlog -d to see old log entries.

VSL is 80MB large by default, which gives a certain history, but it is not persistent unless you instruct Varnish to do otherwise. To change the size of the VSL, see the option -l in the man page of varnishd.

There is not much you have to do with the VSL, except ensure that it does not cause I/O operations. You can avoid I/O by mounting the VSL as a temporary file storage (tmpfs). This is typically configured in /etc/fstab, and the shm-log is normally kept under /var/lib/varnish/ or equivalent locations. You need to restart varnishd after mounting it as tmpfs.

VSL is not persistent, i.e., all its data is in memory. This memory space is mapped under /var/lib/varnish/.

Warning

Some Varnish distribution setup the file storage backend option -s file by default. Those distribution set a path that puts the storage file in the same directory as the shm-log. We discourage this practice.

Tunable Parameters

  • In the CLI:

    param.show -l

  • Do not fall for the copy/paste tips

  • Test the parameters in CLI, then store them in the configuration file

Varnish has many different parameters which can be adjusted to make Varnish act better under specific workloads or with specific software and hardware setups. They can all be viewed with param.show in the management interface varnishadm.

You can set up parameters in two different ways. In varnishadm, use the command param.set <param> <value>. Alternatively, you can issue the command varnishd -p param=value.

Remember that changes made in the management interface are not persistent. Therefore, unless you store your changes in a startup script, they will be lost when Varnish restarts.

The general advice with regards to parameters is to keep it simple. Most of the defaults are optimal. If you do not have a very specific need, it is generally better to use the default values.

A few hidden debug commands exist in the CLI, which can be revealed with help -d. These commands are meant exclusively for development or testing, and many of them are downright dangerous. They are hidden for a reason, and the only exception is perhaps debug.health, which is somewhat common to use.

Tip

Parameters can also be configured via the Varnish Administration Console (VAC) as shown in the figure below.

ui/img/vac_parameters.png

Figure :counter:`figure`: GUI to configure parameters via the Varnish Administration Console (VAC).

Varnish Tuner

  • Command varnishtuner
  • Suggested values for system variables and Varnish parameters are installation specific
  • With our without user input
  • Available for Varnish Plus only

The biggest potential for improvement is outside Varnish. First and foremost in tuning the network stack and the TCP/IP connection handling.

Varnish Tuner is a program toolkit based on the experience and documentation we have built. The toolkit tries to gather as much information as possible from your installation and decides which parameters need tuning.

The tuning advice that the toolkit gives is specific to that system. The Varnish Tuner gathers information from the system it is running in. Based on that information, it suggests values for systems variables of your OS and parameters for your Varnish installation that can be beneficial to tune. Varnish Tuner includes the following information for each suggested system variable or parameter:

  • current value
  • suggested value
  • text explaining why it is advised to be changed

varnishtuner requires by default user input to produce its output. If you are not sure about the requested input, you can instruct varnishtuner to do not suggest parameters that require user input. For this, you issue varnishtuner -n.

Varnish Tuner is valuable to both experts and non-experts. Varnish Tuner is available for Varnish Plus series only.

Warning

Copying Varnish Tuner suggestions to other systems might not be a good idea.

Varnish Tuner Persistence

The output of varnishtuner updates every time you introduce a new input or execute a suggested command. However, the result of the suggested commands are not necessarily persistent, which means that they do not survive a reboot or restart of Varnish Cache. To make the tuning persistent, you can add do the following:

  • Specify the Varnish parameters in the configuration file.
  • Specify the sysctl system variables in /etc/sysctl.conf or in /etc/sysctl.d/varnishtuner.conf (if /etc/sysctl.d/ is included).

To see the usage documentation of Varnish Tuner, execute: varnishtuner --help.

Install Varnish Tuner

Ubuntu Trusty 14.04

Packages in our repositories are signed and distributed via https. You need to enable https support in the package manager and install our public key first:

apt-get install -y apt-transport-https
curl https://<username>:<password>@repo.varnish-software.com/GPG-key.txt |
apt-key add -

You add the Varnish Plus repository to /etc/apt/sources.list.d/varnish-plus.list:

# Varnish Tuner
deb https://<username>:<password>@repo.varnish-software.com/ubuntu
<distribution_codename> non-free

Where <distribution_codename> is the codename of your Linux distribution, for example: trusty, debian, or wheezy.

Then:

apt-get update
apt-get install varnishtuner

Above are the installation instructions for Ubuntu to get Varnish Tuner from our repositories. Replace the <username> and <password> with the ones of your Varnish Plus subscription. If you do not know them, please send an email to our support email to recover them.

Red Hat Enterprise Linux 6

To install Varnish Plus on RHEL6, put the following lines into /etc/yum.repos.d/varnish-plus.repo:

[varnishtuner]
name=Varnishtuner
baseurl=https://<username>:<password>@repo.varnish-software.com/redhat/ \
varnishtuner/el6
enabled=1
gpgcheck=0

Threading Model

  • The child process runs multiple threads in two thread pools
  • Threads accept new connections and delegate them
  • One worker threads per client request – it's common to use hundreds of worker threads
  • Expire-threads evict old content from the cache
The child process runs multiple threads in two thread pools. The threads of these pools are called worker threads. Table 9 presents relevant threads.

Threading Parameters

  • Thread pools can safely be ignored
  • Start threads better sooner than later
  • Maximum and minimum values are per thread pool

When tuning Varnish, think about the expected traffic. The most important thread setting is the number of cache-worker threads. You may configure thread_pool_min and thread_pool_max. These parameters are per thread pool.

Although Varnish threading model allows you to use multiple thread pools, we recommend you to do not modify this parameter. Based on our experience and tests, we have seen that 2 thread pools are enough. In other words, the performance of Varnish does not increase when adding more than 2 pools.

Note

If you run across the tuning advice that suggests to have a thread pool per CPU core, rest assured that this is old advice. We recommend to have at most 2 thread pools, but you may increase the number of threads per pool.

Details of Threading Parameters

  • Default values have proved to be sufficient in most cases
  • thread_pool_min, thread_pool_max and thread_pools are the most common threading parameters to tune.
  • Run extra threads to avoid creating them on demand

Varnish runs one thread per session, so the maximum number of threads is equal to the number of maximum sessions that Varnish can serve concurrently. If you seem to need more threads than the default, it is very likely that there is something wrong in your setup. Therefore, you should investigate elsewhere before you increase the maximum value.

You can observe if the default values are enough by looking at MAIN.sess_queued through varnishstat. Look at the counter over time, because it is fairly static right after startup.

When tuning the number of threads, thread_pool_min and thread_pool_max are the most important parameters. Values of these parameters are per thread pool. The thread_pools parameter is mainly used to calculate the total number of threads. For the sake of keeping things simple, the current best practice is to leave thread_pools at the default |def_thread_pools|.

Varnish operates with multiple pools of threads. When a connection is accepted, the connection is delegated to one of these thread pools. Afterwards, the thread pool either delegates the connection request to an available thread, queue the request otherwise, or drop the connection if the queue is full. By default, Varnish uses 2 thread pools, and this has proven sufficient for even the most busy Varnish server.

Varnish has the ability to spawn new worker threads on demand, and remove them once the load is reduced. This is mainly intended for traffic spikes. It's a better approach to keep a few threads idle during regular traffic, than to run on a minimum amount of threads and constantly spawn and destroy threads as demand changes. As long as you are on a 64-bit system, the cost of running a few hundred threads extra is very low.

The thread_pool_min parameter defines how many threads run for each thread pool even when there is no load. thread_pool_max defines the maximum amount of threads that could be used per thread pool. That means that with the minimum defaults |def_thread_pool_min| and |def_thread_pool_max| of minimum and maximums threads per pool respectively, you have:

Warning

New threads use preallocated workspace, which should be enough for the required task. If threads have not enough workspace, the child process is unable to process the task and it terminates. To avoid this situation, evaluate your setup and consider to increase the workspace_client or workspace_backend parameter.

Time Overhead per Thread Creation

  • thread_pool_add_delay: Wait at least this long after creating a thread.
  • thread_pool_timeout: Thread idle threshold.
  • thread_pool_fail_delay: After a failed thread creation, wait at least this long before trying to create another thread.

Varnish can use several thousand threads, and has had this capability from the very beginning. However, not all operating system kernels were prepared to deal with this capability. Therefore the parameter thread_pool_add_delay was added to ensure that there is a small delay between each thread that spawns. As operating systems have matured, this has become less important and the default value of thread_pool_add_delay has been reduced dramatically, from 20 ms to 2 ms.

There are a few, less important parameters related to thread timing. The thread_pool_timeout is how long a thread is kept around when there is no work for it before it is removed. This only applies if you have more threads than the minimum, and is rarely changed.

Another less important parameter is the thread_pool_fail_delay. After the operating system fails to create a new thread, thread_pool_fail_delay defines how long to wait for a re-trial.

System Parameters

As Varnish has matured, fewer and fewer parameters require tuning. The workspace_client and workspace_backend are parameters that could still be relevant.

  • workspace_client – incoming HTTP header workspace from the client
  • workspace_backend – bytes of HTTP protocol workspace for backend HTTP req/resp
  • ESI typically requires exponential growth
  • Remember: it is virtual, not physical memory

Workspaces are some of the things you can change with parameters. Sometimes you may have to increase them to avoid running out of workspace.

The workspace_client parameter states how much memory can be allocated for each HTTP session. This space is used for tasks like string manipulation of incoming headers. The workspace_backend parameter indicates how much memory can be allocated to modify objects returned from the backend. After an object is modified, its exact size is allocated and the object is stored read-only.

As most of the parameters can be left unchanged, we will not go through all of them. You can take a look at the list of parameter by issuing varnishadm param.show -l to get information about what they can do.

Timers

The timeout-parameters are generally set to pretty good defaults, but you might have to adjust them for unusual applications. The default value of connect_timeout is |def_connect_timeout|. This value is more than enough when having the Varnish server and the backend in the same server room. Consider to increase the connect_timeout value if your Varnish server and backend have a higher network latency.

Keep in mind that the session timeout affects how long sessions are kept around, which in turn affects file descriptors left open. It is not wise to increase the session timeout without taking this into consideration.

The cli_timeout is how long the management thread waits for the worker thread to reply before it assumes it is dead, kills it and starts it back up. The default value seems to do the trick for most users today.

Warning

If connect_timeout is set too high, it does not let Varnish handle errors gracefully.

Note

Another use-case for increasing connect_timeout occurs when virtual machines are involved as they can increase the connection time significantly.

Exercise: Tune first_byte_timeout

  • Set first_byte_timeout to 1 seconds.
  • Check how Varnish times out the request to the backend.
You can solve this exercise either by interacting with a real backend, or simulating it with varnishtest. To check how first_byte_timeout impacts the behavior of Varnish, analyze varnishlog and varnishstat. Again, you can do that by executing them in shell or by reading and asserting VSL and counters in varnishtest.

Solution: Tune first_byte_timeout and test it against your real backend

  • Create a CGI script in /usr/lib/cgi-bin/test.cgi containing:

    #! /bin/sh
sleep 5
echo "Content-type: text/plain"
echo "Cache-control: max-age=0"
echo
echo "Hello world"
date

  • Make it executable.

  • Test that your CGI works without involving Varnish by issuing http localhost:8080/cgi-bin/test.cgi

  • Test your CGI through Varnish by issuing http localhost:80/cgi-bin/test.cgi

In this solution, we have used HTTPie to send a request to the backend, but you can also test your real backend with varnishtest.

Tip

Remember to enable the CGI module in Apache. One way to do that is by issuing the commands: a2enmod cgid, and then service apache2 restart.

Solution: Tune first_byte_timeout and test it against mock-up server

vtc/b00006.vtc

In this example, we introduce -vcl+backend and feature in VTC. -vcl+backend is one way to pass inline VCL code and backend to v1. In this example, v1 receives no inline VCL injects declaration of the backend s1. Thus, -vcl+backend{} is equivalent to -arg "-b ${s1_addr}:${s1_port}" in this case.

feature checks for features to be present in the test environment. If the feature is not present, the test is skipped. SO_RCVTIMEO_WORKS checks for the socket option SO_RCVTIMEO before executing the test.

b00006.vtc is copied from Varnish-Cache/bin/varnishtest/tests/b00023.vtc We advise you to take a look at the many tests under Varnish-Cache/bin/varnishtest/tests/. You will learn so much about Varnish when analyzing them.

Exercise: Configure Threading

  • Change the thread_pool_min and thread_pool_max parameters to get 10 threads running at any given time, but never more than 15.
  • Execute varnishadm param.show <parameter> to see parameter details.
These exercises are for educational purposes, and not intended as an encouragement to change the values. You can learn from this exercise by using varnishstat, varnishadm and varnishstat

Solution: Configure Threading with varnishadm and varnishstat

  • Use varnishadm param.set to set the value of thread_pool_min and thread_pool_max.
  • Monitor the MAIN.threads counter in varnishstat to see how many threads are running while performing this exercise.

Solution: Configure Threading with varnishtest

c00001.vtc

The test above shows you how to set parameters in two ways; passing the argument -p to varnishd or calling param.set. -p vsl_mask=+WorkThread is used to turn on WorkThread debug logging.

The test proves that varnishd starts with the number of threads indicated in thread_pool_min. Changes in thread_pool_min and thread_pool_max are applied by the thread herder, which handles the thread pools and adds or removes threads if necessary. To learn more about other maintenance threads see https://www.varnish-cache.org/trac/wiki/VarnishInternals.

c00001.vtc is a simplified version of Varnish-Cache/bin/varnishtest/tests/r01490.vtc.

HTTP

This chapter is for the web-developer course only

This chapter covers:

  • Protocol basics
  • Requests and responses
  • HTTP request/response control flow
  • Statelessness and idempotence
  • Cache related headers
Varnish is designed to be used with HTTP semantics. These semantics are specified in the version called HTTP/1.1. This chapter covers the basics of HTTP as a protocol, its semantics and the caching header fields most commonly used.

Protocol Basics

ui/img/httprequestflow.png

Figure :counter:`figure`: HTTP request/response control flow diagram

  • Hyper-Text Transfer Protocol, HTTP, is at the core of the web
  • Specified by the IETF, there are two main versions: HTTP/1.1 and HTTP/2
  • Varnish 4.0 supports HTTP/1.1

HTTP is a networking protocol for distributed systems. It is the foundation of data communication for the web. The development of this standard is done by the IETF and the W3C. In 2014, RFCs 723X were published to clarify HTTP/1.1 and they obsolete RFC 2616.

A new version of HTTP called HTTP/2 has been released under RFC 7540. HTTP/2 is an alternative to HTTP/1.1 and does not obsolete the HTTP/1.1 message syntax. HTTP's existing semantics remain unchanged.

The protocol allows multiple requests to be sent in serial mode over a single connection. If a client wants to fetch resources in parallel, it must open multiple connections.

Resources and Representations

  • Resource: target of an HTTP request
  • A resource may have different representations
  • Representation: a particular instantiation of a resource
  • A representation may have different states: past, current and desired

Each resource is identified by a Uniform Resource Identifier (URI), as described in Section 2.7 of [RFC7230]. A resource can be anything and such a thing can have different representations. A representation is an instantiation of a resource. An origin server, a.k.a. backend, produces this instantiation based on a list of request field headers, e.g., User-Agent and Accept-encoding.

When an origin server produces different representations of one resource, it includes a Vary response header field. This response header field is used by Varnish to differentiate between resource variations. More details on this are in the Vary subsection.

An origin server might include metadata to reflect the state of a representation. This metadata is contained in the validator header fields ETag and Last-Modified.

In order to construct a response for a client request, an algorithm is used to evaluate and select one representation with a particular state. This algorithm is implemented in Varnish and you can customize it in your VCL code. Once a representation is selected, the payload for a 200 (OK) or 304 (Not Modified) response can be constructed.

Requests and Responses

  • A request is a message from a client to a server that includes the method to be applied to a requested resource, the identifier of the resource, the protocol version in use and an optional message body
  • A method is a token that indicates the method to be performed on a URI
  • Standard request methods are: GET, POST, HEAD, OPTIONS, PUT, DELETE, TRACE, or CONNECT
  • Examples of URIs are /img/image.png or /index.html
  • Header fields are allowed in requests and responses
  • Header fields allow client and servers to pass additional information
  • A response is a message from a server to a client that consists of a response status, headers and an optional message body

The first line of a request message is called Request-Line, whereas the first line of a response message is called Status-Line. The Request-Line begins with a method token, followed by the requested resource (URI) and the protocol version. .. Methods

A request method informs the web server what sort of request this is: Is the client trying to fetch a resource (GET), update some data (POST) at the backend, or just get the headers of a resource (HEAD)? Methods are case-sensitive.

After the Request-Line, request messages may have an arbitrary number of header fields. For example: Accept-Language, Cookie, Host and User-Agent.

Message bodies are optional but they must comply to the requested method. For instance, a GET request should not contain a request body, but a POST request may contain one. Similarly, a web server cannot attach a message body to the response of a HEAD request.

The Status-Line of a response message consists of the protocol version followed by a numeric status code and its associated textual phrase. This associated textual phrase is also called reason. Important is to know that the reason is intended for the human user. That means that the client is not required to examine the reason, as it may change and it should not affect the protocol. Examples of status codes with their reasons are: 200 OK, 404 File Not Found and 304 Not Modified.

Responses also include header fields after the Status-Line, which allow the server to pass additional information about the response. Examples of response header fields are: Age, ETag, Cache-Control and Content-Length.

Note

Requests and responses share the same syntax for headers and message body, but some headers are request- or response-specific.

Request Example

GET / HTTP/1.1
Host: localhost
User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; fr; rv:1.9.2.16) \
Gecko/20110319 Firefox/3.6.16
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: fr,fr-fr;q=0.8,en-us;q=0.5,en;q=0.3
Accept-Encoding: gzip,deflate
Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
Keep-Alive: 115
Connection: keep-alive
Cache-Control: max-age=0

The above example is a typical HTTP request that includes a Request-Line, and headers, but no message body. The Request-Line consists of the GET request for the / resource and the HTTP/1.1 version. The request includes the header fields Host, User-Agent, Accept, Accept-Language, Accept-Encoding, Accept-Charset, Keep-Alive, Connection and Cache-Control.

Note that the Host header contains the hostname as seen by the browser. The above request was generated by entering http://localhost/ in the browser. Most browsers automatically add a number of headers.

Some of the headers will vary depending on the configuration and state of the client. For example, language settings, cached content, forced refresh, etc. Whether the server honors these headers will depend on both the server in question and the specific header.

The following is an example of an HTTP request using the POST method, which includes a message body:

POST /accounts/ServiceLoginAuth HTTP/1.1
Host: www.google.com
User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; fr; rv:1.9.2.16) \
Gecko/20110319 Firefox/3.6.16
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: fr,fr-fr;q=0.8,en-us;q=0.5,en;q=0.3
Accept-Encoding: gzip,deflate
Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
Keep-Alive: 115
Connection: keep-alive
Referer: https://www.google.com/accounts/ServiceLogin
Cookie: GoogleAccountsLocale_session=en;[...]
Content-Type: application/x-www-form-urlencoded
Content-Length: 288

ltmpl=default[...]&signIn=Sign+in&asts=

Response Example

HTTP/1.1 200 OK
Server: Apache/2.2.14 (Ubuntu)
X-Powered-By: PHP/5.3.2-1ubuntu4.7
Cache-Control: public, max-age=86400
Last-Modified: Mon, 04 Apr 2011 04:13:41 +0000
Expires: Sun, 11 Mar 1984 12:00:00 GMT
Vary: Cookie,Accept-Encoding
ETag: "1301890421"
Content-Type: text/html; charset=utf-8
Content-Length: 23562
Date: Mon, 04 Apr 2011 09:02:26 GMT
X-Varnish: 1886109724 1886107902
Age: 17324
Via: 1.1 varnish
Connection: keep-alive

[data]

The example above is an HTTP response that contains a Status-Line, headers and message body. The Status-Line consists of the HTTP/1.1 version, the status code 200 and the reason OK. The response status code informs the client (browser) whether the server understood the request and how it replied to it. These codes are fully defined in https://tools.ietf.org/html/rfc2616#section-10, but here is an overview of them:

  • 1xx: Informational – Request received, continuing process
  • 2xx: Success – The action was successfully received, understood, and accepted
  • 3xx: Redirection – Further action must be taken in order to complete the request
  • 4xx: Client Error – The request contains bad syntax or cannot be fulfilled
  • 5xx: Server Error – The server failed to fulfill an apparently valid request

HTTP Properties

  • HTTP is a stateless protocol
  • Properties of methods: safe, idempotent and cacheable
  • Most common cacheable request methods are GET and HEAD

HTTP is by definition a stateless protocol meaning that each request message can be understood in isolation. Hence, a server MUST NOT assume that two requests on the same connection are from the same user agent unless the connection is secured and specific to that agent.

HTTP/1.1 persists connections by default. This is contrary to most implementations of HTTP/1.0, where each connection is established by the client prior to the request and closed by the server after sending the response. Therefore, for compatibility reasons, persistent connections may be explicitly negotiated as they are not the default behavior in HTTP/1.0 [https://tools.ietf.org/html/rfc7230#appendix-A.1.2]. In practice, there is a header called Keep-Alive you may use if you want to control the connection persistence between the client and the server.

Safe methods are considered "safe" if they are read-only; i.e., the client request does not alter any state on the server. GET, HEAD, OPTIONS, and TRACE methods are defined to be safe. An idempotent method is such that multiple identical requests have the same effect as a single request. PUT, DELETE and safe requests methods are idempotent. Cacheable methods are those that allow to store their responses for future reuse. RFC7231 specifies GET, HEAD and POST as cacheable. However, responses from POST are very rarely treated as cacheable. [https://tools.ietf.org/html/rfc7231#section-4.2]

Cache-related Headers Fields

  • HTTP provides a list of headers fields to control cache behavior

  • The most important caching header fields are:

    • Expires
    • Cache-Control
    • Etag
    • Last-Modified
    • If-Modified-Since
    • If-None-Match
    • Vary
    • Age

A cached object is a local store of HTTP response messages. These objects are stored, controlled, retrieved and deleted by a subsystem, in this case Varnish. For this purpose, Varnish uses the caching header fields defined in https://tools.ietf.org/html/rfc7232 and https://tools.ietf.org/html/rfc7234.

If a matched cache is valid, Varnish constructs responses from caches. As a result, the origin server is freed from creating and transmitting identical response bodies.

Constructing Responses from Caches

When to serve a cached object?

  • The cached object is properly matched: cache-hit
  • The requested method and its matched object allows it
  • The freshness of the cached object is acceptable

When Varnish matches a request with a cached object (aka cache-hit), it evaluates whether the cache or origin server should be used to construct the response. There are many rules that should be taken into consideration when validating a cache. Most of those rules use caching header fields.

This subsection describes first the concept of cache-hit and cache-miss. After that, it describes three header fields commonly used to effectively match caches. These fields are Vary, Etag and If-None-Match.

Cache Matching

ui/img/httpcachehit.png

Figure :counter:`figure`: Cache-hit control flow diagram

  • Cache-hits are used to reuse, update or invalidate caches
  • Objects may have variants (Vary and Etag)
ui/img/httpcachemiss.png

Figure :counter:`figure`: Cache-miss control flow diagram

Figure 18 shows the flow diagram of a cache-hit. A cache-hit occurs when the requested object (URI) matches a stored HTTP response message (cache). If the matched stored message is valid to construct a response for the client, Varnish serves construct a response and serves it without contacting the origin server.

Figure 19 shows the flow diagram of a cache-miss. A cache-miss happens when Varnish does not match a cache. In this case, Varnish forwards the request to the origin server.

Vary

  • Selects a representation of a resource
  • Be careful when using Vary
  • Wrong usage can create a very large number of cached objects and reduce efficiency

If the origin server sends Vary in a response, Varnish does not use this response to satisfy a later request unless the later request has the same values for the listed fields in Vary as the original request. As a consequence, Vary expands the cache key required to match a new request to the stored cache entry.

Vary is one of the trickiest headers to deal with when caching. A caching server like Varnish does not necessarily understand the semantics of a header, or what part triggers different variants of a response. In other words, an inappropriate use of Vary might create a very large number of cached objects, and reduce the efficiency of your cache server. Therefore, you must be extremely cautious when using Vary.

Caching objects taking into consideration all differences from requesters creates a very fine-grained caching policy. This practice is not recommended, because those cached objects are most likely retrieved only by their original requester. Thus, fine-grained caching strategies do not scale well. This is a common mistake if Vary is not used carefully.

An example of wrong usage of Vary is setting Vary: User-Agent. This tells Varnish that for absolutely any difference in User-Agent, the response from the origin server might look different. This is not optimal because there are probably thousands of User-Agent strings out there.

Another example of bad usage is when using only Vary: Cookie to differentiate a response. Again, there could be a very large number of cookies and hence a very large number of cached objects, which are going to be retrieved most likely only by their original requesters.

The most common usage of Vary is Vary: Accept-Encoding, which tells Varnish that the content might look different depending on the request Accept-Encoding header. For example, a web page can be delivered compressed or uncompressed depending on the client. For more details on how to use Vary for compressions, refer to https://www.varnish-cache.org/docs/trunk/users-guide/compression.html.

One way to assist Vary is by building the response body from cached and non-cached objects. We will discuss this further in the Content Composition chapter.

Note

Varnish can handle Accept-Encoding and Vary: Accept-Encoding, because Varnish has support for gzip compression.

Understanding Vary in varnishtest

vtc/c00002.vtc

In c00002.vtc, c1 requests /same-url three times. Since the backend s1 returns Vary: Foobar, Varnish maps the cached object to both req.url and http.foobar. Therefore, the second request misses the cached object and fetches from s1 a new variation mapped to Foobar: 2.

The third request from c1 matches both req.url and http.foobar values from the first request. Thus, this request does not trigger a backend request.

Recall that X-Varnish contains the transaction ID of the client request and if applicable, the ID of the backend transaction that stored the object delivered. You can see this behavior in the third request. That also means that the VXID counting does not increase to 1006 in the third client request.

If your backend returns Vary:, it must also handle situations when clients do not send the request header to identify a variation. For example, when c1 does not send Footbar::

txreq -url "/same-url"
rxresp

your backend should handle the lack of that header field specifically. You can test it as the following assertion shows:

rxreq
expect req.http.foobar == <undef>
txresp -hdr "Vary: Foobar" -hdr "Snafu: 3" -body "3333\n"

Be aware that the lack of a header field sent by a client is not the same as sending the field with an empty value. Therefore, requests like:

txreq -hdr "Foobar:  "

should be handled in your backend specifically. You can test it as:

rxreq
expect req.http.foobar == ""
txresp -hdr "Vary: Foobar" -hdr "Snafu: 4" -body "4444\n"

c00002.vtc is a modified version for teaching purposes from Varnish-Cache/bin/varnishtest/tests/c00004.vtc. We advise you to look at the many tests included in Varnish-Cache.

Next we cover four important header fields used in conditional requests. Two validator fields: ETag and Last-Modified; and two precondition header fields: If-None-Match and If-Modified-Since.

ETag

  • An Entity Tag (ETag) is metadata to differentiate between multiple states of a resource's representation
  • A differentiator key of presentations in addition to Vary
  • An ETag is a validator header field
  • Response header field

Origin servers normally add metadata to further describe the representation of a resource. This metadata is used in conditional requests. "Conditional requests are HTTP requests that include one or more header fields indicating a precondition to be tested before applying the method semantics to the target resource" [RFC7232]. ETag is a validator header field.

The ETag header field provides a state identifier for the requested variant (resource representation). ETags are used to differentiate between multiple states based on changes over time of a representation. In addition, ETags are also used differentiate between multiple representations based on content negotiation regardless their state.

Example of an ETag header:

ETag: "1edec-3e3073913b100"

The response ETag field is validated against the request If-None-Match header field. We will see the details of If-None-Match later in this subsection, but before, we learn about the other validator header field: Last-Modified.

Last-Modified

  • Time-based state of presentations
  • Validator header field
  • Response header field

The Last-Modified response header field is a timestamp that indicates when the variant was last modified. This headers may be used in conjunction with If-Modified-Since and If-None-Match.

Example of a Last-Modified header:

Last-Modified: Wed, 01 Sep 2004 13:24:52 GMT

ETag and Last-Modified are validator header fields, which help to differentiate between representations. Normally, origin servers send both fields in successful responses. Whether you use one, another or both, depends on your use cases. Please refer to Section 2.4 in https://tools.ietf.org/html/rfc7232#section-2.4 for a full description on when to use either of them.

If-None-Match

  • Precondition Header Field
  • Request header field
  • Validates local caches against ETag

A client that has obtained a response message and stored it locally, may reuse the obtained ETag value in future requests to validate its local cache against the selected cache in Varnish. The obtained value from an ETag is sent from the client to Varnish in the request If-None-Match header field. In fact, a client may have stored multiple resource representations and therefore a client may send an If-None-Match field with multiple ETag values to validate.

The purpose of this header field is to reuse local caches without compromising its validity. If the local cache is valid, Varnish replies with a 304 (Not Modified) response, which does not include a message body. In this case, the client reuses its local cache to construct the requested resource.

Example of an If-None-Match header:

If-None-Match: "1edec-3e3073913b100"
ui/img/httpifnonematch.png

Figure :counter:`figure`: If-None-Match control diagram.

If-Modified-Since

  • Validates local caches by modification date
  • Precondition Header Field
  • Request header field

A request containing an If-Modified-Since header field indicates that the client wants to validate one or more of its local caches by modification date. If the requested representation has not been modified since the time specified in this field, Varnish returns a 304 (not modified) response. A 304 response does not contain message body. This behavior is similar to as when using If-None-Match.

Example of an If-Modified-Since header:

If-Modified-Since: Wed, 01 Sep 2004 13:24:52 GMT
ui/img/httpifmodifiedsince.png

Figure :counter:`figure`: If-Modified-Since control flow diagram.

Understanding Last-Modified and If-Modified-Since in varnishtest

vtc/b00007.vtc

The example above is a modified version of Varnish-Cache/bin/varnishtest/tests/b00039.vtc and it shows the usage of Last-Modified and If-Modified-Since header fields. The example introduces how to insert VCL code in varnishtest:

sub vcl_backend_response {
   set beresp.ttl = 2s;
   set beresp.grace = 5s;
   # beresp.was_304 is ``true`` if the response from the backend was
   # a positive result of a conditional fetch (``304 Not Modified``).
   set beresp.http.was-304 = beresp.was_304;
}

You will learn all details about VCL in the following sections, but for now it is enough to understand that this code sets the time to live TTL and grace time of cached objects to 2 and 5 seconds respectively. Recall the object lifetime from Figure 2 to understand the expected behavior.

The code also adds a HTTP response header field was-304 with the boolean value of the beresp.was_304. This variable is set to true if the response from the backend was a positive result of a conditional fetch (304 Not Modified).

We hope that this exercise motivates you to use varnishtest when designing your cache policies. As you can see, varnishtest is very precise when testing caching objects against different time settings.

Note

beresp.was_304 is a variable available in Varnish 4.1

Allowance

  • How to control which caches can be served?
  • Cache-Control and Pragma (for backwards compatibility only)
Varnish allows you to validate whether the stored response (cache) can be reused or not. Validation can be done by checking whether the presented request does not contain the no-cache directive. This subsection reviews two common header fields, Cache-Control and Pragma, to check caching allowance.

Cache-Control

The Cache-Control header field specifies directives that must be applied by all caching mechanisms (from proxy cache to browser cache).

Table 12 summarizes the directives you may use for each context. The most relevant directives of Cache-Control are:

  • public: The response may be cached by any cache.
  • no-store: The response body must not be stored by any cache mechanism.
  • no-cache: Authorizes a cache mechanism to store the response in its cache but it must not reuse it without validating it with the origin server first.
    In order to avoid any confusion with this argument think of it as a "store-but-do-no-serve-from-cache-without-revalidation" instruction.
  • max-age: Specifies the period in seconds during which the cache is considered fresh.
  • s-maxage: Like max-age but it applies only to public caches.
  • must-revalidate: Indicates that a stale cache item can not be served without revalidation with the origin server first.

Example of a Cache-Control header:

Cache-Control: public, must-revalidate, max-age=2592000

Note

Cache-Control always overrides Expires.

Note

By default, Varnish does not care about the Cache-Control request header. If you want to let users update the cache via a force refresh you need to do it yourself.

Understanding Cache-Control in varnishtest

b00009.vtc

The example above shows how Cache-control: max-age=3 overwrites TTL for cached objects. The default TTL is 120 seconds, but we set it here to 1 just to explicitly show that the cached object is not expired after a delay of 2 seconds, because max-age=3. Therefore, the second assert:

expect resp.bodylen == 3

is 3 but not 6 (size of FOOBAR).

If you are curious, you can remove:

-hdr "Cache-control: max-age=3"

from the first txresp, and you you will that the second request will contain a body length of 5.

Tip

Take a look at b00941.vtc, b00956.vtc and r01578.vtc in Varnish-Cache/bin/varnishtest/tests/ to learn more.

Pragma

  • Only for legacy
  • Treat Pragma: no-cache as Cache-Control: no-cache
The Pragma request header is a legacy header and should no longer be used. Some applications still send headers like Pragma: no-cache but this is for backwards compatibility reasons only. Any proxy cache should treat Pragma: no-cache as Cache-Control: no-cache, and should not be seen as a reliable header especially when used as a response header.

Freshness

  • Fresh object: age has not yet exceeded its freshness lifetime
  • Stale object: has exceeded its freshness lifetime, i.e., expired object

When reusing a stored response (cached object), you should always check its freshness and evaluate whether to deliver expired objects or not. A response's freshness lifetime is the length of time between its generation by the origin server and its expiration time. A stale (expired) object can also be reused, but only after further validation with the origin server.

As defined in RFC7234 [https://tools.ietf.org/html/rfc7234#section-4.2]:

A response's age is the time that has passed since it was generated by, or successfully validated with, the origin server. The primary mechanism for determining freshness is for an origin server to provide an explicit expiration time in the future, using either the ``Expires`` header field or the ``max-age`` response directive.

Age

  • Response header field calculated at the cache server, i.e., Varnish
  • Varnish send an additional response header field, Age, to indicate the age of the cache
  • Clients (and Varnish) will use the Age header field to determine the freshness of a cache
  • max-age-based equation: cache duration = max-age - Age
  • Age can be used to disallow caches at the client side

Consider what happens if you let Varnish cache content for a week. If Varnish does not calculate the age of a cached object, Varnish might happily inform clients that the content is fresh, but the cache could be older than the maximum allowed max-age. By age we mean an estimate amount of time since the response was generated or successfully validated at the origin server.

Client browsers calculate a cache duration based on the Age header field and the max-age directive from Cache-Control. If this calculation results in a negative number, the browser does not cache the response locally. Negative cache duration times, however, do not prevent browsers from using the received object. Varnish does the same, if you put one Varnish server in front of another.

Exercise: Use article.php to test Age
  1. Copy article.php from Varnish-Book/material/webdev/ to /var/www/html/
  2. Send a request to article.php via Varnish, and see the response Age header field in varnishlog -g request -i ReqHeader,RespHeader
  3. Click the link several times and refresh your browser. Can you identify patterns?
  4. Analyze the output of varnishstat -f MAIN.client_req -f MAIN.cache_hit in addition to varnishlog
  5. Can you use the Age field to determine whether Varnish made a cache hit?
  6. What is the difference between caching time in Varnish and the client?
  7. Use different browsers and analyze whether their internal caches work different

You might encounter that different browsers have different behaviors. Some of them might cache content locally, and their behavior when refreshing might be different, which can be very confusing. This just highlights that Varnish is not the only part of your web-stack that parses and honors cache-related headers. There might also be other caches along the way which you do not control, like a company-wide proxy server.

Since browsers might interpret cache headers differently, it is a good idea to control them in your cache server. In the next chapters, you will learn how to modify the response headers Varnish sends. This also allows your origin server to emit response headers that should be seen and used by Varnish only, not in your browser.

When browsers decide to load a resource from their local cache, requests are never sent. Therefore this exercise and these type of tests are not possible to be simulated in varnishtest.

Expires

  • Used to stale objects
  • Response header field only

The Expires response header field gives the time after which the response is considered stale. Normally, a stale cache item should not be returned by any cache (proxy cache or client cache). The syntax for this header is:

Expires: GMT formatted date

It is recommended not to define Expires too far in the future. Setting it to 1 year is usually enough. The use of Expires does not prevent the cached object from being updated. For example, if the name of the resource is updated.

Expires and Cache-Control do more or less the same job, but Cache-Control gives you more control. The most significant differences between these two headers is:

  • Cache-Control uses relative times in seconds
  • Cache-Control is both a request and a response header field.
  • Expires always returns an absolute time
  • Expires is a response header field only

Expires works best for files that are part of a website design like JavaScripts stylesheets or images.

Understanding Expires in varnishtest

b00008.vtc

In Varnish, an expired object is an object that has exceeded the TTL + grace + keep time. In the example above, the Expires header field sets TTL to 1, and changes default_grace from 10 to 2. default_keep is already 0, but we show it explicitly anyway.

Tip

Take a look at s00000.vtc and s00001.vtc in Varnish-Cache/bin/varnishtest/tests/.

Tip

To get more information about n_expire, issue man varnish-counters.

Availability of Header Fields

Exercise: Test Various Cache Headers Fields with a Real Browser

Against a real backend:

  1. Use httpheadersexample.php via your Varnish server to experiment and get a sense of what it is all about.
  2. Copy the PHP file from Varnish-Book/material/webdev/ to /var/www/html/
  3. Use varnishstat -f MAIN.client_req -f MAIN.cache_hit and varnishlog -g request -i ReqHeader,RespHeader to analyze the responses.
  4. Try every link several times by clicking on them and refreshing your browser.
  5. Analyze the response in your browser and the activity in your Varnish server.
  6. Discuss what happens when having the Cache-Control and Expires fields in the third link.
  7. When testing Last-Modified and If-Modified-Since, does your browser issue a request to Varnish? If the item was in cache, does Varnish query the origin server?
  8. Try the Vary header field from two different browsers.
When performing this exercise, try to see if you can spot the patterns. There are many levels of cache on the web, and you have to consider them besides Varnish. If it has not happened already, it is likely that the local cache of your browser will confuse you at least a few times through this course. When that happens, pull up varnishlog, varnishstat and another browser, or use client mock-ups in varnishtest instead of browsers.

VCL Basics

In this chapter, you will learn the following topics:

  • The Varnish Configuration Language (VCL) is a domain-specific language
  • VCL as a finite state machine
  • States as subroutines
  • Varnish includes built-in subroutines
  • Available functions, legal return actions and variables

The Varnish Configuration Language (VCL) is a domain-specific language designed to describe request handling and document caching policies for Varnish Cache. When a new configuration is loaded, the VCC process, created by the Manager process, translates the VCL code to C. This C code is compiled typically by gcc to a shared object. The shared object is then loaded into the cacher process.

This chapter focuses on the most important tasks to write effective VCL code. For this, you will learn the basic syntax of VCL, and the most important VCL built-in subroutines: vcl_recv and vcl_backend_fetch. All other built-in subroutines are taught in the next chapter.

Tip

Remember that Varnish has many reference manuals. For more details about VCL, check its manual page by issuing man vcl.

Varnish Finite State Machine

  • VCL workflow seen as a finite state machine – See Figure 22 in the book
  • States are conceptualized and implemented as subroutines, e.g., sub vcl_recv
  • Built-in subroutines start with vcl_, which is a reserved prefix
  • return (action) terminates subroutines, where action is a keyword that indicates the next step to do

Snippet from vcl_recv subroutine

Figure :counter:`figure`: Simplified Version of the Varnish Finite State Machine

VCL is also often described as a finite state machine. Each state has available certain parameters that you can use in your VCL code. For example: response HTTP headers are only available after vcl_backend_fetch state.

Figure 22 shows a simplified version of the Varnish finite state machine. This version shows by no means all possible transitions, but only a traditional set of them. Figure 23 and Figure 24 show the detailed version of the state machine for the frontend and backend worker respectively. The figures represent a request flow diagram.

States in VCL are conceptualized as subroutines, with the exception of the waiting state described in Waiting State

Subroutines in VCL take neither arguments nor return values. Each subroutine terminates by calling return (action), where action is a keyword that indicates the desired outcome. Subroutines may inspect and manipulate HTTP header fields and various other aspects of each request. Subroutines instruct how requests are handled.

Subroutine example:

sub pipe_if_local {
  if (client.ip ~ local) {
    return (pipe);
  }
}

To call a subroutine, use the call keyword followed by the subroutine's name:

call pipe_if_local;

Varnish has built-in subroutines that are hook into the Varnish workflow. These built-in subroutines are all named vcl_*. Your own subroutines cannot start their name with vcl_.

Detailed Varnish Request Flow for the Client Worker Thread

  • Figure 23 shows the detailed request flow diagram of the backend worker.
  • The grayed box is detailed in Figure 24.

Figure :counter:`figure`: Detailed Varnish Request Flow for the Client Worker Thread

The VCL Finite State Machine

  • Each request is processed separately
  • Each request is independent from others at any given time
  • States are related, but isolated
  • return(action); exits one state and instructs Varnish to proceed to the next state
  • Built-in VCL code is always present and appended below your own VCL

Before we begin looking at VCL code, it's worth trying to understand the fundamental concepts behind VCL. When Varnish processes a request, it starts by parsing the request itself. Next, Varnish separates the request method from headers, verifying that it's a valid HTTP request and so on.

When the basic parsing has completed, the very first policies are checked to make decisions. Policies are a set of rules that the VCL code uses to make a decision. Policies help to answer questions such as: should Varnish even attempt to find the requested resource in the cache? In this example, the policies are in the vcl_recv subroutine.

VCL Syntax

  • VCL files start with vcl 4.0;
  • //, # and /* foo */ for comments
  • Subroutines are declared with the sub keyword
  • No loops, state-limited variables
  • Terminating statements with a keyword for next action as argument of the return() function, i.e.: return(action)
  • Domain-specific
  • Add as little or as much as you want

Starting with Varnish 4.0, each VCL file must start by declaring its version with a special vcl 4.0; marker at the top of the file. If you have worked with a programming language or two before, the basic syntax of Varnish should be reasonably straightforward. VCL is inspired mainly by C and Perl. Blocks are delimited by curly braces, statements end with semicolons, and comments may be written as in C, C++ or Perl according to your own preferences.

Subroutines in VCL neither take arguments, nor return values. Subroutines in VCL can exchange data only through HTTP headers.

VCL has terminating statements, not traditional return values. Subroutines end execution when a return(*action*) statement is made. The action tells Varnish what to do next. For example, "look this up in cache", "do not look this up in the cache", or "generate an error message". To check which actions are available at a given built-in subroutine, see the Legal Return Actions section or see the manual page of VCL.

Warning

If you define your own subroutine and call it from one of the built-in subroutines, executing return(foo) does not return execution from your custom subroutine to the default function, but returns execution from VCL to Varnish.

VCL in varnishtest

Snippet from vcl_recv subroutine in varnishtest

varnishtest allows you to insert VCL code with the -vcl directive when declaring a Varnish server. This VCL code is inserted above the subroutines in built-in code in {varnish-source-code}/bin/varnishd/builtin.vcl. Since builtin.vcl already includes vcl 4.0;, you do not need to add it in varnishtest.

Example: built-in vcl_recv

The built-in VCL for vcl_recv is designed to ensure a safe caching policy even with no modifications in VCL. It has two main uses:

  1. Only handle recognized HTTP methods.
  2. Cache requests with GET and HEAD headers.

Policies for no caching data are to be defined in your VCL. Built-in VCL code is executed right after any user-defined VCL code, and is always present. You can not remove built-in subroutines, however, you can avoid them if your VCL code reaches one of the terminating actions: pass, pipe, hash, or synth. These terminating actions return control from the VRT (VCL Run-Time) to Varnish.

For a well-behaving Varnish server, most of the logic in the built-in VCL is needed. Consider either replicating all the built-in VCL logic in your own VCL code, or let your client requests be handled by the built-in VCL code.

We will revisit and discuss in more detail the vcl_recv subroutine in VCL Built-in Subroutines, but before, let's learn more about built-in functions, keywords, variables and return actions

VCL Built-in Functions and Keywords

Functions:

  • regsub(str, regex, sub)
  • regsuball(str, regex, sub)
  • ban(boolean expression)
  • hash_data(input)
  • synthetic(str)

Keywords:

  • call subroutine
  • return(action)
  • new()
  • set()
  • unset()

All functions are available in all subroutines, except the listed in the table below.

VCL offers a handful of simple to use built-in functions that allow you to modify strings, add bans, restart the VCL state engine and return control from the VCL Run Time (VRT) environment to Varnish. This book describes the most important functions in later sections, so the description at this point is brief.

regsub() and regsuball() have the same syntax and does the almost same thing: They both take a string str as input, search it with a regular-expression regex and replace it with another string. The difference between regsub() and regsuball() is that the latter changes all occurrences while the former only affects the first match.

The ban(boolean expression) function invalidates all objects in cache that match the boolean expression. banning and purging in detailed in the Cache Invalidation chapter.

Legal Return Actions

The table above shows the VCL built-in subroutines and their legal returns. return is a built-in keyword that ends execution of the current VCL subroutine and continue to the next action step in the request handling state machine. Legal return actions are: lookup, synth, purge, pass, pipe, fetch, deliver, hash, restart, retry, and abandon.

Note

In Varnish 4 purge is used as a return action.

Variables in VCL subroutines

The State column lists the different states in a request work-flow. States are handled by subroutines, which have a leading vcl_ prefix name.

The Variables columns list the prefix of the available variables. Most but not all of the prefixed variables are readable (R) or writable (W) at the given state. To have a detailed availability of each variable, refer to the VCL man page by typing: man vcl.

Table 17 shows the availability of variables in different states of the Varnish finite state machine. In addition to the variable prefixes in Table 17, there are other three variables prefixes; client.*, server.*, and storage.*, which are accessible from all subroutines at the frontend (client) side. Another variable is now, which is accessible from all subroutines.

These additional prefixes and variables are practically accessible everywhere.

Remember that changes made to beresp. variables are stored in obj. afterwards. And the resp. variables are copies of what is about to be returned to the client. The values of resp. variables come possibly from obj..

A change to beresp. variables, in other words, affects obj. and resp. variables. Similar semantics apply to req. and bereq. variables.

Variables belonging to the backend request (bereq.) are assigned with values from the original request (req.). However, their values may slightly differ, because Varnish may modify HTTP requests methods. For example, client requests with HEAD methods may be converted to backend requests with GET methods.

Many but not all of the variables are self-explanatory. To get more information about a particular variable, consult the VCL man page or ask the instructor at the course.

VCL – vcl_backend_fetch

Figure :counter:`figure`: Varnish Request Flow for the Backend Worker Thread.

Figure 24 shows the vcl_backend_fetch, vcl_backend_response and vcl_backend_error subroutines. These subroutines are the backend-counterparts to vcl_recv. You can use data provided by the client in vcl_recv or even vcl_backend_fetch to define your caching policy. An important difference is that you have access to bereq.* variables in vcl_backend_fetch.

vcl_backend_fetch can be called from vcl_miss or vcl_pass. When vcl_backend_fetch is called from vcl_miss, the fetched object may be cached. If vcl_backend_fetch is called from vcl_pass, the fetched object is not cached even if obj.ttl or obj.keep variables are greater than zero.

A relevant variable is bereq.uncacheable. This variable indicates whether the object requested from the backend may be cached or not. However, all objects from pass requests are never cached, regardless the bereq.uncacheable variable.

vcl_backend_fetch has two possible terminating actions, fetch or abandon. The fetch action sends the request to the backend, whereas the abandon action calls the vcl_synth routine. The built-in vcl_backend_fetch subroutine simply returns the fetch action. The backend response is processed by vcl_backend_response or vcl_backend_error. You will learn more about vcl_backend_response in the VCL – vcl_backend_response section.

Summary of VCL

  • VCL is all about policies
  • VCL provides a state machine for controlling Varnish
  • Each request is handled independently
  • Building a VCL file is done one line at a time

VCL provides subroutines that allow you to affect the handling of any single request almost anywhere in the execution chain. This provides pros and cons as any other programming language.

This book is not a complete reference guide to how you can deal with every possible scenario in VCL, but if you master the basics of VCL you can solve complex problems that nobody has thought about before. And you can usually do it without requiring too many different sources of documentation.

Whenever you are working on VCL, you should think of what that exact line you are writing has to do. The best VCL is built by having many independent sections that do not interfere with each other more than what they have to.

Remember that there is a built-in VCL. If your own VCL code does not reach a return statement, the built-in VCL subroutine is executed after yours. If you just need a little modification of a subroutine, you can use the code from {varnish-source-code}/bin/varnishd/builtin.vcl as a template.

VCL Built-in Subroutines

  • Cover the VCL built-in subroutines: vcl_recv, vcl_pass, vcl_backend_fetch, vcl_backend_response, vcl_hash, vcl_hit, vcl_miss, vcl_deliver, and vcl_synth
  • If your VCL code does not reach a return statement, the built-in VCL subroutine is executed after yours.

This chapter covers the VCL subroutines where you customize the behavior of Varnish. However, this chapter does not define caching policies. VCL subroutines can be used to: add custom headers, change the appearance of the Varnish error message, add HTTP redirect features in Varnish, purge content, and define what parts of a cached object is unique.

After this chapter, you should know what all the VCL subroutines can be used for. You should also be ready to dive into more advanced features of Varnish and VCL.

Note

It is strongly advised to let the default built-in subroutines whenever is possible. The built-in subroutines are designed with safety in mind, which often means that they handle any flaws in your VCL code in a reasonable manner.

Tip

Looking at the code of built-in subroutines can help you to understand how to build your own VCL code. Built-in subroutines are in the file /usr/share/doc/varnish/examples/builtin.vcl.gz or {varnish-source-code}/bin/varnishd/builtin.vcl. The first location may change depending on your distro.

VCL – vcl_recv

  • Normalize client-input
  • Pick a backend web server
  • Re-write client-data for web applications
  • Decide caching policy based on client-input
  • Access Control Lists (ACL)
  • Security barriers, e.g., against SQL injection attacks
  • Fixing mistakes, e.g., index.htlm -> index.html

vcl_recv is the first VCL subroutine executed, right after Varnish has parsed the client request into its basic data structure. vcl_recv has four main uses:

  1. Modifying the client data to reduce cache diversity. E.g., removing any leading "www." in the Host: header.
  2. Deciding which web server to use.
  3. Deciding caching policy based on client data. For example; no caching POST requests but only caching specific URLs.
  4. Executing re-write rules needed for specific web applications.

In vcl_recv you can perform the following terminating actions:

pass: It passes over the cache lookup, but it executes the rest of the Varnish request flow. pass does not store the response from the backend in the cache.

pipe: This action creates a full-duplex pipe that forwards the client request to the backend without looking at the content. Backend replies are forwarded back to the client without caching the content. Since Varnish does no longer try to map the content to a request, any subsequent request sent over the same keep-alive connection will also be piped. Piped requests do not appear in any log.

hash: It looks up the request in cache.

purge: It looks up the request in cache in order to remove it.

synth - Generate a synthetic response from Varnish. This synthetic response is typically a web page with an error message. synth may also be used to redirect client requests.

It's also common to use vcl_recv to apply some security measures. Varnish is not a replacement for intrusion detection systems, but can still be used to stop some typical attacks early. Simple Access Control Lists (ACLs) can be applied in vcl_recv too.

For further discussion about security in VCL, take a look at the Varnish Security Firewall (VSF) application at https://github.com/comotion/VSF. The VSF supports Varnish 3 and above. You may also be interested to look at the Security.vcl project at https://github.com/comotion/security.vcl. The Security.vcl project, however, supports only Varnish 3.x.

Tip

The built-in vcl_recv subroutine may not cache all what you want, but often it's better not to cache some content instead of delivering the wrong content to the wrong user. There are exceptions, of course, but if you can not understand why the default VCL does not let you cache some content, it is almost always worth it to investigate why instead of overriding it.

Revisiting built-in vcl_recv

Example: Basic Device Detection

One way of serving different content for mobile devices and desktop browsers is to run some simple parsing on the User-Agent header. The following VCL code is an example to create custom headers. These custom headers differentiate mobile devices from desktop computers.

You can read more about different types of device detection at https://www.varnish-cache.org/docs/trunk/users-guide/devicedetection.html

This simple VCL will create a request header called X-Device which will contain either mobile or desktop. The web server can then use this header to determine what page to serve, and inform Varnish about it through Vary: X-Device.

It might be tempting to just send Vary: User-Agent, but that requires you to normalize the User-Agent header itself because there are many tiny variations in the description of similar User-Agents. This normalization, however, leads to loss of detailed information of the browser. If you pass the User-Agent header without normalization, the cache size may drastically inflate because Varnish would keep possibly hundreds of different variants per object and per tiny User-Agent variants. For more information on the Vary HTTP response header, see the Vary section.

Note

If you do use Vary: X-Device, you might want to send Vary: User-Agent to the users after Varnish has used it. Otherwise, intermediary caches will not know that the page looks different for different devices.

Exercise: Rewrite URL and Host Header Fields

  1. Copy the Host header field (req.http.Host) and URL (req.url) to two new request headers: req.http.x-host and req.http.x-url.
  2. Ensure that www.example.com and example.com are cached as one, using regsub().
  3. Rewrite all URLs under http://sport.example.com to http://example.com/sport/. For example: http://sport.example.com/index.html to http://example.com/sport/index.html.
  4. Use HTTPie or varnishtest to verify the result.
  • Extra: Make sure / and /index.html are cached as one object.
  • Extra 2: Make the redirection work for any domain with sport. at the front. E.g: sport.example.com, sport.foobar.example.net, sport.blatti, etc.

For the first point, use set req.http.headername = "value"; or set req.http.headername = regsub(...);.

In point 2, change req.http.host by calling the function regsub(str, regex, sub). str is the input string, in this case, req.http.host. regex is the regular-expression matching whatever content you need to change. Use ^ to match what begins with www, and \. to finish the regular-expression, i.e. ^www.. sub is what you desire to change it with, an empty string "" can be used to remove what matches regex.

For point 3, you can check host headers with a specific domain name, for example: if (req.http.host == "sport.example.com"). An alternative is to check for all hosts that start with sport, regardless the domain name: if (req.http.host ~ "^sport\."). In the first case, setting the host header is straight forward: set req.http.host = "example.com". In the second case, you can set the host header by removing the string that precedes the domain name set req.http.host = regsub(req.http.host,"^sport\.", ""); Finally, you rewrite the URL in this way: set req.url = regsub(req.url, "^", "/sport");.

To simulate client requests, you can either use HTTPie or varnishtest.

Tip

Remember that man vcl contains a reference manual with the syntax and details of functions such as regsub(str, regex, sub). We recommend you to leave the default VCL file untouched and create a new file for your VCL code. Remember to update the location of the VCL file in the Varnish configuration file and reload it.

Solution: Rewrite URL and Host Header Fields
sub vcl_recv {

    set req.http.x-host = req.http.host;
    set req.http.x-url = req.url;

    set req.http.host = regsub(req.http.host, "^www\.", "");

    /* Alternative 1 */
    if (req.http.host == "sport.example.com") {
        set req.http.host = "example.com";
        set req.url = "/sport" + req.url;
     }

     /* Alternative 2 */
     if (req.http.host ~ "^sport\.") {
       set req.http.host = regsub(req.http.host,"^sport\.", "");
       set req.url = regsub(req.url, "^", "/sport");
     }
}

You can test this solution via HTTPie or varnishtest. Using HTTPie:

http -p hH --proxy=http:http://localhost sport.example.com/index.html

Then you verify your results by issuing the following command and analyzing the output:

varnishlog -i ReqHeader,ReqURL

varnishtest solution:

VCL – vcl_pass

  • Called upon entering pass mode

The vcl_pass subroutine is called after a previous subroutine returns the pass action. This actions sets the request in pass mode. vcl_pass typically serves as an important catch-all for features you have implemented in vcl_hit and vcl_miss.

vcl_pass may return three different actions: fetch, synth, or restart. When returning the fetch action, the ongoing request proceeds in pass mode. Fetched objects from requests in pass mode are not cached, but passed to the client. The synth and restart return actions call their corresponding subroutines.

hit-for-pass

  • Used when an object should not be cached
  • hit-for-pass object instead of fetched object
  • Has TTL

Some requested objects should not be cached. A typical example is when a requested page contains the Set-Cookie response header, and therefore it must be delivered only to the client that requests it. In this case, you can tell Varnish to create a hit-for-pass object and stores it in the cache, instead of storing the fetched object. Subsequent requests are processed in pass mode.

When an object should not be cached, the beresp.uncacheable variable is set to true. As a result, the cacher process keeps a hash reference to the hit-for-pass object. In this way, the lookup operation for requests translating to that hash find a hit-for-pass object. Such requests are handed over to the vcl_pass subroutine, and proceed in pass mode.

As any other cached object, hit-for-pass objects have a TTL. Once the object's TTL has elapsed, the object is removed from the cache.

VCL – vcl_backend_response

  • Sanitize server-response
  • Override cache duration
  • See Figure 24 in the book

Figure 24 shows that vcl_backend_response may terminate with one of the following actions: deliver, abandon, or retry. The deliver terminating action may or may not insert the object into the cache depending on the response of the backend.

Backends might respond with a 304 HTTP headers. 304 responses happen when the requested object has not been modified since the timestamp If-Modified-Since in the HTTP header. If the request hits a non fresh object (see Figure 2), Varnish adds the If-Modified-Since header with the value of t_origin to the request and sends it to the backend.

304 responses do not contain a message-body. Thus, Varnish tries to steal the body from cache, merge it with the header response and deliver it. This process updates the attributes of the cached object.

Typical tasks performed in vcl_backend_fetch or vcl_backend_response include:

  • Overriding cache time for certain URLs
  • Stripping Set-Cookie header fields that are not needed
  • Stripping bugged Vary header fields
  • Adding helper-headers to the object for use in banning (more information in later sections)
  • Applying other caching policies

vcl_backend_response

built-in vcl_backend_response

sub vcl_backend_response {
    if (beresp.ttl <= 0s ||
      beresp.http.Set-Cookie ||
      beresp.http.Surrogate-control ~ "no-store" ||
      (!beresp.http.Surrogate-Control &&
        beresp.http.Cache-Control ~ "no-cache|no-store|private") ||
      beresp.http.Vary == "*") {
        /*
         * Mark as "Hit-For-Pass" for the next 2 minutes
         */
        set beresp.ttl = 120s;
        set beresp.uncacheable = true;
    }
    return (deliver);
}

The vcl_backend_response built-in subroutine is designed to avoid caching in conditions that are most probably undesired. For example, it avoids caching responses with cookies, i.e., responses with Set-Cookie HTTP header field. This built-in subroutine also avoids request serialization described in the Waiting State section.

To avoid request serialization, beresp.uncacheable is set to true, which in turn creates a hit-for-pass object. The hit-for-pass section explains in detail this object type.

If you still decide to skip the built-in vcl_backend_response subroutine by having your own and returning deliver, be sure to never set beresp.ttl to 0. If you skip the built-in subroutine and set 0 as TTL value, you are effectively removing objects from cache that could eventually be used to avoid request serialization.

Note

Varnish 3.x has a hit_for_pass return action. In Varnish 4, this action is achieved by setting beresp.uncacheable to true. The hit-for-pass section explains this in more detail.

The Initial Value of beresp.ttl

Before Varnish runs vcl_backend_response, the beresp.ttl variable has already been set to a value. beresp.ttl is initialized with the first value it finds among:

  • The s-maxage variable in the Cache-Control response header field
  • The max-age variable in the Cache-Control response header field
  • The Expires response header field
  • The default_ttl parameter.

Only the following status codes will be cached by default:

  • 200: OK
  • 203: Non-Authoritative Information
  • 300: Multiple Choices
  • 301: Moved Permanently
  • 302: Moved Temporarily
  • 304: Not modified
  • 307: Temporary Redirect
  • 410: Gone
  • 404: Not Found

You can cache other status codes than the ones listed above, but you have to set the beresp.ttl to a positive value in vcl_backend_response. Since beresp.ttl is set before vcl_backend_response is executed, you can modify the directives of the Cache-Control header field without affecting beresp.ttl, and vice versa. Cache-Control directives are defined in RFC7234 Section 5.2.

A backend response may include the response header field of maximum age for shared caches s-maxage. This field overrides all max-age values throughout all Varnish servers in a multiple Varnish-server setup. For example, if the backend sends Cache-Control: max-age=300, s-maxage=3600, all Varnish installations will cache objects with an Age value less or equal to 3600 seconds. This also means that responses with Age values between 301 and 3600 seconds are not cached by the clients' web browser, because Age is greater than max-age.

A sensible approach is to use the s-maxage directive to instruct Varnish to cache the response. Then, remove the s-maxage directive using regsub() in vcl_backend_response before delivering the response. In this way, you can safely use s-maxage as the cache duration for Varnish servers, and set max-age as the cache duration for clients.

Warning

Bear in mind that removing or altering the Age response header field may affect how responses are handled downstream. The impact of removing the Age field depends on the HTTP implementation of downstream intermediaries or clients.

For example, imagine that you have a three Varnish-server serial setup. If you remove the Age field in the first Varnish server, then the second Varnish server will assume Age=0. In this case, you might inadvertently be delivering stale objects to your client.

Example: Setting TTL of .jpg URLs to 60 seconds

The above example caches all URLs ending with .jpg for 60 seconds. Keep in mind that the built-in VCL is still executed. That means that images with a Set-Cookie field are not cached.

Example: Cache .jpg for 60 seconds only if s-maxage is not present

The purpose of the above example is to allow a gradual migration to using a backend-controlled caching policy. If the backend does not supply s-maxage, and the URL is a jpg file, then Varnish sets beresp.ttl to 60 seconds.

The Cache-Control response header field can contain a number of directives. Varnish parses this field and looks for s-maxage and max-age.

By default, Varnish sets beresp.ttl to the value of s-maxage if found. If s-maxage is not found, Varnish uses the value max-age. If neither exists, Varnish uses the Expires response header field to set the TTL. If none of those header fields exist, Varnish uses the default TTL, which is 120 seconds.

The default parsing and TTL assignment are done before vcl_backend_response is executed. The TTL changing process is recorded in the TTL tag of varnishlog.

Exercise: Avoid Caching a Page

  • Write a VCL which avoids caching the index page at all
  • Your VCL should cover both resource targets: / and /index.html
When trying this out, remember that Varnish keeps the Host header field in req.http.host and the requested resource in req.url. For example, in a request to http://www.example.com/index.html, the http:// part is not seen by Varnish at all, req.http.host has the value www.example.com and req.url the value /index.html. Note how the leading / is included in req.url.
Solution: Avoid caching a page
// Suggested solution A
sub vcl_recv {
     if (req.url ~ "^/index\.html" || req.url ~ "^/$") {
             return(pass);
     }
}

// Suggested solution B
sub vcl_backend_response {
    if (bereq.url ~ "^/index\.html" || bereq.url ~ "^/$") {
       set beresp.uncacheable = true;
    }
}
Usually it is most convenient to do as much as possible in vcl_recv. The usage of bereq.uncacheable in vcl_backend_fetch creates a hit-for-pass object. See the hit-for-pass section for detailed description about this type of object.

Exercise: Either use s-maxage or set TTL by file type

Write a VCL that:

  • uses Cache-Control: s-maxage when present,
  • caches .jpg for 30 seconds if s-maxage is not present,
  • caches .html for 10 seconds if s-maxage isn't present, and
  • removes the Set-Cookie header field if s-maxage or the above rules indicate that Varnish should cache.

Tip

Divide and conquer! Most somewhat complex VCL tasks are easily solved when you divide the tasks into smaller problems and solve them individually. Try solving each part of the exercise by itself first.

Note

Varnish automatically parses s-maxage for you, so you only need to check if it is there or not. Remember that if s-maxage is present, Varnish has already used it to set beresp.ttl.

Solution: Either use s-maxage or set ttl by file type
There are many ways to solve this exercise, and this solution is only one of them. The first condition checks the presence of s-maxage and handles .jpg and .html files to make them cacheable for 30 and 10 seconds respectively. If s-maxage is present with a positive TTL, we consider the response cacheable by removing beresp.http.Set-Cookie.

Waiting State

  • Request serialization is a non desired side-effect that is handled in the vcl_backend_response subroutine
  • Designed to improve response performance

The waiting state is reached when a request n arrives while a previous identical request 0 is being handled at the backend. In this case, request 0 is set as busy and all subsequent requests n are queued in a waiting list. If the fetched object from request 0 is cacheable, all n requests in the waiting list call the lookup operation again. This retry will hopefully hit the desired object in cache. As a result, only one request is sent to the backend.

The waiting state is designed to improve response performance. However, a counterproductive scenario, namely request serialization, may occur if the fetched object is uncacheable, and so is recursively the next request in the waiting list. This situation forces every single request in the waiting list to be sent to the backend in a serial manner. Serialized requests should be avoided because their performance is normally poorer than sending multiple requests in parallel. The built-in vcl_backend_response subroutine avoids request serialization.

VCL – vcl_hash

  • Defines what is unique about a request.
  • Executed when vcl_recv returns the hash action keyword.

vcl_hash defines the hash key to be used for a cached object. Hash keys differentiate one cached object from another. The default VCL for vcl_hash adds the hostname or IP address, and the requested URL to the cache hash.

One usage of vcl_hash is to add a user-name in the cache hash to identify user-specific data. However, be warned that caching user-data should only be done cautiously. A better alternative might be to hash cache objects per session instead.

The vcl_hash subroutine returns the lookup action keyword. Unlike other action keywords, lookup is an operation, not a subroutine. Depending on what lookup finds in the cache, it has four possible outcomes: hit, miss, hit-for-pass, or purge.

When the lookup operation does not match any hash, it creates an object with a busy flag and inserts it in cache. Then, the request is sent to the vcl_miss subroutine. The busy flag is removed once the request is handled, and the object is updated with the response from the backend.

Subsequent similar requests that hit busy flagged objects are sent into a waiting list. This waiting list is designed to improve response performance, and it is explain the Waiting State section.

Note

One cache hash may refer to one or many object variations. Object variations are created based on the Vary header field. It is a good practice to keep several variations under one cache hash, than creating one hash per variation.

VCL – vcl_hit

  • Executed after the lookup operation, called by vcl_hash, finds (hits) an object in the cache.

The vcl_hit subroutine typically terminate by calling return() with one of the following keywords: deliver, restart, or synth.

deliver returns control to vcl_deliver if the TTL + grace time of an object has not elapsed. If the elapsed time is more than the TTL, but less than the TTL + grace time, then deliver calls for background fetch in parallel to vcl_deliver. The background fetch is an asynchronous call that inserts a fresher requested object in the cache. Grace time is explained in the Grace Mode section.

restart starts again the transaction, and increases the restart counter. If the number of restarts is higher than max_restarts counter, Varnish emits a guru meditation error.

synth(status code, reason) returns the specified status code to the client and abandon the request.

VCL – vcl_miss

  • Subroutine called if a requested object is not found by the lookup operation.
  • Contains policies to decide whether or not to attempt to retrieve the document from the backend, and which backend to use.
The subroutines vcl_hit and vcl_miss are closely related. It is rare that you customize them, because modification of HTTP request headers is typically done in vcl_recv. However, if you do not wish to send the X-Varnish header to the backend server, you can remove it in vcl_miss or vcl_pass. For that case, you can use unset bereq.http.x-varnish;.

VCL – vcl_deliver

  • Common last exit point for all request workflows, except requests through vcl_pipe
  • Often used to add and remove debug-headers

The vcl_deliver subroutine is simple, and it is also very useful to modify the output of Varnish. If you need to remove a header, or add one that is not supposed to be stored in the cache, vcl_deliver is the place to do it.

The variables most useful and common to modify in vcl_deliver are:

resp.http.*
Headers that are sent to the client. They can be set and unset.
resp.status
The status code (200, 404, 503, etc).
resp.reason
The HTTP status message that is returned to the client.
obj.hits
The count of cache-hits on this object. Therefore, a value of 0 indicates a miss. This variable can be evaluated to easily reveal whether a response comes from a cache hit or miss.
req.restarts
The number of restarts issued in VCL - 0 if none were made.

VCL – vcl_synth

  • Used to generate content from within Varnish, without talking to the backend
  • Error messages go here by default
  • Other use cases: Redirecting users (301/302 redirects)

To make a distinction between VCL synthetic responses and internally generated errors (when trying to fetch an object), there are two subroutines that handle errors in Varnish. One is vcl_synth, and the other is vcl_backend_error. To return control to vcl_synth,call the synth() function like this:

return (synth(status_code, "reason"));

Note that the syntax synth is not a keyword, but a function with arguments.

You must explicitly return the status code and reason arguments for vcl_synth. Setting headers on synthetic response bodies are done on resp.http.

Note

Note how you can use {" and "} to make multi-line strings. This is not limited to the synthetic() function, but can be used anywhere.

Note

A vcl_synth defined object is never stored in cache, contrary to a vcl_backend_error defined object, which may end up in cache. vcl_synth and vcl_backend_error replace vcl_error from Varnish 3.

Example: Redirecting requests with vcl_synth

Redirecting with VCL is fairly easy – and fast. Basic HTTP redirects work when the HTTP response is either 301 Moved Permanently or 302 Found. These response have a Location header field telling the web browser where to redirect.

Note

The 301 response can affect how browsers prioritize history and how search engines treat the content. 302 responses are temporary and do not affect search engines as 301 responses do.

Exercise: Modify the HTTP response header fields

  • Add a header field holding the string HIT if the requested resourced was found in cache, or MISS otherwise
  • "Rename" the Age header field to X-Age

Solution: Modify the HTTP response header fields

It is a good practice to ensure that a variable has the expected type. You should do this before acting on the content of a variable. Therefore, in this solution, we use the greater than comparison operator obj.hits > 0 instead of the not equal to operator obj.hits != 0.

There have been some bugs when converting strings. Those bugs happened when the variable to be converted had an unexpected value. This may apply to all variable types – and all languages for that matter. Thus it is important that you always check the variable type.

Exercise: Change the error message

  • Make the default error message more friendly.

Solution: Change the error message

vcl/customized_error.vcl

The suggested solution forces a 503 error by misconfiguring .port in the default backend. You can also force a 503 response by using ${bad_ip} in varnishtest. The macro ${bad_ip} translates to 192.0.2.255.

vtc/b00011.vtc

Note that in the proposed solution the client receives a 200 response code.

Cache Invalidation

  • Cache invalidation is an important part of your cache policy
  • Varnish automatically invalidates expired objects
  • You can however pro-actively invalidate objects with Varnish
  • You should define cache invalidation rules before caching objects specially in production environments

There are four mechanisms to invalidate caches in Varnish:

  1. HTTP PURGE
  • Use the vcl_purge subroutine
  • Invalidate caches explicitly using objects' hashes
  • vcl_purge is called via return(purge) from vcl_recv
  • vcl_purge removes all variants of an object from cache, freeing up memory
  • The restart return action can be used to update immediately a purged object
  1. Banning
  • Use the built-in function ban(regex)
  • Invalidates objects in cache that match the regular-expression
  • Does not necessarily free up memory at once
  • Also accessible from the management interface
  1. Force Cache Misses
  • Use req.hash_always_miss in vcl_recv
  • If set to true, Varnish disregards any existing objects and always (re)fetches from the backend
  • May create multiple objects as side effect
  • Does not necessarily free up memory at once
  1. Hashtwo -> Varnish Plus only!
  • For websites with the need for cache invalidation at a very large scale
  • Varnish Software's implementation of surrogate keys
  • Flexible cache invalidation based on cache tags

Purge vs. Bans vs. Hashtwo vs. Cache Misses

Which one to use and when?

Whenever you deal with caching, you have to eventually deal with the challenge of cache invalidation, or content update. Varnish has different mechanisms to addresses this challenge, but which one to use?

There is rarely a need to pick only one solution, as you can implement many of them. However, you can try to answer the following questions:

  • Am I invalidating one or many specific objects?
  • Do I need to free up memory or just replace the content?
  • How long time does it take to replace the content?
  • Is this a regular or a one-off task?

or follow these guidelines:

  • If you need to invalidate more than one item at a time, consider to use bans or hashtwo.
  • If it takes a long time to pull content from the backend into Varnish, consider to force cache misses by using req.hash_always_miss.

The rest of the chapter teaches you more about each mechanism.

Note

Purge and Hashtwo work very similar. The main difference is that they have and act on different hash keys.

HTTP PURGE

  • If you know exactly what to remove, use HTTP PURGE
  • Frees up memory, removes all Vary:-variants of the object
  • Leaves it to the next client to refresh the content
  • Often combined with return(restart);
  • As easy as handling any other HTTP request

A purge is what happens when you pick out an object from the cache and discard it along with its variants. A resource can exist in multiple Vary:-variants. For example, you could have a desktop version, a tablet version and a smartphone version of your site, and use the Vary HTTP header field in combination with device detection to store different variants of the same resource.

Usually a purge is invoked through HTTP with the method PURGE. A HTTP PURGE is another request method just as HTTP GET. Actually, you can call the PURGE method whatever you like, but PURGE has become the de-facto naming standard. Squid, for example, uses the PURGE method name for the same purpose.

Purges apply to a specific object, since they use the same lookup operation as in vcl_hash. Therefore, purges find and remove objects really fast!

There are, however, two clear down-sides. First, purges cannot use regular-expressions, and second, purges evict content from cache regardless the availability of the backend. That means that if you purge some objects and the backend is down, Varnish will end up having no copy of the content.

VCL – vcl_purge

  • You may add actions to be executed once the object and its variants is purged
  • Called after the purge has been executed

Note

Cache invalidation with purges is done by calling return (purge); from vcl_recv in Varnish 4. The keyword purge; from Varnish 3 has been retired.

Example: PURGE

vcl/PURGE.vcl

acl is a reserved keyword that is used to create Access Control Lists (ACLs). ACLs are used to control which client IP addresses are allowed to purge cached objects.

In the example above, return (purge) ends execution of vcl_recv and jumps to vcl_hash. When vcl_hash calls return(lookup), Varnish purges the object and then calls vcl_purge.

You can test this code with HTTPie by issuing:

http -p hH --proxy=http:http://localhost PURGE www.example.com

Alternatively, you can test it with varnishtest:

vtc/b00012.vtc

The example above is a modification of Varnish-Cache/bin/varnishtest/tests/b00036.vtc. In this VTC you can see how the second request of c1 is constructed out from the cached object. After purging the implicit resource /, the third request from c1 fetches a new object from s1.

A quick way for you to double check the acl authorization is by changing the IP address in it. So you should see that the PURGE request coming from localhost will not pass the condition:

if (!client.ip ~ purgers)

Exercise: PURGE an article from the backend

  • Send a PURGE request to Varnish from your backend server after an article is published.
    • Simulate the article publication.
    • The result is that the article is evicted in Varnish.

You are provided with article.php, which fakes an article. It is recommended to create a separate php file to implement purging.

article.php

Tip

Remember to place your php files under /var/www/html/.

Solution: PURGE an article from the backend

purgearticle.php

solution-purge-from-backend.vcl

PURGE with restart return action

  • Start the VCL processing again from the top of vcl_recv
  • Any changes made are kept

The restart return action allows Varnish to re-run the VCL state machine with different variables. This is useful in combination with PURGE, in the way that a purged object can be immediately restored with a new fetched object.

Every time a restart occurs, Varnish increments the req.restarts counter. If the number of restarts is higher than the max_restarts parameter, Varnish emits a guru meditation error. In this way, Varnish safe guards against infinite loops.

Warning

Restarts are likely to cause a hit against the backend, so do not increase max_restarts thoughtlessly.

Banning

  • Use ban to prevent Varnish from serving a cached object

  • Does not free up memory

  • Examples in the varnishadm command line interface:

    • ban req.url ~ /foo
    • ban req.http.host ~ example.com && obj.http.content-type ~ text
    • ban.list

  • Example in VCL:

    • ban("req.url ~ /foo");

  • Example of VCL code to catch the HTTP BAN request method:

    sub vcl_recv {
    if (req.method == "BAN") {
        ban("req.http.host == " + req.http.host +
            " && req.url == " + req.url);
        # Throw a synthetic page so the request won't go to the backend.
        return(synth(200, "Ban added"));
    }
}

Banning in the context of Varnish refers to adding a ban expression that prohibits Varnish to serve certain objects from the cache. Ban expressions are more useful when using regular-expressions.

Bans work on objects already in the cache, i.e., it does not prevent new content from entering the cache or being served. Cached objects that match a ban are marked as obsolete. Obsolete objects are expunged by the expiry thread like any other object with obj.ttl == 0.

Ban expressions match against req.* or obj.* variables. Think about a ban expression as; "the requested URL starts with /sport", or "the cached object has a Server-header matching lighttpd". You can add ban expressions in VCL syntax via VCL code, HTTP request method, or CLI.

Ban expressions are inserted into a ban-list. The ban-list contains:

  • ID of the ban,
  • timestamp when the ban entered the ban-list,
  • counter of objects that have matched the ban expression,
  • a C flag for completed that indicates whether a ban is invalid because it is duplicated,
  • the ban expression.

To inspect the current ban-list, issue the ban.list command in the CLI:

0xb75096d0 1318329475.377475    10      obj.http.x-url ~ test0
0xb7509610 1318329470.785875    20C     obj.http.x-url ~ test1

Varnish tests bans whenever a request hits a cached object. A cached object is checked only against bans added after the last checked ban. That means that each object checks against a ban expression only once.

Bans that match only against obj.* are also checked by a background worker thread called the ban lurker. The parameter ban_lurker_sleep controls how often the ban lurker tests obj.* bans. The ban lurker can be disabled by setting ban_lurker_sleep to 0.

Note

You may accumulate a lot of ban expressions based in req.* variables if you have many objects with long TTL that are seldom accessed. This accumulation occurs because bans are kept until all cached objects have been checked against them. This might impact CPU usage and thereby performance.

Therefore, we recommend you to avoid req.* variables in your ban expressions, and to use obj.* variables instead. Ban expressions using only obj.* are called lurker-friendly bans.

Note

If the cache is completely empty, only the last added ban stays in the ban-list.

Tip

You can also execute ban expressions via the Varnish Administration Console (VAC).

ui/img/vac_bans.png

Figure :counter:`figure`: Executing ban expressions via the Varnish Administration Console (VAC).

Lurker-Friendly Bans

  • Ban expressions that match only against obj.*
  • Evaluated asynchronously by the ban lurker thread
  • Similar to the concept of garbage collection

Ban expressions are checked in two cases: 1) when a request hits a cached object, or 2) when the ban lurker wakes up. The first case is efficient only if you know that the cached objects to be banned are frequently accessed. Otherwise, you might accumulate a lot of ban expressions in the ban-list that are never checked. The second case is a better alternative because the ban lurker can help you keep the ban-list at a manageable size. Therefore, we recommend you to create ban expressions that are checked by the ban lurker. Such ban expressions are called lurker-friendly bans.

Lurker-friendly ban expressions are those that use only obj.*, but not req.* variables. Since lurker-friendly ban expressions lack of req.*, you might need to copy some of the req.* contents into the obj structure. In fact, this copy operation is a mechanism to preserve the context of client request in the cached object. For example, you may want to copy useful parts of the client context such as the requested URL from req to obj.

The following snippet shows an example on how to preserve the context of a client request in the cached object:

sub vcl_backend_response {
   set beresp.http.x-url = bereq.url;
}

sub vcl_deliver {
   # The X-Url header is for internal use only
   unset resp.http.x-url;
}

Now imagine that you just changed a blog post template that requires all blog posts that have been cached. For this you can issue a ban such as:

$ varnishadm ban 'obj.http.x-url ~ ^/blog'

Since it uses a lurker-friendly ban expression, the ban inserted in the ban-list will be gradually evaluated against all cached objects until all blog posts are invalidated. The snippet below shows how to insert the same expression into the ban-list in the vcl_recv subroutine:

sub vcl_recv {
   if (req.method == "BAN") {
      # Assumes the ``X-Ban`` header is a regex,
      # this might be a bit too simple.

      ban("obj.http.x-url ~ " + req.http.x-ban);
      return(synth(200, "Ban added"));
   }
}

Exercise: Write a VCL program using purge and ban

  • Write a VCL program that handles the PURGE and BAN HTTP methods.
  • When handling the BAN method, use the request header fields req.http.x-ban-url and req.http.x-ban-host
  • Use Lurker-Friendly Bans
  • To build further on this, you can also use the REFRESH HTTP method that fetches new content, using req.hash_always_miss, which is explained in the next subsection

To test this exercise, you can use HTTPie:

http -p hH PURGE http://localhost/testpage
http -p hH BAN http://localhost/ 'X-Ban-Url: .*html$' \
                                 'X-Ban-Host: .*\.example\.com'
http -p hH REFRESH http://localhost/testpage

You can also send PURGE, BAN and REFRESH requests in varnishtest:

client c1 {
   txreq -req BAN
   rxres
} -run

Remember that you still need specify the requested URL in c1 if is other than /. We advise you to search for purge and ban in Varnish-Cache/bin/varnishtest/tests/ to learn more on how to invalidate caches.

Solution: Write a VCL program using purge and ban

Force Cache Misses

  • set req.hash_always_miss = true; in vcl_recv
  • Causes Varnish to look the object up in cache, but ignore any copy it finds
  • Useful way to do a controlled refresh of a specific object
  • If the server is down, the cached object is left untouched
  • Depending on Varnish-version, it may leave extra copies in the cache
  • Useful to refresh slowly generated content

Setting a request in pass mode instructs Varnish to always ask a backend for content, without storing the fetched object into cache. The vcl_purge removes old content, but what if the web server is down?

Setting req.has_always_miss to true tells Varnish to look up the content in cache, but always miss a hit. This means that Varnish first calls vcl_miss, then (presumably) fetches the content from the backend, cache the updated object, and deliver the updated content.

The distinctive behavior of req.hash_always_miss occurs when the backend server is down or unresponsive. In this case, the current cached object is untouched. Therefore, client requests that do not enable req.hash_always_miss keep getting the old and untouched cached content.

Two important use cases for using req.hash_always_miss are when you want to: 1) control who takes the penalty for waiting around for the updated content (e.g. a script you control), and 2) ensure that content is not evicted before it is updated.

Note

Forcing cache misses do not evict old content. This means that causes Varnish to have multiple copies of the content in cache. The newest copy is always used. If you cache your content for a long period of time, the memory usage will increase gradually.

Hashtwo (Varnish Software Implementation of Surrogate Keys)

  • Hashtwo is Varnish Software's implementation of surrogate keys
  • Cache invalidation based on cache tags
  • Adds patterns easily to be matched against
  • Highly scalable
  • Hashtwo is distributed as a VMOD for Varnish Plus only!
  • Install the Varnish Plus VMODs

The idea is that you can use any arbitrary string for cache invalidation. You can then key your cached objects on, for example, product ID or article ID. In this way, when you update the price of a certain product or a specific article, you have a key to evict all those objects from the cache.

So far, we have discussed purges and bans as methods for cache invalidation. Two important distinctions between them is that purges remove a single object (with its variants), whereas bans perform cache invalidation based on matching expressions. However, there are cases where none of these mechanisms are optimal.

Hashtwo creates a second hash key to link cached objects based on cache tags. This hash keys provide the means to invalidate cached objects with common cache tags.

In practice, Hashtwo create cache invalidation patterns, which can be tested and invalidated immediately just as purges do. In addition, Hashtwo is much more efficient than bans because of two reasons: 1) looking up hash keys is much more efficient than traversing ban-lists, and 2) every time you test a ban expression, it checks every object in the cache that is older than the ban itself.

The hashtwo VMOD is pre-built for supported versions and can be installed using regular package managers from the Varnish Software repositories. Once the repository is in place you can issue the following commands to install the VMOD:

On Debian or Ubuntu:

apt-get install libvmod-hashtwo

On Red Hat Enterprise Linux:

yum install libvmod-hashtwo

Finally, you can use this VMOD by importing it in your VCL code:

import hashtwo;

VCL Example Using Hashtwo

On an e-commerce site the backend application issues a X-HashTwo HTTP header field for every product that is referenced on that page. The header for a certain page might look like this:

HTTP/1.1 200 OK
Server: Apache/2.2.15
X-HashTwo: 8155054
X-HashTwo: 166412
X-HashTwo: 234323

The VCL example code:

import hashtwo;

# In this example the key to be purged is specified on the
# X-HashTwo-Purge header.
sub vcl_recv {
  if (req.http.X-HashTwo-Purge) {
    if (hashtwo.purge(req.http.X-HashTwo-Purge) != 0) {
       return (purge);
    } else {
      return (synth(404, "Key not found"));
    }
  }
}

# Normally the backend is responsible for setting the header.
# If you were to do it in VCL it will look something like this:
sub vcl_backend_response {
  set beresp.http.X-HashTwo = "secondary_hash_key";
}

In order to keep the web pages in sync with the database, a trigger is set up in the database. When a stock keeping unit (SKU) is updated, an HTTP request towards the Varnish server is triggered. This request invalidates every cached object with the matching X-HashTwo header:

GET / HTTP/1.1
Host: www.example.com
X-HashTwo-Purge: 166412

Note the X-HashTwo-Purge HTTP header field.

Based on the VCL code above, Varnish finds the objects and purge them. After that, Varnish responds with:

HTTP/1.1 200 Purged
Date: Thu, 24 Apr 2014 17:08:28 GMT
X-Varnish: 1990228115
Via: 1.1 Varnish

The objects are now cleared.

Warning

You should protect purges with ACLs from unauthorized hosts.

Saving a Request

This chapter is for the system administration course only

Varnish offers many mechanisms to save a request. By saving a request we mean:

  1. Rescue: mechanisms to handle requests when backends are in problematic situations.
  2. Economization: mechanisms to spend less resources, i.e., send less requests to the backend.
  3. Protection: mechanisms to restrict access cache invalidation from unauthorized entities.

Table 19 shows how different mechanisms are mapped to their saving meaning. This chapter explains how to make your Varnish setup more robust by using these mechanisms.

Directors

  • Loadable VMOD
  • Contains 1 or more backends
  • All backends must be known
  • Selection methods:
    • round-robin
    • fallback
    • random
      • seeded with a random number
      • seeded with a hash key

Round-robin director example:

Varnish can have several backends defined, and it can set them together into clusters for load balancing purposes. Backend directors, usually just called directors, provide logical groupings of similar web servers by re-using previously defined backends. A director must have a name.

There are several different director selection methods available, they are: random, round-robin, fallback, and hash. The next backend to be selected depends on the selection method. The simplest directors available are the round-robin and the random director.

A round-robin director takes only a backend list as argument. This director type picks the first backend for the first request, then the second backend for the second request, and so on. Once the last backend have been selected, backends are selected again from the top. If a health probe has marked a backend as sick, a round-robin director skips it.

A fallback director will always pick the first backend unless it is sick, in which case it would pick the next backend and so on. A director is also considered a backend so you can actually stack directors. You could for instance have directors for active and passive clusters, and put those directors behind a fallback director.

Random directors are seeded with either a random number or a hash key. Next section explains their commonalities and differences.

Note

Health probes are explain in the Health Checks section.

Note

Directors are defined as loadable VMODs in Varnish 4. See the vmod_directors man page for more information.

Warning

If you declare backend servers, but do not use them, varnishd returns error by default. You can avoid this situation by turning off the runtime parameter vcc_err_unref. However, this practice is strongly discouraged. Instead, we advise to declare only what you use.

Random Directors

  • Random director: seeded with a random number
  • Hash director: seeded with hash key typically from a URL or a client identity string

Hash director that uses client identity for backend selection

The random director picks a backend randomly. It has one per-backend parameter called weight, which provides a mechanism for balancing the selection of the backends. The selection mechanism of the random director may be regarded as traffic distribution if the amount of traffic is the same per request and per backend. The random director also has a director-wide counter called retries, which increases every time the director selects a sick backend.

Both, the random and hash director select a backend randomly. The difference between these two is the seed they use. The random director is seeded with a random number, whereas the hash director is seeded with a hash key.

Hash directors typically use the requested URL or the client identity (e.g. session cookie) to compute the hash key. Since the hash key is always the same for a given input, the output of the hash director is always the same for a given hash key. Therefore, hash directors select always the same backend for a given input. This is also known as sticky session load balancing. You can learn more about sticky sessions in https://www.varnish-software.com/blog/proper-sticky-session-load-balancing-varnish.

Hash directors are useful to load balance in front of other Varnish caches or other web accelerators. In this way, cached objects are not duplicated across different cache servers.

Note

In Varnish 3 there is a client director type, which is removed in Varnish 4. This client director type is a special case of the hash director. Therefore, the semantics of a client director type are achieved using hash.backend(client.identity).

Health Checks

  • Poke your web server every N seconds
  • Affects backend selection
  • std.healthy(req.backend_hint)
  • Varnish allows at most threshold amount of failed probes within a set of the last window probes
  • threshold and window are parameters
  • Set using .probe
  • varnishlog: Backend_health

You can define a health check for each backend. A health check defines a probe to verify whether a backend replies on a given URL every given interval.

The above example causes Varnish to send a request to http://server1.example.com/healthtest every 4 seconds. This probe requires that at least 3 requests succeed within a sliding window of 5 request.

Varnish initializes backends marked as sick. .initial is another variable of .probe. This variable defines how many times the probe must succeed to mark the backend as healthy. The .initial default value is equal to .threshold – 1.

The variable Backend_health of varnishlog shows the result of a backend health probe. Issue man vsl to see its detailed syntax.

When Varnish has no healthy backend available, it attempts to use a graced copy of the cached object that a request is looking for. The next section Grace Mode explains this concept in detail.

You can also declare standalone probes and reuse them for several backends. It is particularly useful when you use directors with identical behaviors, or when you use the same health check procedure across different web applications.

Note

Varnish does NOT send a Host header with health checks. If you need that, you can define an entire request using .request instead of .url.

Note

The healthy function is implemented as VMOD in Varnish 4. req.backend.healthy from Varnish 3 is replaced by std.healthy(req.backend_hint). Do not forget to include the import line: import std;

Grace Mode

  • A graced object is an object that has expired, but is still kept in cache.
  • Grace mode is when Varnish uses a graced object.
  • Grace mode is a feature to mitigate thread pile-ups, that allows Varnish to continue serving requests when the backend cannot do it.
  • Varnish can use a graced objects in many different ways.
  • beresp.grace defines the time that Varnish keeps an object after beresp.ttl has elapsed.

To understand best grace mode, recall Figure 2, which shows the lifetime of cached objects. When Varnish is in grace mode, Varnish is capable of delivering a stale object and issues an asynchronous refresh request. When possible, Varnish delivers a fresh object, otherwise Varnish builds a response from a stale object. This procedure is also known as stale-while-revalidate.

The typical way to use grace is to store an object for several hours after its TTL has elapsed. In this way, Varnish has always a copy to be delivered immediately, while fetching a new object asynchronously. If the backend is healthy, a graced object does not get older than a few seconds (after its TTL has elapsed). If the backend is sick, Varnish may be delivering a graced object up to its maximum grace time. The following VCL code illustrates a normal usage of grace.

Graced objects are those with a grace time that has not yet expired. The grace time is stored in obj.grace, which default is 10 seconds. You can change this value by three means:

  1. by parsing the HTTP Cache-Control field stale-while-revalidate that comes from the backend,
  2. by setting the variable beresp.grace in VCL, or
  3. by changing the grace default value with varnishadm param.set default_grace <value>.

In the first case, Varnish parses stale-while-revalidate automatically, as in: "Cache-control: max-age=5, stale-while-revalidate=30". In this example, the result of the variables of the fetched objects are: obj.ttl=5 and obj.grace=30. The second and third case are self descriptive.

Note

obj.ttl and obj.grace are countdown timers. Objects are valid in cache as long as they have a positive remaining time equal to obj.ttl + obj.grace.

Timeline Example

Backend response HTTP Cache-Control header field:

"Cache-control: max-age=60, stale-while-revalidate=30"

or set in VCL:

set req.ttl = 60s;
set beresp.grace = 30s;
  • 50s: Normal delivery
  • 62s: Normal cache miss, but grace mode possible
  • 80s: Normal cache miss, but grace mode possible
  • 92s: Normal cache miss, object is removed from cache
In this timeline example, it is assumed that the object is never refreshed. If you do not want that objects with a negative TTL are delivered, set beresp.grace = 0. The downside of this is that all grace functionality is disabled, regardless any reason.

Understanding Grace

  • A request is already pending for some specific content
  • No healthy backend is available
The main goal of grace mode is to avoid requests to pile up whenever a popular object has expired in cache. As long as a request is waiting for new content, Varnish delivers graced objects instead of queuing incoming requests. These requests may come from different clients, thus, large number of clients benefit from grace mode setups.

Understanding Grace using varnishtest

Varnish-Cache/bin/varnishtest/tests/b00043.vtc:

This example shows you how the HTTP response header field Cache-Control sets max-age to ttl and stale-while-revalidate to grace. ttl and grace are attributes of cached objects. The VCL code in v1 includes these attributes in the HTTP response header fields http.ttl and http.grace that are sent to the client. c1 asserts the values of these fields.

Exercise: Grace

  1. Copy the following CGI script in /usr/lib/cgi-bin/test.cgi:

    #!/bin/sh
sleep 10
echo "Content-type: text/plain"
echo "Cache-control: max-age=10, stale-while-revalidate=20"
echo
echo "Hello world"
date

  2. Make the script executable.

  3. Issue varnishlog -i VCL_call,VCL_return in one terminal.

  4. Test that the script works outside Varnish by typing http http://localhost:8080/cgi-bin/test.cgi in another terminal.

  5. Send a single request, this time via Varnish, to cache the response from the CGI script. This should take 10 seconds.

  6. Send three requests: one before the TTL (10 seconds) elapses, another after 10 seconds and before 30 seconds, and a last one after 30 seconds.

  7. Repeat until you understand the output of varnishlog.

  8. Play with the values of max-age and stale-while-revalidate in the CGI script, and the beresp.grace value in the VCL code.

With this exercise you should see that as long as the cached object is within its TTL, Varnish delivers the cached object as normal. Once the TTL expires, Varnish delivers the graced copy, and asynchronously fetches an object from the backend. Therefore, after 10 seconds of triggering the asynchronous fetch, an updated object is available in the cache.

retry Return Action

  • Available in vcl_backend_response and vcl_backend_error
  • Re-enters vcl_backend_fetch
  • Any changes made are kept
  • Parameter max_retries safe guards against infinite loops
  • Counter bereq.retries registers how many retries are done

The retry return action is available in vcl_backend_response and vcl_backend_error. This action re-enters the vcl_backend_fetch subroutine. This only influences the backend thread, the client-side handling is not affected.

You may want to use this action when the backend fails to respond. In this way, Varnish can retry the request to a different backend. For this, you must define multiple backends.

You can use directors to let Varnish select the next backend to try. Alternatively, you may use bereq.backend to specifically select another backend.

return (retry) increments the bereq.retries counter. If the number of retries is higher than max_retries, control is passed to vcl_backend_error.

Note

In Varnish 3.0 it is possible to do return (restart) after the backend response failed. This is now called return (retry), and jumps back up to vcl_backend_fetch.

Tune Backend Properties

If a backend has not enough resources, it might be advantageous to set max_connections. So that a limited number of simultaneous connections are handled by a specific backend. All backend-specific timers are available as parameters and can be overridden in VCL on a backend-specific level.

Tip

Varnish only accepts hostnames for backend servers that resolve to a maximum of one IPv4 address and one IPv6 address. The parameter prefer_ipv6 defines which IP address Varnish prefer.

Access Control Lists (ACLs)

  • An ACL is a list of IP addresses
  • VCL programs can use ACLs to define and control the IP addresses that are allowed to purge, ban, or do any other regulated task.
  • Compare with client.ip or server.ip

An Access Control List (ACL) declaration creates and initializes a named list of IP addresses and ranges, which can later be used to match client or server IP addresses. ACLs can be used for anything. They are typically used to control the IP addresses that are allowed to send PURGE or ban requests, or even to avoid the cache entirely.

You may also setup ACLs to differentiate how your Varnish servers behave. You can, for example, have a single VCL program for different Varnish servers. In this case, the VCL program evaluates server.ip and acts accordingly.

ACLs are fairly simple to create. A single IP address or hostname should be in quotation marks, as "localhost". ACL uses the CIDR notation to specify IP addresses and their associated routing prefixes. In Varnish's ACLs the slash "/" character is appended outside the quoted IP address, for example "192.168.1.0"/24.

To exclude an IP address or range from an ACL, and exclamation mark "!" should precede the IP quoted address. For example !"192.168.1.23". This is useful when, for example, you want to include all the IP address in a range except the gateway.

Warning

If you declare ACLs, but do not use them, varnishd returns error by default. You can avoid this situation by turning off the runtime parameter vcc_err_unref. However, this practice is strongly discouraged. Instead, we advise to declare only what you use.

Content Composition

This chapter is for the web-developer course only

This chapter teaches you how to glue content from independent sources into one web page.

  • Cookies and how to work with them
  • Edge Side Includes (ESI) and how to compose a single client-visible page out of multiple objects
  • Combining ESI and Cookies
  • AJAX and masquerading AJAX requests through Varnish

A Typical Website

Most websites follow a pattern: they have easily distinguishable parts:

  • A front page
  • Articles or sub-pages
  • A login-box or "home bar"
  • Static elements, like CSS, JavaScript and graphics

To truly utilize Varnish to its full potential, start by analyzing the structure of the website. Ask yourself this:

  • What makes web pages in your server different from each other?
  • Does the differences apply to entire pages, or only parts of them?
  • How can I let Varnish to know those differences?

Beginning with the static elements should be easy. Previous chapters of this book cover how to handle static elements. How to proceed with dynamic content?

An easy solution is to only cache content for users that are not logged in. For news-papers, that is probably enough, but not for web-shops.

Web-shops re-use objects frequently. If you can isolate the user-specific bits, like the shopping cart, you can cache the rest. You can even cache the shopping cart, if you tell Varnish when to change it.

The most important lessons is to start with what you know.

Cookies

  • Be careful when caching cookies!
  • Cookies are frequently used to identify unique users, or user's choices.
  • They can be used for anything from identifying a user-session in a web-shop to opting for a mobile version of a web page.
  • Varnish can handle cookies coming from two different sources:
  • req.http.Cookie header field from clients
  • beresp.http.Set-Cookie header field from servers

By default Varnish does not cache a page if req.http.Cookie or beresp.http.Set-Cookie are present. This is for two main reasons: 1) to avoid littering the cache with large amount of copies of the same content, and 2) to avoid delivering cookie-based content to a wrong client.

It is far better to either cache multiple copies of the same content for each user or cache nothing at all, than caching personal, confidential or private content and deliver it to a wrong client. In other words, the worst is to jeopardize users' privacy for saving backend resources. Therefore, it is strongly advised to take your time to write a correct VCL program and test it thoroughly before caching cookies in production deployments.

Despite cookie-based caching being discouraged, Varnish can be forced to cache content based on cookies. If a client request contains req.http.Cookie, use return (hash); in vcl_recv. If the cookie is a Set-Cookie HTTP response header field from the server, use return (deliver); in vcl_backend_response.

Note

If you need to handle cookies, consider using the cookie VMOD from https://github.com/lkarsten/libvmod-cookie. This VMOD handles cookies with convenient parsing and formatting functions without the need of regular expressions.

Vary and Cookies

  • Not advised
  • Used to cache content that varies on cookies
  • By default, Varnish does not store responses when cookies are involved
  • The Vary response header field can be used to store responses that are based on the value of cookies
  • Cookies are widely used, but not Vary: Cookie

Varnish uses a different hash value for each cached resource. Resources with several representations, i.e. variations containing the Vary response header field, share the same hash value in Varnish. Despite this common hash value, caching based on the Vary: Cookie response header is not advised, because of its poor performance. For a more detailed explanation on Vary, please refer to the Vary subsection.

Note

Consider using Edge Side Includes to let Varnish build responses that combine content with and without cookies, i.e. combining caches and responses from the origin server.

Best Practices for Cookies

  • Remove all cookies that you do not need

  • Organize the content of your web site in a way that let you easily determine if a page needs a cookie or not. For example:

    • /common/ -- no cookies
    • /user/ -- has user-cookies
    • /voucher/ -- has only the voucher-cookie
    • etc.

  • Add the req.http.Cookie request header to the cache hash by issuing hash_data(req.http.cookie); in vcl_hash.

  • Never cache a Set-Cookie header. Either remove the header before caching or do not cache the object at all.

  • To ensure that all cached pages are stripped of Set-Cookie, finish vcl_backend_response with something similar to:

    if (beresp.ttl > 0s) {
    unset beresp.http.Set-cookie;
}

Exercise: Compare Vary and hash_data in varnishtest

In this exercise you have to use two cache techniques; first Vary and then hash_data. The exercise uses the Cookie header field, but the same rules apply to any other field.

Vary: Part 1:

  1. Write a VTC program that forces Varnish to cache client requests with cookies.
  2. Send two client requests for the same URL; one for user Alice and one for user Bob.
  3. Does Varnish use different backend responses to build and deliver the response to the client?
  4. Make the your simulated server send the Vary: Cookie response header field, then analyze the response to the client.
  5. Remove beresp.http.Vary in vcl_backend_response and see if Varnish still honors the Vary header.

Vary: Part 2:

  1. Purge the cached object for resource /cookies.php.
  2. Check if it affects all, none or just one of the objects in cache (e.g: change the value of the cookie and see if the PURGE method has purged all of them).

hash_data(): Part 1:

  1. Write another VTC program or add conditions and asserts to differentiate requests handled by Vary and hash_data().
  2. Add hash_data(req.http.Cookie); in vcl_hash.
  3. Check how multiple values of Cookie give individual cached objects.

hash_data(): Part 2:

  1. Purge the cache again and check the result after using hash_data() instead of Vary: Cookie.
This exercise is all about Vary and hash mechanisms. After this exercise, you should have a very good idea on how Vary and hash_data() work.
Solution: Handle Cookies with Vary in varnishtest

vtc/c00003.vtc

Vary and hash_data() might behave very similar at first sight and they might even seem like alternatives for handling cookies. However, cached objects are referenced in different ways.

If Varnish is forced to store responses with cookies, Vary ensures that Varnish stores resources per URL and Cookie. If Vary: Cookie is used, objects are purged in this way:

txreq -req PURGE -url "/cookie.php"

but something different is needed when using hash_data(req.http.Cookie), as you can see it in the next suggested solution.

Solution: Handle Cookies with hash_data() in varnishtest

vtc/c00004.vtc

hash_data(req.http.Cookie) adds the request header field Cookie to the hash key. So Varnish can discern between backend responses linked to a specific request header field.

To purge cached objects in this case, you have to specify the header field used in hash_data():

txreq -req PURGE -url "/cookie.php" -hdr "Cookie: user: Alice"

Exercise: Compare Vary and hash_data with HTTPie

In this exercise you have to use two cache techniques; first Vary and then hash_data(). The exercise uses the Cookie header field, but the same rules apply to any other field. For that, prepare the testbed and test with HTTPie:

  1. Copy the file material/webdev/cookies.php to /var/www/html/cookies.php.

  2. Send different requests in HTTPie changing /cookies.php and user=Alice for /article.html and user=Bob, e.g.:

    http -p hH http://localhost/cookies.php "Cookie: user=Alice"

Vary: Part 1:

  1. Write a VCL program to force Varnish to cache client requests with cookies.
  2. Send two client requests for the same URL; one for user Alice and one for user Bob.
  3. Does Varnish use different backend responses to build and deliver the response to the client?
  4. Make cookies.php send the Vary: Cookie response header field, then analyze the response to the client.
  5. Remove beresp.http.Vary in vcl_backend_response and see if Varnish still honors the Vary header.

Vary: Part 2:

  1. Purge the cached object for resource /cookies.php.
  2. Check if it affects all, none or just one of the objects in cache (e.g: change the value of the cookie and see if the PURGE method has purged all of them).

hash_data(): Part 1:

  1. Write another VCL program or add conditions to differentiate requests handled by Vary and hash_data().
  2. Add hash_data(req.http.Cookie); in vcl_hash.
  3. Check how multiple values of Cookie give individual cached objects.

hash_data(): Part 2:

  1. Purge the cache again and check the result after using hash_data() instead of Vary: Cookie.
This exercise is all about Vary and hash mechanisms. After this exercise, you should have a very good idea on how Vary and hash_data(); work.

Edge Side Includes

  • What is ESI?
  • How to use ESI?
  • Testing ESI without Varnish
ui/img/esi.png

Figure :counter:`figure`: Web page assembling using ESI via Varnish

Edge Side Includes or ESI is a small markup language for dynamic web page assembly at the reverse proxy level. The reverse proxy analyses the HTML code, parses ESI specific markup and assembles the final result before flushing it to the client. Figure 26 depicts this process.

With ESI, Varnish can be used not only to deliver objects, but to glue them together. The most typical use case for ESI is a news article with a most recent news box at the side. The article itself is most likely written once and possibly never changed, and can be cached for a long time. The box at the side with most recent news, however, changes frequently. With ESI, the article can include a most recent news box with a different TTL.

When using ESI, Varnish fetches the news article from a web server, then parses the <esi:include src="/url" /> ESI tag, and fetches the URL via a normal request. Either finding it already cached or getting it from a web server and inserting it into cache.

The TTL of the ESI element can be 5 minutes while the article is cached for two days. Varnish delivers the two different objects in one glued page. Thus, Varnish updates parts independently and makes possible to combine content with different TTL.

Basic ESI usage

Enabling ESI in Varnish is simple enough:

To include a page in another, the <esi:include> ESI tag is used:

<esi:include src="/url" />

You can also strip off cookies per ESI element. This is done in vcl_recv.

Varnish only supports three ESI tags:

  • <esi:include>: calls the page defined in the src attribute and replaces the ESI tag with the content of src.

  • <esi:remove>: removes any code inside this opening and closing tag.

  • <!--esi ``(content) -->``: Leaves (content) unparsed. E.g., the following does not process the <esi:include> tag:

    <!--esi
    This ESI tag is not processed: <esi:include src="example">
-->

Note

Varnish outputs ESI parsing errors in varnishstat and varnishlog.

Understanding ESI in varnishtest

Varnish-Cache/bin/varnishtest/tests/e00004.vtc:

e00004.vtc shows how ESI substitution works. When Varnish reads <!--esi <esi:include src="/body"/> -->, it triggers a request with URL /body. The result of this request replaces the <!--esi --> tag.

We have counted the expected body length after the substitution and assert it in the VTC, but if you do not trust us, you can easily see the replacement by executing:

varnishtest -v e00004.vtc | grep "chunk|"

In the result:

**** c1    0.4 chunk| \n
**** c1    0.4 chunk| \t\t<html>\n
**** c1    0.4 chunk| \t\tBefore include\n
**** c1    0.4 chunk| \t\t
**** c1    0.4 chunk| \n
**** c1    0.4 chunk| \t\tIncluded file\n
**** c1    0.4 chunk| \t \n
**** c1    0.4 chunk| \t\tAfter include\n
**** c1    0.4 chunk| \t

you can see the HTML document after ESI has been processed.

Example: Using ESI

Copy material/webdev/esi-date.php to /var/www/html/. This file contains an ESI include tag:

Copy material/webdev/esi-date.cgi to /usr/lib/cgi-bin/. This file is a simple CGI that outputs the date of the server:

For ESI to work, load the following VCL code:

Then reload Varnish and issue the command http http://localhost/esi-date.php. The output should show you how Varnish replaces the ESI tag with the response from esi-date.cgi. Note the different TTLs from the glued objects.

Exercise: Enable ESI and Cookies

  1. Use material/webdev/esi-top.php and material/webdev/esi-user.php to test ESI.
  2. Visit esi-top.php and identify the ESI tag.
  3. Enable ESI for esi-top.php in VCL and test.
  4. Strip all cookies from esi-top.php and make it cache.
  5. Let esi-user.php cache too. It emits Vary: Cookie, but might need some help.

Try to avoid return (hash); in vcl_recv and return (deliver); in vcl_backend_response as much as you can. This is a general rule to make safer Varnish setups.

During the exercise, make sure you understand all the cache mechanisms at play. You can also try removing the Vary: Cookie header from esi-user.php.

You may also want to try PURGE. If so, you have to purge each of the objects, because purging just /esi-top.php does not purge /esi-user.php.

Testing ESI without Varnish

  • Test ESI Using JavaScript to fill in the blanks.

During development of different web pages to be ESI-glued by Varnish, you might not need Varnish all the time. One important reason for this, is to avoid caching during the development phase. There is a solution based on JavaScript to interpret ESI syntax without having to use Varnish at all. You can download the library at the following URL:

Once downloaded, extract it in your code base, include esiparser.js and include the following JavaScript code to trigger the ESI parser:

$(document).ready( function () { do_esi_parsing(document); });

Masquerading AJAX requests

ui/img/ajaxok.png ui/img/ajaxko.png
This works This does not work
With AJAX it is not possible by default to send requests across another domain. This is a security restriction imposed by browsers. If this represents an issue for your web pages, you can be easily solve it by using Varnish and VCL.

Exercise: write a VCL that masquerades XHR calls

material/webdev/ajax.html

Use the provided ajax.html page. Note that function getNonMasqueraded() fails because the origin is distinct to the google.com domain. Function getMasqueraded() can do the job if a proper VCL code handles it. Write the VCL code that masquerades the Ajax request to http://www.google.com/robots.txt.

Solution: write a VCL that masquerades XHR calls

vcl/solution-vcl_fetch-masquerade-ajax-requests.vcl

Note that the getMasqueraded() works now after being processed in vcl_recv().

Varnish Plus Software Components

The Varnish Plus offer of software products includes:

For more information about the complete Varnish Plus offer and their documentation, please visit:

Varnish Administration Console (VAC)

  • Single point of control for simultaneous administration of multiple Varnish Cache servers
  • VAC provides:
    • GUI
    • API
    • Super Fast Purger
  • VAC has its own documentation

The Varnish Administration Console (VAC) consists of a GUI and an API. VAC is most commonly used in production environments where real-time graphs and statistics help identify bottlenecks and issues within Varnish Cache servers. VAC is a management console for groups of Varnish Cache servers, also known as cache groups. A cache group is a collection of Varnish Cache servers that have identical configuration. Attributes on a cache group includes:

  • One or more Varnish Cache servers
  • Active Varnish Configuration Language (VCL) file
  • Homogeneous parameter configuration across all servers in a group

VAC distributes and store VCL files for you. A parameter set is a list of Varnish cache parameters. These parameters can be applied to one or more cache groups simultaneously, as long as all cache groups consist of cache servers of the same version.

VAC ships with a JSON-based RESTful API to integrate your own systems with the VAC. All actions performed via the user interface can be replicated with direct access to the API. This includes fetching all real-time graph data.

The Super Fast Purger is a high performance cache invalidation delivery mechanism for multiple installations of Varnish. Super Fast Purger is capable of distributing purge requests to cache groups across data centers via the Restful interface. Super Fast Purger uses HMAC as security mechanism to protect your purge requests and thus ensure data integrity.

In order to install VAC on either Debian/Ubuntu or Red Hat Enterprise, one would require access to the Varnish Plus Software repository. As a Varnish Plus customer, you have access to the installation guide document. This document has instructions to install, configure, maintain, and troubleshoot your VAC installation If you have any questions on how to set up your repository or where to obtain the VAC installation guide, please ask the instructor or send us an email to support@varnish-software.com.

Figures 26, Figures 27 and Figures 28 show screenshots of the GUI. You may also be interested in trying the VAC demo at https://vacdemo.varnish-software.com. The instructor of the course provides you the credentials.

Overview Page of the Varnish Administration Console

ui/img/vac_screenshot_1.png

Figure :counter:`figure`: Overview page of the Varnish Administration Console

Configuration Page of the Varnish Administration Console

ui/img/vac_screenshot_2.png

Figure :counter:`figure`: Configuration page of the Varnish Administration Console

Banning Page of the Varnish Administration Console

ui/img/vac_screenshot_3.png

Figure :counter:`figure`: Banning page of the Varnish Administration Console

Varnish Custom Statistics (VCS)

  • Data stream management system (DSMS) for your Varnish servers
  • Real-time statistics engine to aggregate, display and analyze user web traffic
  • Provides an API to retrieve statistics
  • Provides a GUI that presents lists and charts to get a quick overview of the key metrics that matters you

Figure :counter:`figure`: VCS Data Flow

Varnish Custom Statistics (VCS) is our data stream management system (DSMS) implementation for Varnish. VCS allows you to analyze the traffic from multiple Varnish servers in near real-time to compute traffic statistics and detect critical conditions. This is possible by continuously extracting transactions with the vcs-key tags in your VSL. Thus, VCS does not slow down your Varnish servers.

You can add as many custom vcs-key tags as you need in your VCL code. This allows you to define your own metrics.

VCS can be used to produce statistical data or even apply complex event processing techniques. Thus, VCS offers endless opportunities for tracking all aspects of websites' behavior. Typical cases include:

  • A/B testing
  • Measuring click-through rate
  • Track slow pages and cache misses
  • Analyze what is "hot" right now in a news website
  • Track changes in currency conversions in e-commerce
  • Track changes in Stock Keeping Units (SKUs) <behavior in e-commerce
  • Track number of unique consumers of HLS/HDS/DASH video streams

VCS is a great tool when you want to test some functionality in your backend. For that, you can separate your requests into different groups, handle their requests accordingly, analyze the results and conclude whether your new functionality should be applied to all groups. This type of tests are called A/B testing. If you want to learn how to implement A/B testing in Varnish, please refer to https://www.varnish-software.com/blog/live-ab-testing-varnish-and-vcs.

Figure 31 and Figure 32 are screenshots of the VCS GUI. These screenshots are from the demo on http://vcsdemo.varnish-software.com. Your instructor can provide you credential for you to try the demo online.

Note

For further details on VCS, please look at its own documentation at https://www.varnish-software.com/resources/.

VCS Data Model

  • Represents a finite relation from an infinite stream.
  • Uses time-based tumbling windows
  • API to query data model
  • API outputs in JSON and JSONP format

VCS uses the time-based tumbling windows technique to segment the data stream into finite parts. These windows are created based on the vcs-key tag that you specify in your VCL code. Each window aggregates the data within a configurable period of time.

Table 20 shows the data model in VCS. This table is basically a representation of two windows seen as two records in a conventional database. In this example, data shows two windows of 30 second based on the example.com vcs-key. For presentation purposes in this page, the distribution of this table is of a database that grows from left to right.

The VCS data model has the following fields:

vcs-key
common key name for transactions making this record
timestamp
Timestamp at the start of the window
n_req
Number of requests
n_req_uniq
Number of unique requests, if configured
n_miss
Number of backend requests (i.e. cache misses) Number of hits can be calculated as n_hit = n_req - n_miss
avg_restart
Average number of VCL restarts triggered per request
n_bodybytes
Total number of bytes transferred for the response bodies
ttfb_miss
Average time to first byte for requests that ended up with a backend request
ttb_hit
Average time to first byte for requests that were served directly from varnish cache
resp_1xx -- resp_5xx
Counters for response status codes.
reqbytes
Number of bytes received from clients.
respbytes
Number of bytes transmitted to clients.
berespbytes
Number of bytes received from backends.
bereqbytes
Number of bytes transmitted to backends.

You can think of each window as a record of a traditional database that resides in memory. This database is dynamic, since the engine of VCS updates it every time a new window (record) is available. VCS provides an API to retrieve this data from the table above in JSON format:

{
 "example.com": [
     {
         "timestamp": "2013-09-18T09:58:30",
         "n_req": 76,
         "n_req_uniq": "NaN",
         "n_miss": 1,
         "avg_restarts": 0.000000,
         "n_bodybytes": 10950,
         "ttfb_miss": 0.000440,
         "ttfb_hit": 0.000054,
         "resp_1xx": 0,
         "resp_2xx": 76,
         "resp_3xx": 0,
         "resp_4xx": 0,
         "resp_5xx": 0,
         ...
     },
     {
         "timestamp": "2013-09-18T09:58:00",
         "n_req": 84,
         "n_req_uniq": "NaN",
         "n_miss": 0,
         "avg_restarts": 0.000000,
         "n_bodybytes": 12264,
         "ttfb_miss": "NaN",
         "ttfb_hit": 0.000048,
         "resp_1xx": 0,
         "resp_2xx": 84,
         "resp_3xx": 0,
         "resp_4xx": 0,
         "resp_5xx": 0,
         ...
     },

     ...
     ]
}

VCS API

  • API provides ready-to-use queries
  • Queries over HTTP
  • Top-most sorting
  • Results in JSON and JSONP format

Examples:

For vcs-key with names ending with .gif, retrieve a list of the top 10:

/match/(.*)%5C.gif$/top

Find a list of the top 50 slowest backend requests:

/all/top_ttfb/50

The VCS API queries the VCS data model and the output is in JSON format. The API responds to requests for the following URLs:

/key/<vcs-key>
Retrieves stats for a single vcs-key. <vcs-key> name must be URL encoded.
/match/<regex>
Retrieves a list of vcs-key matching the URL encoded regular-expression. Accepts the query parameter verbose=1, which displays all stats collected for the <vcs-keys> matched.
/all
Retrieves a list of all the <vcs-keys> currently in the data model.

For /match/<regex> and /all, VCS can produce sorted lists. For that, you can append one of the following sorting commands.

/top
Sort based on number of requests.
/top_ttfb
Sort based on the ttfb_miss field.
/top_size
Sort based on the n_bodybytes field.
/top_miss
Sort based on the n_miss field.
/top_respbytes
Sort based on number of bytes transmitted to clients.
/top_reqbytes
Sort based on number of bytes received from clients.
/top_berespbytes
Sort based on number of bytes fetched from backends.
/top_bereqbytes
Sort based on number of bytes transmitted to backends.
/top_restarts
Sort based on the avg_restarts field.
/top_5xx, /top_4xx, ..., /top_1xx
Sort based on number of HTTP response codes returned to clients for 5xx, 4xx, 3xx, etc.
/top_uniq
Sort based on the n_req_uniq field.

Further, a /k parameter can be appended, which specifies the number of keys to include in the top list. If no k value is provided, the top 10 is displayed.

Note

For installation instructions, please refer to http://files.varnish-software.com/pdfs/installation-guide_vcs-latest.pdf. Once you have installed all necessary components, take a look at the man pages of vstatd and vstatdprobe for more documentation.

Screenshots of GUI

ui/img/vcsui_header_2.png

Figure :counter:`figure`: Header of Varnish Custom Statistics



ui/img/vcs-ui-chart.png

Figure :counter:`figure`: Summary of metrics with time based graphs

Varnish High Availability (VHA)

  • Content replicator
  • Increases resiliency and performance
  • Two-server, circular, multi-master replication
  • Requests to replicate content against Varnish servers, not the backend

Figure :counter:`figure`: VHA Sequence Diagram

The Varnish High Availability agent (vha-agent) is a content replicator with the aim of copying the cached objects from an origin Varnish server to a neighboring Varnish server. This increases resiliency and performance, specially when backend traffic surges.

vha-agent reads the log of Varnish, and for each object insertion detected it fires a request to the neighboring Varnish server. This server fetches the object from the origin Varnish server. As a result, the same object is cached in both servers with only one single backend fetch.

This solution requires vha-agent to be installed on the origin Varnish server, and some simple VCL configuration on the replicated Varnish server. Ideally, vha-agent is installed on both servers so they can both replicate object insertions from each other in an active/active configuration.

Typical uses of VHA include:

  • Business critical Varnish Plus installations
  • Any multi-cache Varnish setup
  • Multi node CDN POP installations

The replication of cached objects may bring the need for multiple cache invalidation. For that purpose, you can use the Varnish Administration Console (VAC). Remember: you should define the rules on how to invalidate cached objects before caching them in production environments.

SSL/TLS Support

  • SSL/TLS support on both the HTTP backend and client side.

  • Easy configuration:

    backend default {
    .host = "host.name";
    .port = "https";                # This defaults to https when SSL
    .ssl = 1;                       # Turns on SSL support
    .ssl_nosni = 1;                 # Disable SNI extension
    .ssl_noverify = 1;              # Don't verify peer
}

Varnish Plus allows you to improve your website security without having to rely on third-party solutions. SSL/TLS support allows you to encrypt and secure communication on both the front and backend as Varnish acts as both HTTP server and client. On the client side, the HTTP server intercepts web requests before they reach a web server. The SSL/TLS support on this side enables traffic encryption between the client and Varnish.

On the backend, the HTTP client fetches content missing in the cache from the web server. This enables content to be fetched over the encrypted SSL/TLS, which particularly benefits customers who run a fully encrypted data center or have web servers that reside in a different location to their Varnish Plus servers.

Appendix A: Resources

Community driven:

Commercial:

Appendix B: Varnish Programs

Varnish Shared memory Log (VSL) tools:

  • varnishlog
  • varnishncsa
  • varnishhist
  • varnishtop

Administration:

  • varnishadm

Global counters:

  • varnishstat

Misc:

  • varnishtest
varnishlog, varnishadm and varnishstat are explained in the Examining Varnish Server's Output chapter. Next sections explain varnishtop, varnishncsa, and varnishhist.

varnishtop

$varnishtop -i BereqURL,RespStatus

list length 5                                                 trusty-amd64

     7.20 RespStatus     200
     5.26 RespStatus     404
     0.86 BereqURL       /test.html
     0.68 BereqURL       /
     0.39 BereqURL       /index.html
  • Group tags and tag-content by frequency

varnishtop groups tags and their content together to generate a sorted list of the most frequently appearing tag/tag-content pair. This tool is sometimes overlooked, because its usefulness is visible after you start filtering. The above example lists status codes that Varnish returns.

Two of the perhaps most useful variants of varnishtop are:

  • varnishtop -i BereqURL: creates a list of URLs requested at the backend. Use this to find out which URL is the most requested.
  • varnishtop -i RespStatus: lists what status codes Varnish returns to clients.

You may also combine taglist as in the above example. Even more, you may apply Query Language -q options. For example, varnishtop -q 'respstatus > 400' shows you counters for responses where client seem to have erred.

Some other possibly useful examples are:

  • varnishtop -i ReqUrl: displays what URLs are most frequently requested from clients.
  • varnishtop -i ReqHeader -C -I 'User-Agent:.*Linux.*': lists User-Agent headers that contain the ignoring case Linux string. This example is useful for Linux users, since most web browsers in Linux report themselves as Linux.
  • varnishtop -i RespStatus: lists status codes received in clients from backends.
  • varnishtop -i VCL_call: shows what VCL functions are used.
  • varnishtop -i ReqHeader -I Referrer shows the most common referrer addresses.

varnishncsa

10.10.0.1 - - [24/Aug/2008:03:46:48 +0100] "GET \
http://www.example.com/images/foo.png HTTP/1.1" 200 5330 \
"http://www.example.com/" "Mozilla/5.0"

If you already have tools in place to analyze NCSA Common log format, varnishncsa can be used to print the VSL in this format. varnishncsa dumps everything pointing to a certain domain and subdomains.

Filtering works similar to varnishlog.

varnishhist

1:1, n = 71                                                    localhost

                      #
                      #
                      #
                      #
                      ##
                     ###
                     ###
                     ###
                     ###
                     ###
                  |  ###
                  |  ###
                | |  ###
                |||| ###                    #
                |||| ####                   #
                |##|#####   #           #   #  # #
+-------+-------+-------+-------+-------+-------+-------+-------+-------
|1e-6   |1e-5   |1e-4   |1e-3   |1e-2   |1e-1   |1e0    |1e1    |1e2

The varnishhist utility reads the VSL and presents a continuously updated histogram showing the distribution of the last n requests. varnishhist is particularly useful to get an idea about the performance of your Varnish Cache server and your backend.

The horizontal axis shows a time range from 1e-6 (1 microsecond) to 1e2 (100 seconds). This time range shows the internal processing time of your Varnish Cache server and the time it takes to receive a response from the backend. Thus, this axis does not show the time perceived at the client side, because other factors such as network delay may affect the overall response time.

Hits are marked with a pipe character ("|"), and misses are marked with a hash character ("#"). These markers are distributed according to the time taken to process the request. Therefore, distributions with more markers on the left side represent a faster performance.

When the histogram grows in vertical direction more than what the terminal can display, the 1:m rate changes on the left top corner. Where m represent the number of times that each marker is found. On the right top corner, you can see the name of the host.

Exercise: Try the tools

  • Send a few requests to Varnish using http -p hH http://localhost/
  • verify you have some cached objects using varnishstat
  • look at the communication with the clients, using varnishlog. Try sending various headers and see them appear in varnishlog.
  • Install siege
  • Run siege against localhost while looking at varnishhist

Appendix C: Extra Material

This appendix contains code needed for some exercises.

ajax.html

article.php

cookies.php

esi-top.php

esi-user.php

httpheadersexample.php

purgearticle.php

test.php

set-cookie.php

VCL Migrator from Varnish 3 to Varnish 4

The script aims to replace most of the syntactical changes in VCL code from Varnish 3 to Varnish 4, but it is not exhaustive. That said, you should use it under your own responsibility.

You can download the script from https://github.com/fgsch/varnish3to4. Usage and up-to-date details about the script is at the same web address.

Appendix D: VMOD Development

  • VMOD basics
  • varnishtest
< Hello, World! >

    \   ^__^
     \  (oo)\_______
        (__)\       )\/\
                   ||----w |
            ||     ||

This appendix explains the concepts you should know to develop your own VMODs. The appendix takes you through the simplest possible VMOD: the Hello, World VMOD.

To learn most out of this appendix, you should have understood at least the following chapters of this book: Design Principles, Getting Started, VCL Basics, VCL Built-in Subroutines.

VMOD Basics

  • What is a VMOD
  • When to use in-line C or VMODs?
  • The workspace memory model
  • varnishtest

A VMOD is a shared library with some C functions which can be called from VCL code. The standard (std) VMOD, for instance, is a VMOD included in Varnish Cache. We have already used the std VMOD in this book to check whether a backend is healthy with by calling std.healthy().

VCL is the domain specific language of Varnish. This language is very powerful and efficient for most tasks of a cache server. However, sometimes you might need more functionalities, e.g., look up an IP address in a database.

VCL allows you to add inline C code, but this is not the most convenient approach. If you use the memory management API provided by Varnish, your VMODs are generally easier to maintain, more secure, and much easier to debug in collaboration with other developer. In addition, VMODs do not require to reload VCL files to take effect.

When writing VMODs, you should test them towards varnishtest. In fact, we recommend you first to write your tests as part of your design, and then implement your VMOD. Therefore, we introduce next varnishtest before proceeding with the implementation the VMOD itself.

varnishtest script program

  • Script driven program used to test programs written in VCL and VMODs

With varnishtest you can create mock-ups of clients and origin servers to interact with your Varnish installation. This is useful to simulate transactions and provoke a specific behavior. You can use varnishtest when writing VCL code or VMODs. varnishtest is also useful to reproduce bugs when filing a bug report.

The best way to learn how to create Varnish tests is by running the ones included in Varnish Cache and then write your own tests based on them. Tests of Varnish Cache are in bin/varnishtest/tests/. You should also take a look at the README file bin/varnishtest/tests/README to learn about the naming convention.

VTC

helloworldtest.vtc:

varnishtest "Hello, World"

server s1 {
   rxreq
   txresp
} -start

varnish v1 -vcl+backend {
   sub vcl_deliver {
      set resp.http.hello = "Hello, World";
   }
} -start

client c1 {
   txreq -url "/"
   rxresp
   expect resp.http.hello == "Hello, World"
}

varnish v1 -expect cache_miss == 0
varnish v1 -expect cache_hit == 0

client c1 -run

varnish v1 -expect cache_miss == 1
varnish v1 -expect cache_hit == 0

client c1 -run
client c1 -run

varnish v1 -expect cache_miss == 1
varnish v1 -expect cache_hit == 2

varnishtest does not follow the unit testing framework (up/test/assert/tear down) nor behavior-driven development (given/when/then). Depending on your use case, there might be test preparations, executions and assertions all over the place. VTC is not compiled but simply interpreted on the fly. When run, the above script simulates an origin server s1, starts a real Varnish instance v1, and simulates a client c1.

server s1 declares a simulated origin server that receives a request rxreq, and transmits a response txresp. The -start directive makes an instance of that declaration to be started. -start can also be at a later point in your code as server s1 -start.

varnish v1 -vcl+backend declares an instance of your real Varnish server and loads the VCL code inside the brackets. varnishtest controls v1 through the manager process (see The Parent Process: The Manager). The +backend directive injects the backend (origin server s1) to the VCL code. Alternatively, you might want to add backends manually, for example:

varnish v1 -vcl {
   backend default {
      .host = "${s1_addr}";
      .port = "${s1_port}";
   }
}

Once s1 is started, the macros ${s1_addr} and ${s1_port} with the IP address and port of your simulated backend are automatically made available. Since varnishtest launches a real varnishd instance, it is possible to use a real backend instead of mock servers. Thus, you can test your actual backend.

client c1 declares a simulated client that transmits a request txreq for the slash URL: -url "/". c1 receives a response rxresp.

varnishtest supports assertions with the keyword expect. For example, c1 expects the response header field resp.http.hello with value Hello, World. Assertions can be inside the declaration of the origin server and client, but not inside the Varnish server. Since Varnish is a proxy, checking requests and responses in it is irrelevant. Instead, you have access to the counters exposed by varnishstat at any time:

varnish v1 -expect counter == value

Assertions are evaluated by reading the shared memory log, which ensures that your assertions are tested against a real Varnish server. Therefore, Varnish tests might take a bit longer than what you are used to in other testing frameworks. Finally, client c1 -run starts the simulated client c1.

You will learn how to extend this script to test your VMOD later in this chapter, but before, run and analyze the output of helloworldtest.vtc to understand varnishtest better.

Run Your Varnish Tests

$varnishtest helloworldtest.vtc
   top  TEST helloworldtest.vtc passed (1.554)

To run your test, you simply issue varnishtest helloworldtest.vtc By default, varnishtest outputs the summary of passed tests, and a verbose output for failed tests. It is strongly recommended that you look at the verbose output to understand what happens under the hood. For that, you run varnishtest with the -v option.

Later in this appendix, when testing VMODs, we will call make check instead of typing varnishtest in the command line. make check calls varnishtest with the needed options.

Note

man varnishtest shows you all options

Hello, World! VMOD

  • Build Varnish Cache from source
  • Try libvmod-example

VMODs require the source code from Varnish Cache that you are running. The easiest way to be sure you have everything in place, is to build your Varnish Cache from source. Although you can also develop VMODs against a Varnish installation from a package repository. The git repository and building instructions are at https://github.com/varnish/Varnish-Cache.git.

Once you have built Varnish Cache, build the libvmod-example from https://github.com/varnish/libvmod-example by following the instructions in the repository. When building the libvmod-example, make sure that its branch matches with the branch version of Varnish-Cache. For example, to build libvmod-example against Varnish Cache 4.0, make sure that you checkout the branch 4.0 in both, the libvmod-example and Varnish-Cache.

Next, we explain the content inside libvmod-example.

Declaring and Documenting Functions

  • VCC: VCL to C Compiler

vmod_example.vcc:

$Module example
$Init init_function
$Function STRING hello(STRING)

vcc_if.h:

struct VCL_conf;
struct vmod_priv;

/* Functions */
VCL_STRING vmod_hello(VRT_CTX, VCL_STRING);
int init_function(struct vmod_priv *, const struct VCL_conf *);

In vmod_example.vcc you declare the module name, initialization function and other functions you need. Definitions are stored in files with .vcc extension. Please note the $ sign leading the definitions in the vmod_example.vcc.

$Module example

The first line gives the name to the module.

$Init init_function

The second line declares an optional initial function, which is called when a VCL program loads this VMOD.

$Function STRING hello(STRING)

The third line declares the only function in this VMOD.

The source tree is based on autotools to configure the building. When you run the Makefile,it passes vmod_example.vcc to the script vmodtool.py (included in Varnish Cache) and translates the VCC code to C. The translation is stored in vcc_if.h and vcc_if.c, and since they are machine generated, you should not modify them.

The .vcc file also contains the documentation that is used to generate the man page of your VMOD. The documentation is parsed by vmodtool.py and stored in src/vmod_example.man.rst. This RST file is then used to create the man page when compiling your VMOD.

We advise you to write the documentation of your VMOD in the .vcc file, but if you prefer to have it in another location, such as the README.rst file, you can change the source for documentation in Makefile.am. For example, change from:

vmod_example.3: src/vmod_example.man.rst

to:

vmod_example.3: README.rst

Next, you implement the hello functions in the vmod_example.c.

Implementing Functions

vmod_example.c:

VCL_STRING
vmod_hello(const struct vrt_ctx *ctx, VCL_STRING name)
{
   char *p;
   unsigned u, v;

   u = WS_Reserve(ctx->ws, 0); /* Reserve some work space */
   p = ctx->ws->f;         /* Front of workspace area */
   v = snprintf(p, u, "Hello, %s", name);
   v++;
   if (v > u) {
      /* No space, reset and leave */
      WS_Release(ctx->ws, 0);
      return (NULL);
   }
   /* Update work space with what we've used */
   WS_Release(ctx->ws, v);
   return (p);
}
You reserve the memory you need by calling WS_Reserve(ctx->ws, 0). It is important to release the memory you used with WS_Release(ctx->ws, v), otherwise you are introducing a memory leak.

The Workspace Memory Model

Figure :counter:`figure`: Work Space Memory Model

Every worker thread has its own workspace ws in virtual memory. This workspace is a contiguous char array defined in cache/cache.h as:

struct ws {
   unsigned                magic;
   #define WS_MAGIC        0x35fac554
   char                    id[4];          /* identity */
   char                    *s;             /* (S)tart of buffer */
   char                    *f;             /* (F)ree/front pointer */
   char                    *r;             /* (R)eserved length */
   char                    *e;             /* (E)nd of buffer */
};

magic and WS_MAGIC are used for sanity checks by workspace functions. The id field is self descriptive. The parts that most likely you are interested in are the SFRE fields.

s and e point to the start and end of the char array respectively. f points to the currently available memory, it can be seen as a head that moves forward every time memory is allocated. f can move up to the end of the buffer pointed by e.

r points to the reserved memory space of the workspace. This space is reserved to allow incremental allocation. You should remember to release this space by calling WS_Release(struct ws *ws, unsigned bytes) once your VMOD does not need it any longer.

The cache/cache.h is automatically included when you compile your .vcc file. Next, we describe in detail the headers that are included in vmod_example.c.

Headers

#include "vrt.h"
#include "cache/cache.h"

#include "vcc_if.h"

The vrt.h header provides data structures and functions needed by compiled VCL programs and VMODs. cache.h declares the function prototypes for the workspace memory model among others. These functions are implemented in cache_ws.c.

cache.h:

void WS_Init(struct ws *ws, const char *id, void *space, unsigned len);
unsigned WS_Reserve(struct ws *ws, unsigned bytes);
void WS_MarkOverflow(struct ws *ws);
void WS_Release(struct ws *ws, unsigned bytes);
void WS_ReleaseP(struct ws *ws, char *ptr);
void WS_Assert(const struct ws *ws);
void WS_Reset(struct ws *ws, char *p);
char *WS_Alloc(struct ws *ws, unsigned bytes);
void *WS_Copy(struct ws *ws, const void *str, int len);
char *WS_Snapshot(struct ws *ws);
int WS_Overflowed(const struct ws *ws);
void *WS_Printf(struct ws *ws, const char *fmt, ...) __printflike(2, 3);

vcc_if.h is generated out from the definitions in your .vcc file. This header contains the declaration of your VMOD functions in C code.

Exercise: Build and Test libvmod_example

$./autogen.sh
$./configure
$make
$make check
#make install
  • Examine how libvmod_example is imported in src/tests/test01.vtc

The source tree is based on travis and Autotools to configure the building. More detailed building instructions are in: https://github.com/varnish/libvmod-example/blob/master/README.rst

Note that make check calls varnishtest with the needed options.

Cowsay: Hello, World!

  1. Use rename-vmod-script to rename libvmod-example
  2. Create tests
  3. Define and document functions in vmod_cowsay.vcc file
  4. Implement functions in vmod_cowsay.c file
  5. make, make check and make install your VMOD

In this subsection you will learn how to build your own VMOD on top of libvmod-example. rename-vmod-script prepares a basic VMOD that can be extended to your needs:

$./rename-vmod-script your_vmod_name

We have prepared a cowsay VMOD for you to follow easier this subsection, but you can create your own VMOD from scratch as we explain further. The cowsay VMOD adds the output of the Linux program, which generates ASCII pictures of a cow or a different animal with a message. You can download the code of the cowsay VMOD from https://github.com/aondio/libvmod-cowsay.git.

Cowsay Varnish Tests

test01.vtc:

varnishtest "Test cowsay header"

server s1 {
       rxreq
       txresp
} -start

varnish v1 -vcl+backend {
        import cowsay from "${vmod_topbuild}/src/.libs/libvmod_cowsay.so";
        sub vcl_recv {
            if (req.url ~ "/cowsay") {
                set req.http.x-cow = cowsay.cowsay_canonical();
                }
        }

        sub vcl_deliver {
                set resp.http.x-cow = req.http.x-cow;
        }
} -start

client c1 {
txreq -url "/cowsay"
rxresp
expect resp.http.x-cow == "Cowsay: Hello World!"
} -run

Snippet of test02.vtc:

sub vcl_recv {
        if (req.url ~ "/cowsay") {
                return(synth(700, "OK"));
        }
}

sub vcl_synth {
        if (resp.status == 700) {
                set resp.status = 200;
                set resp.http.Content-Type = "text/plain; charset=utf-8";
                synthetic(cowsay.cowsay_vsb());
                return (deliver);
         }
}

...

client c1 {
        txreq -url "/cowsay"
        rxresp
        expect resp.body == {** mybody **
 ^__^
 (oo)\_______
 (__)\       )\/\
      ||----w |
      ||     ||
}
} -run

Snippet of test03.vtc:

sub vcl_recv {
        if (req.url ~ "/cowsay" || req.url ~ "/bunny") {
                return(synth(700, "OK"));
        }
}

sub vcl_synth {
        if (resp.status == 700) {
                set resp.status = 200;
                set resp.http.Content-Type = "text/plain; charset=utf-8";
                if (req.url ~ "/cowsay"){
                    synthetic(cowsay.cowsay_friends("cow", "moo"));
                }
                if (req.url ~ "/bunny"){
                    synthetic(cowsay.cowsay_friends("bunny", "Varnish"));
                }
                return (deliver);
                }
        }

...

client c1 {
        txreq -url "/cowsay"
        rxresp
        expect resp.body == {** moo **

 ^__^
 (oo)\_______
 (__)\       )\/\
      ||----w |
      ||     ||
}
} -run

client c2 {
        txreq -url "/bunny"
        rxresp
        expect resp.body == {** Varnish **
 (\/)
 (..)
(")(")
}
} -run

We advise you to start designing your tests. test01.vtc, test02.vtc and test03.vtc are examples in https://github.com/franciscovg/libvmod-cowsay.git. test01.vtc shows how the HTTP response header field is assigned. When client c1 requests the /cowsay URL, Varnish server v1 assigns the output of the VMOD function cowsay_canonical() to the HTTP request header field req.http.x-cow.

test02.vtc shows how to alter the message body in the vcl_synth subroutine. In this second test, we use cowsay.cowsay_vsb(), which return the cowsay ASCII picture. The important difference between cowsay_canonical() and cowsay.cowsay_vsb() is the library used to manipulate the returned string. We discuss this difference later in this section.

test03.vtc reuses most of test02.vtc. test03.vtc replaces cowsay.cowsay_vsb() for cowsay.cowsay_friends(arg1, arg2), and adds some basic conditions. cowsay.cowsay_friends() returns the ASCII figure given in arg1, which utters arg2. After the design of your tests, you declare the functions in the .vcc file and implement them in the .c file.

Exercise: Add Assertions To Your Varnish Tests
  • Add more assertions using the keyword expect to check values of Varnish counters

vmod_cowsay.vcc

$Function STRING cowsay_canonical()
$Function STRING cowsay_vsb()
$Function STRING cowsay_friends(STRING, STRING)

cowsay_canonical() uses the canonical string library, and cowsay_vsb() uses the Varnish String Buffer string library. The VSB library is very useful to manipulate strings. Therefore we recommend you to use it instead of the canonical string libraries.

Note

Although VCL allows it, we do not recommend to assign multiple lines to HTTP context header fields.

vmod_cowsay.c

int
init_function(struct vmod_priv *priv, const struct VCL_conf *conf)
{
     /* init global state valid for the whole VCL life */
     cow =
         "\n ^__^\n"
         " (oo)\\_______\n"
         " (__)\\       )\\/\\\n"
         "      ||----w |\n"
         "      ||     ||\n";
     /* this 'cow' is now available for every other functions that will
        be defined in this vmod */
     return (0);
}

VCL_STRING
vmod_cowsay_friends(VRT_CTX, VCL_STRING animal, VCL_STRING talk)
{
     unsigned  u;
     struct vsb *vsb;
     u = WS_Reserve(ctx->ws, 0);
     vsb = VSB_new(NULL, ctx->ws->f, u, VSB_AUTOEXTEND);
     if(!strcmp(animal, "cow")) {
             VSB_printf(vsb, "** %s **\n", talk);
             VSB_cat(vsb, cow);
     }

     if(!strcmp(animal, "bunny")) {
             VSB_printf(vsb, "** %s **\n", talk);
             VSB_cat(vsb, baby_bunny());
     }
     VSB_finish(vsb);
     WS_Release(ctx->ws, VSB_len(vsb) + 1);
     return (vsb->s_buf);
}

We explain here two relevant functions in this file. The first is the init_function(), where we declare a global variable holding the cow figure. The second part is the vmod_cowsay_friends() function, where we use string manipulation functions provided by VSB. vmod_cowsay_vsb() is a simplified version of vmod_cowsay_friends(). The implementation of cowsay_canonical() is practically the same as vmod_hello().

Finally, it is time to make, make check and make install your VMOD. Note that make check calls varnishtest with the needed options.

Resources

The best way to learn more about VMODs is by writing them and seeing how others VMOD works. There are many VMODs written by the Varnish community and Varnish Software. Please take a look at the list in https://www.varnish-cache.org/vmods.

In addition, you can also look at the following blogs, which were used, besides other sources, to write this section:

Appendix E: Varnish Three Letter Acronyms

VAV
Varnish Arg Vector -- Argv parsing.
VBE
Varnish Back End -- Code for contacting backends (bin/varnishd/cache_backend.c)
VBP
Varnish Backend Polling -- Health checks of backends (bin/varnishd/cache_backend_poll.c)
VCA
Varnish Connection Acceptor -- The code that receives/accepts the TCP connections (bin/varnishd/cache_acceptor.c)
VCC
VCL to C Compiler -- The code that compiles VCL to C code. (lib/libvcl)
VCL
Varnish Configuration Language -- The domain-specific programming language used for configuring a varnishd.
VCT
Varnish CType(3) -- Character classification for RFC2616 and XML parsing.
VDD
Varnish (Core) Developer Day -- Quarterly invite-only meeting strictly for Varnish core (C) developers, packagers and VMOD hackers.
VEV
Varnish EVent -- library functions to implement a simple event-dispatcher.
VGB
Varnish Governing Board -- May or may not exist. If you need to ask, you are not on it.
VGC
Varnish Generated Code -- Code generated by VCC from VCL.
VIN
Varnish Instance Naming -- Resolution of -n arguments.
VLU
Varnish Line Up -- library functions to collect stream of bytes into lines for processing. (lib/libvarnish/vlu.c)
VRE
Varnish Regular-Expression -- library functions for regular-expression based matching and substring replacement. (lib/libvarnish/vre.c)
VRT
Varnish Run Time -- functions called from compiled code. (bin/varnishd/cache_vrt.c)
VRY
VaRY -- Related to processing of Vary: HTTP headers. (bin/varnishd/cache_vary.c)
VSL
Varnish Shared memory Log -- The log written into the shared memory segment for varnish{log,ncsa,top,hist} to see.
VSB
Varnish string Buffer -- a copy of the FreeBSD "sbuf" library, for safe string handling.
VSC
Varnish Statistics Counter -- counters for various stats, exposed via varnishapi.
VSS
Varnish Session Stuff -- library functions to wrap DNS/TCP. (lib/libvarnish/vss.c)
VTC
Varnish Test Code -- a test-specification for the varnishtest program.
VTLA
Varnish Three Letter Acronym -- No rule without an exception.
VUG
Varnish User Group meeting -- Half-yearly event where the users and developers of Varnish Cache gather to share experiences and plan future development.
VWx
Varnish Waiter 'x' -- A code module to monitor idle sessions.
VWE
Varnish Waiter Epoll -- epoll(2) (linux) based waiter module.
VWK
Varnish Waiter Kqueue -- kqueue(2) (freebsd) based waiter module.
VWP
Varnish Waiter Poll -- poll(2) based waiter module.
VWS
Varnish Waiter Solaris -- Solaris ports(2) based waiter module.

Appendix F: Apache as Backend

  • Install Apache. We will use it as backend.

To install Apache in Ubuntu, type the command: apt-get install apache2. Install the HTTPie utility with the command: apt-get install httpie. HTTPie allows you to issue arbitrary HTTP requests in the terminal. Next:

  1. Verify that Apache works by typing http -h localhost. You should see a 200 OK response from Apache.
  2. Change Apache's port from 80 to 8080 in /etc/apache2/ports.conf and /etc/apache2/sites-enabled/000-default.conf.
  3. Restart Apache: service apache2 restart.
  4. Verify that Apache still works by typing http -h localhost:8080.