Skip to content
This repository

Symfony2 Bundle that provides a way to set path based cache expiration headers via the app configuration

branch: master
README.md

CacheControlBundle

This Bundle provides a way to set path based cache expiration headers via the app configuration and provides a helper to control the reverse proxy varnish.

Build Status

Installation with composer

Just add the following line to your projects composer.json require section:

"liip/cache-control-bundle": "~1.0"

Enable the module

Add this bundle to your application's kernel:

// application/ApplicationKernel.php
public function registerBundles()
{
  return array(
      // ...
      new Liip\CacheControlBundle\LiipCacheControlBundle(),
      // ...
  );
}

Cache control

Simply configure as many paths as needed with the given cache control rules:

# app/config.yml
liip_cache_control:
    rules:
        # the controls section values are used in a call to Response::setCache();
        - { path: ^/, controls: { public: true, max_age: 15, s_maxage: 30, last_modified: "-1 hour" }, vary: [Accept-Encoding, Accept-Language] }

        # only match login.example.com
        - { host: ^login.example.com$, controls: { public: false, max_age: 0, s_maxage: 0, last_modified: "-1 hour" }, vary: [Accept-Encoding, Accept-Language] }

        # match a specific controller action
        - { controller: ^AcmeBundle:Default:index$, controls: { public: true, max_age: 15, s_maxage: 30, last_modified: "-1 hour" }, vary: [Accept-Encoding, Accept-Language] }

The matches are tried from top to bottom, the first match is taken and applied.

Run app/console config:dump-reference liip_cache_control to get the full list of configuration options.

About the path parameter

The path, host and controller parameter of the rules represent a regular expression that a page must match to use the rule.

For this reason, and it's probably not the behaviour you'd have expected, the path ^/ will match any page.

If you just want to match the homepage you need to use the path ^/$.

To match pages URLs with caching rules, this bundle uses the class Symfony\Component\HttpFoundation\RequestMatcher.

The unless_role makes it possible to skip rules based on if the current authenticated user has been granted the provided role.

Debug information

The debug parameter adds a X-Cache-Debug header to each response that you can use in your Varnish configuration.

# app/config.yml
liip_cache_control:
    debug: true

Add the following code to your Varnish configuration to have debug headers added to the response if it is enabled:

#in sub vcl_deliver
# debug info
# https://www.varnish-cache.org/trac/wiki/VCLExampleHitMissHeader
if (resp.http.X-Cache-Debug) {
    if (obj.hits > 0) {
        set resp.http.X-Cache = "HIT";
        set resp.http.X-Cache-Hits = obj.hits;
    } else {
       set resp.http.X-Cache = "MISS";
    }
    set resp.http.X-Cache-Expires = resp.http.Expires;
} else {
    # remove Varnish/proxy header
    remove resp.http.X-Varnish;
    remove resp.http.Via;
    remove resp.http.X-Purge-URL;
    remove resp.http.X-Purge-Host;
}

Custom Varnish Parameters

Additionally to the default supported headers, you may want to set custom caching headers for varnish.

# app/config.yml
liip_cache_control:
    rules:
        # the controls section values are used in a call to Response::setCache();
        - { path: /, controls: { stale_while_revalidate=9000, stale_if_error=3000, must-revalidate=false, proxy_revalidate=true } }

Custom Varnish Time-Outs

Varnish checks the Cache-Control header of your response to set the TTL. Sometimes you may want that varnish should cache your response for a longer time than the browser. This way you can increase the performance by reducing requests to the backend.

To achieve this you can set the reverse_proxy_ttl option for your rule:

# app/config.yml
liip_cache_control:
    rules:
        # the controls section values are used in a call to Response::setCache();
        - { path: /, reverse_proxy_ttl: 300, controls: { public: true, max_age: 15, s_maxage: 30, last_modified: "-1 hour" } }

This example will add the header X-Reverse-Proxy-TTL: 300 to your response.

But by default, varnish will not know anything about it. To get it to work you have to extend your varnish vcl_fetch configuration:

sub vcl_fetch {

    /* ... */

    if (beresp.http.X-Reverse-Proxy-TTL) {
        C{
            char *ttl;
            ttl = VRT_GetHdr(sp, HDR_BERESP, "\024X-Reverse-Proxy-TTL:");
            VRT_l_beresp_ttl(sp, atoi(ttl));
        }C
        unset beresp.http.X-Reverse-Proxy-TTL;
    }

    /* ... */

}

Varnish will then look for the X-Reverse-Proxy-TTL header and if it exists, varnish will use the found value as TTL and then remove the header. There is a beresp.ttl field in VCL but unfortunately it can only be set to absolute values and not dynamically. Thus we have to use a C code fragment.

Note that if you are using this, you should have a good purging strategy.

Varnish helper

This helper can be used to talk back to varnish to invalidate cached URLs. Configure the location of the varnish reverse proxies (be sure not to forget any, as each varnish must be notified separately):

# app/config.yml
liip_cache_control:
    varnish:
        host: http://www.liip.ch
        ips: 10.0.0.10, 10.0.0.11
        port: 80
        headers: ["Authorization: Basic Zm9vOmJhcg==", "X-Another-Header: here"]
  • host: This must match the web host clients are using when connecting to varnish. You will not notice if this is mistyped, but cache invalidation will never happen. You can also add a regexp here like "." to clear all host entries. The regexp will be surrounded by "^(" and ")$" ending in "^(.)$" in this example.
  • ips: List of IP adresses of your varnish servers. Comma separated.
  • port: The port varnish is listening on for incoming web connections.
  • headers: (optional) If you want to send special headers with each request sent to varnish, you can add them here (as array)

To use the varnish cache helper you must inject the liip_cache_control.varnish service or fetch it from the service container:

// using a "manual" url
$varnish = $this->container->get('liip_cache_control.varnish');
/* $response Is an associative array with keys 'headers', 'body', 'error' and 'errorNumber' for each configured IP.
   A sample response will look like:
   array('10.0.0.10' => array('body'    => 'raw-request-body',
                              'headers' => 'raw-headers',
                              'error'   =>  'curl-error-msg',
                              'errorNumber'   =>  integer-curl-error-number),
          '10.0.0.11' => ...)
*/
$response = $varnish->invalidatePath('/some/path');

// using the router to generate the url
$router = $this->container->get('router');
$varnish = $this->container->get('liip_cache_control.varnish');
$response = $varnish->invalidatePath($router->generate('myRouteName'));

When using ESI, you will want to purge individual fragments. To generate the corresponding _internal route, inject the http_kernel into your controller and use HttpKernel::generateInternalUri with the parameters as in the twig render tag.

Purging

Add the following code to your Varnish configuration to have it handle PURGE requests (make sure to uncomment the appropiate line(s))

varnish 3.x

#top level:
# who is allowed to purge from cache
# https://www.varnish-cache.org/docs/trunk/users-guide/purging.html
acl purge {
    "127.0.0.1"; #localhost for dev purposes
    "10.0.11.0"/24; #server closed network
}

#in sub vcl_recv
# purge if client is in correct ip range
if (req.request == "PURGE") {
    if (!client.ip ~ purge) {
        error 405 "Not allowed.";
    }

    return(lookup);
}

sub vcl_hit {
  if (req.request == "PURGE") {
     purge;
     error 200 "Purged";
     return (error);
  }
}

sub vcl_miss {
   if (req.request == "PURGE") {
     purge;
     error 404 "Not in cache";
     return (error);
   }
}

In Varnish 2, the purge action is actually just marking caches as invalid. This is called ban in Varnish 3.

Varnish 2.x

#top level:
# who is allowed to purge from cache
# https://www.varnish-cache.org/docs/trunk/users-guide/purging.html
acl purge {
    "127.0.0.1"; #localhost for dev purposes
    "10.0.11.0"/24; #server closed network
}

#in sub vcl_recv
# purge if client is in correct ip range
if (req.request == "PURGE") {
    if (!client.ip ~ purge) {
        error 405 "Not allowed.";
    }

    purge("req.url ~ " req.url);
    purge("req.url ~ " req.url);
    error 200 "Success";
}

NOTE: this code invalidates the url for all domains. If your varnish serves multiple domains, you should improve this configuration.

The varnish path invalidation is about equivalent to doing this:

 netcat localhost 6081 << EOF
 PURGE /url/to/purge HTTP/1.1
 Host: webapp-host.name

 EOF

Banning

Since varnish 3 banning can be used to invalidate the cache. Banning invalidates whole section with regular expressions, so you will need to be careful to not invalidate too much.

Configure the varnish reverse proxies to use ban as purge instruction:

# app/config.yml
liip_cache_control:
    varnish:
        purge_instruction: ban

This will do a purge request and will add X-Purge headers which can be used by your Varnish configuration:

varnish 3.x

#top level:
# who is allowed to purge from cache
# https://www.varnish-cache.org/docs/trunk/users-guide/purging.html
acl purge {
    "127.0.0.1"; #localhost for dev purposes
    "10.0.11.0"/24; #server closed network
}

#in sub vcl_recv
# purge if client is in correct ip range
if (req.request == "PURGE") {
    if (!client.ip ~ purge) {
        error 405 "Not allowed.";
    }
    ban("obj.http.X-Purge-Host ~ " + req.http.X-Purge-Host + " && obj.http.X-Purge-URL ~ " + req.http.X-Purge-Regex + " && obj.http.Content-Type ~ " + req.http.X-Purge-Content-Type);
    error 200 "Purged.";
}

#in sub vcl_fetch
# add ban-lurker tags to object
set beresp.http.X-Purge-URL = req.url;
set beresp.http.X-Purge-Host = req.http.host;

Force refresh

Alternatively one can also force a refresh using the approach

#top level:
# who is allowed to purge from cache
# http://www.varnish-cache.org/trac/wiki/VCLExampleEnableForceRefresh
acl refresh {
    "127.0.0.1"; #localhost for dev purposes
    "10.0.11.0"/24; #server closed network
}

sub vcl_hit {
    if (!obj.cacheable) {
        pass;
    }

    if (req.http.Cache-Control ~ "no-cache" && client.ip ~ refresh) {
        set obj.ttl = 0s;
        return (restart);
    }
    deliver;
}

The vanish path force refresh is about equivalent to doing this:

netcat localhost 6081 << EOF
GET /url/to/refresh HTTP/1.1
Host: webapp-host.name
Cache-Control: no-cache, no-store, max-age=0, must-revalidate

EOF

To use the varnish cache helper you must inject the liip_cache_control.varnish service or fetch it from the service container:

// using a "manual" url
$varnish = $this->container->get('liip_cache_control.varnish');
$varnish->refreshPath('/some/path');

Banning from the console

You can also ban URLs from the console

app/console liip:cache-control:varnish:invalidate

will ban (invalidate) all entries in your configured varnish servers (matching varnish.host)

app/console liip:cache-control:varnish:invalidate /posts.*

will ban (invalidate) all entries in your configured varnish servers, where the URL starts with "/posts". Any regular expression understood by varnish can be used here.

It uses the Varnish Helper class, therefore if you defined more than one varnish server in the config file (in varnish.ips), the entries will be deleted in all servers.

Cache authorization listener

Enable the authorization listener:

# app/config.yml
liip_cache_control:
    authorization_listener: true

This listener makes it possible to stop a request with a 200 "OK" for HEAD requests right after the security firewall has finished. This is useful when one uses Varnish while handling content that is not available for all users.

In this scenario on a cache hit, Varnish can be configured to issue a HEAD request when this content is accessed. This way Symfony2 can be used to validate the authorization, but no work needs to be made to regenerate the content that is already in the Varnish cache.

Note this obviously means that it only works with path based Security. Any additional security implemented inside the Controller will be ignored.

Note further that a HEAD response is supposed to contain the same HTTP header meta data as the GET response to the same URL. However for the purpose of this use case we have no other choice but to assume a 200.

backend default {
    .host = “127.0.0.1″;
    .port = “81″;
}

acl purge {
    “127.0.0.1″; #localhost for dev purposes
}

sub vcl_recv {
    # pipe HEAD requests as we convert all GET requests to HEAD and back later on
    if (req.request == “HEAD”) {
        return (pipe);
    }


    if (req.request == "GET") {
        if (req.restarts == 0) {
            set req.request = "HEAD";
            return (pass);
        } else {
            set req.http.Surrogate-Capability = "abc=ESI/1.0";
            return (lookup);
        }
    }
}

sub vcl_hash {
}

sub vcl_fetch {
    if (beresp.http.Cache-Control ~ “(private|no-cache|no-store)”) {
        return (pass);
    }

    if (beresp.status >= 200 && beresp.status < 300) {
        if (req.request == "HEAD") {
            # if the BE response said OK, change the request type back to GET and restart
            set req.request = "GET";
            restart;
        }
    } else {
        # In any other case (authentication 302 most likely), just pass the response to the client
        # Don't forget to set the content-length, as the HEAD response doesn't have any (and the client will hang)
        if (req.request == "HEAD") {
            set beresp.http.content-length = "0";
        }

        return (pass);
    }

    if (beresp.http.Surrogate-Control ~ "ESI/1.0") {
        unset beresp.http.Surrogate-Control;
        // varnish < 3.0:
        esi;
        // varnish 3.0 and later:
        // set beresp.do_esi = true;
    }
}

Flash message listener

The Response flash message listener moves all flash messages currently set into a cookie. This way it becomes possible to better handle flash messages in combination with ESI. The ESI configuration will need to ignore the configured cookie. It will then be up to the client to read out the cookie, display the flash message and remove the flash message via javascript.

# app/config.yml
liip_cache_control:
    flash_message_listener:
        name: flashes
        path: /
        host: null
        secure: false
        httpOnly: true

If you do not want the flash message listener, you can disable it:

# app/config.yml
liip_cache_control:
    flash_message_listener:
        enabled: false
Something went wrong with that request. Please try again.