Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
13 contributors

Users who have contributed to this file

@meltsufin @gregw @ludoch @aslo @aozarov @dlorenc @ChengyuanZhao @cassand @yubin154 @nkubala @hegemonic @janbartel @ahmetb
340 lines (267 sloc) 17.9 KB

Build Status

Google Cloud Platform Jetty Docker Image

This repository contains the source for the Google-maintained Jetty docker image. This image can be used as the base image for running Java web applications on Google App Engine Flexible Environment and Google Container Engine. It provides the Jetty Servlet container on top of the OpenJDK image.

This image is mirrored at both launcher.gcr.io/google/jetty and gcr.io/google-appengine/jetty.

The layout of this image is intended to mostly mimic the official docker-jetty image and unless otherwise noted, the official docker-jetty documentation should apply.

Configuring the Jetty image

Arguments passed to the docker run command are passed to Jetty, so the configuration of the jetty server can be seen with a command like:

docker run gcr.io/google-appengine/jetty --list-config

Alternate commands can also be passed to the docker run command, so the image can be explored with

docker run -it --rm launcher.gcr.io/google/jetty bash

Various environment variables (see below) can also be used to set jetty properties, enable modules and disable modules. These variables may be set either in an app.yaml or passed in to a docker run command eg.

docker run -it --rm -e JETTY_PROPERTIES=jetty.http.idleTimeout=10000 launcher.gcr.io/google/jetty 

To update the server configuration in a derived Docker image, the Dockerfile may enable additional modules with RUN commands like:

WORKDIR $JETTY_BASE
RUN java -jar "$JETTY_HOME/start.jar" --add-to-startd=jmx,stats

Modules may be configured in a Dockerfile by editing the properties in the corresponding mod files in /var/lib/jetty/start.d/ or the module can be deactivated by removing that file.

Enabling gzip compression

The gzip handler is bundled with Jetty but not activated by default. To activate this module you have to set the environment variable JETTY_MODULES_ENABLE=gzip

For example with docker:

docker run -p 8080 -e JETTY_MODULES_ENABLE=gzip gcr.io/yourproject/yourimage

Or with GAE (app.yaml):

env_variables:
  JETTY_MODULES_ENABLE: gzip

Using Quickstart

Jetty provides mechanisms to speed up the start time of your application by pre-scanning its content and generating configuration files. If you are using an extended image you can active quickstart by executing /scripts/jetty/quickstart.sh in your Dockerfile, after the application WAR is added.

FROM launcher.gcr.io/google/jetty
ADD your-application.war $JETTY_BASE/webapps/root.war

# generate quickstart-web.xml
RUN /scripts/jetty/quickstart.sh

App Engine Flexible Environment

When using App Engine Flexible, you can use the runtime without worrying about Docker by specifying runtime: java in your app.yaml:

runtime: java
env: flex

The runtime image launcher.gcr.io/google/jetty will be automatically selected if you are attempting to deploy a WAR (*.war file).

If you want to use the image as a base for a custom runtime, you can specify runtime: custom in your app.yaml and then write the Dockerfile like this:

FROM launcher.gcr.io/google/jetty
ADD your-application.war $APP_DESTINATION_WAR

That will add the WAR in the correct location for the Docker container.

You can also use exploded-war artifacts:

ADD your-application $APP_DESTINATION_EXPLODED_WAR

Once you have this configuration, you can use the Google Cloud SDK to deploy this directory containing the 2 configuration files and the WAR using:

gcloud app deploy app.yaml

Container Engine & other Docker hosts

For other Docker hosts, you'll need to create a Dockerfile based on this image that copies your application code and installs dependencies. For example:

FROM launcher.gcr.io/google/jetty
COPY your-application.war $APP_DESTINATION_WAR

If your artifact is an exploded-war, then use the APP_DESTINATION_EXPLODED_WAR environment variable instead. You can then build the docker container using docker build or Google Cloud Container Builder. By default, the CMD is set to start the Jetty server. You can change this by specifying your own CMD or ENTRYPOINT.

Entry Point Features

The /docker-entrypoint.bash for the image is inherited from the openjdk-runtime and its capabilities are described in the associated README

This image updates the docker CMD and adds the /setup-env.d/50-jetty.bash script to include options and arguments to run the Jetty container, unless an executable argument is passed to the docker image. Additional environment variables are used/set including:

Env Var Maven Prop Value/Comment
JETTY_VERSION jetty9.version
GAE_IMAGE_NAME jetty
GAE_IMAGE_LABEL docker.tag.long
JETTY_HOME jetty.home /opt/jetty-home
JETTY_BASE jetty.base /var/lib/jetty
TMPDIR /tmp/jetty
JETTY_PROPERTIES Comma separated list of name=value pairs appended to $JETTY_ARGS
JETTY_MODULES_ENABLE Comma separated list of modules to enable by appending to $JETTY_ARGS
JETTY_MODULES_DISABLE Comma separated list of modules to disable by removing from $JETTY_BASE/start.d
JETTY_ARGS Arguments passed to jetty's start.jar. Any arguments used for custom jetty configuration should be passed here.
ROOT_WAR $JETTY_BASE/webapps/root.war
ROOT_DIR $JETTY_BASE/webapps/root
JAVA_OPTS JVM runtime arguments

If a WAR file is found at $ROOT_WAR, it is unpacked to $ROOT_DIR if it is newer than the directory or the directory does not exist. If there is no $ROOT_WAR or $ROOT_DIR, then /app is symbolic linked to $ROOT_DIR. If a $ROOT_DIR is discovered or made by this script, then it is set as the working directory. See Extending the image below for some examples of adding an application as a WAR file or directory.

The command line executed is effectively (where $@ are the args passed into the docker entry point):

java $JAVA_OPTS \
     -Djetty.base=$JETTY_BASE \
     -jar $JETTY_HOME/start.jar \
     "$@"

Logging

This image is configured to use Java Util Logging(JUL) to capture all logging from the container and its dependencies. Applications that also use the JUL API will inherit the same logging configuration.

By default JUL is configured to use a ConsoleHandler to send logs to the stderr of the container process. When run on as a GCP deployment, all output to stderr is captured and is available via the Stackdriver logging console, however more detailed and integrated logs are available if the Stackdriver logging mechanism is used directly (see below).

To alter logging configuration a new logging.properties file must be provided to the image that among other things can: alter log levels generated by Loggers; alter log levels accepted by handlers; add/remove/configure log handlers.

Providing logging.properties via the web application

A new logging configuration file can be provided as part of the application (typically at WEB-INF/logging.properties) and the Java System Property java.util.logging.config.file updated to reference it.

When running in a GCP environment, the system property can be set in app.yaml:

env_variables:
  JETTY_ARGS: -Djava.util.logging.config.file=WEB-INF/logging.properties

If the image is run directly, then a -e argument to the docker run command can be used to set the system property:

docker run \
  -e JETTY_ARGS=-Djava.util.logging.config.file=WEB-INF/logging.properties \
  ...

Providing logging.properties via a custom image

If this image is being used as the base of a custom image, then the following Dockerfile commands can be used to add either replace the existing logging configuration file or to add a new logging.properties file.

The default logging configuration file is located at /var/lib/jetty/etc/java-util-logging.properties, which can be replaced in a custom image is built. The default configuration can be replaced with a Dockerfile like:

FROM gcr.io/google-appengine/jetty
ADD logging.properties /var/lib/jetty/etc/java-util-logging.properties
...

Alternately an entirely new location for the file can be provided and the environment amended in a Dockerfile like:

FROM gcr.io/google-appengine/jetty
ADD logging.properties /etc/logging.properties
ENV JETTY_ARGS -Djava.util.logging.config.file=/etc/logging.properties
...

Providing logging.properties via docker run

A logging.properties file may be added to an existing images using the docker run command if the deployment environment allows for the run arguments to be modified. The -v option can be used to bind a new logging.properties file to the running instance and the -e option can be used to set the system property to point to it:

docker run -it --rm \
-v /mylocaldir/logging.properties:/etc/logging.properties \
-e JETTY_ARGS="-Djava.util.logging.config.file=/etc/logging.properties" \
...

Enhanced Stackdriver Logging (BETA)

When running on the Google Cloud Platform Flex environment, the Java Util Logging can be configured to send logs to Google Stackdriver Logging by providing a logging.properties file that configures a LoggingHandler as follows:

handlers=com.google.cloud.logging.LoggingHandler

# Optional configuration
.level=INFO
com.google.cloud.logging.LoggingHandler.level=FINE
com.google.cloud.logging.LoggingHandler.log=gae_app.log
com.google.cloud.logging.LoggingHandler.formatter=java.util.logging.SimpleFormatter
java.util.logging.SimpleFormatter.format=%3$s: %5$s%6$s

When deployed on the GCP Flex environment, an image so configured will automatically be configured with:

  • a LabelLoggingEnhancer instance, that will add labels from the monitored resource to each log entry.
  • a TraceLoggingEnhancer instance that will add any trace-id set to each log entry.
  • the gcp module will be enabled that configures jetty so that the setCurrentTraceId method is called for any thread handling a request.

When deployed in other environments, logging enhancers can be manually configured by setting a comma separated list of class names as the com.google.cloud.logging.LoggingHandler.enhancers property.

When using Stackdriver logging, it is recommended that io.grpc and sun.net logging level is kept at INFO level, as both these packages are used by Stackdriver internals and can result in verbose and/or initialisation problems.

Distributed Session Storage

The Jetty session mechanism is highly customizable and the options presented below are only a subset of meaningful configurations. Consult the Jetty Sessions documentation for more details.

Google Cloud Session Store

This image can be configured to use Google Cloud Datastore for clustered session storage by enabling the gcp-datastore-sessions jetty module. You can do this in your app.yaml:

env_variables:
  JETTY_MODULES_ENABLE: gcp-datastore-sessions

Jetty will use the default namespace in Datastore as the store for all session data, or jetty.session.gcloud.namespace property can be used to set an alternative namespace. By default gcloud has no request affinity, so all session data will be retrieved and stored from the datastore on every request and no session data will be shared in memory.

Note that the gcp-datastore-sessions module is an aggregate module and the same configuration can be achieved by activating it's dependent modules individually:

env_variables:
  JETTY_MODULES_ENABLE: session-cache-null,gcp-datastore,session-store-gcloud

Cached Google Cloud Session Store

The Google Load Balancer can support instance affinity for more efficient session usage. This can be configured in app.yaml with:

network:
  session_affinity: true

env_variables:
  JETTY_MODULES_ENABLE: session-cache-hash,gcp-datastore,session-store-gcloud

Sessions will be retrieved from the in memory session cache and multiple requests can share a session instance. The Google Data Cloud is only accessed for unknown sessions (if affinity changes) or if a session is modified. Session cache behaviour can be further configured by following the Jetty Session Cache documentation. Note that affinity is achieved by the Google Load Balancer setting a GCLB cookie rather than tracking the JSESSIONID cookie.

Memcached Google Cloud Session Store (Alpha)

Sessions can be cached in memcache (without need for affinity) and backed by Google Cloud Datastore. This can be configured in app.yaml with:

env_variables:
  JETTY_MODULES_ENABLE: gcp-memcache-datastore-sessions

Note that the gcp-memcache-datastore-sessions module is an aggregate module and the same configuration can be achieved by activating it's dependent modules individually:

env_variables:
  JETTY_MODULES_ENABLE: session-cache-null,gcp-datastore,session-store-gcloud,gcp-xmemcached,session-store-cache

The session-cache-null module may be replaced with the session-cache-hash module to achieve 2 levels of caching (in memory and memcache) prior to accessing the Google Cloud Datastore, and network affinity may also be activated as above.

Extending the image

The image produced by this project may be automatically used/extended by the Cloud SDK and/or App Engine maven plugin. Alternately it may be explicitly extended with a custom Dockerfile.

The latest released version of this image is available at launcher.gcr.io/google/jetty, alternately you may build and push your own version with the shell commands:

mvn clean install
docker tag jetty:latest gcr.io/your-project-name/jetty:your-label
gcloud docker -- push gcr.io/your-project-name/jetty:your-label

Adding the root WAR application to an image

A standard war file may be deployed as the root context in an extended image by placing the war file in the docker build directory and using a Dockerfile like:

FROM launcher.gcr.io/google/jetty
COPY your-application.war $APP_DESTINATION_WAR

An exploded-war can also be used:

COPY your-application $APP_DESTINATION_EXPLODED_WAR

Adding the root application to an image

If the application exists as directory (i.e. an expanded war file), then directory must be placed in the docker build directory and using a Dockerfile like:

FROM launcher.gcr.io/google/jetty
COPY your-application-dir $JETTY_BASE/webapps/root

Mounting the root application at local runtime

If no root WAR or root directory is found, the docker-entrypoint.bash script will link the /app directory as the root application. Thus the root application can be added to the image via a runtime mount:

docker run -v /some-path/your-application:/app launcher.gcr.io/google/jetty  

Enabling dry-run

The image's default start command will first run the jetty start.jar as a --dry-run to generate the JVM start command before starting the jetty web server. If you wish to generate the start command in your Dockerfile rather than at container start-time, you can run the /scripts/jetty/generate-jetty-start.sh script to generate it for you, i.e.

RUN /scripts/jetty/generate-jetty-start.sh

NOTE: Make sure that the web application and any additional custom jetty modules have been added to the container BEFORE running this script.

Development Guide

Contributing changes

Licensing

You can’t perform that action at this time.