[🚀 Feature]: how to config use Tracing with OpenTelemetry and Jaeger in Selenium Grid 4 with docker?

[zengfr]

Feature and motivation

how to config use Tracing with OpenTelemetry and Jaeger in Selenium Grid 4 with docker?

Usage example

how to config use Tracing with OpenTelemetry and Jaeger in Selenium Grid 4 with docker?

[jaycapps]

ditto.. I can't get this working either. any clue?

[ceagan]

I did some research into this and got it working in my setup. There may be a better way to do this more generically, so I'm offering my findings for consideration rather than as a PR.

I ended up changing the start-selenium-grid-hub.sh and similar scripts (e.g., start-selenium-node.sh) in other Docker images to add support for the --ext parameter to be set before the command. In the following snippet, the ${EXT_OPT} variable has been added before the hub command.

java ${JAVA_OPTS:-$SE_JAVA_OPTS} -jar /opt/selenium/selenium-server.jar \
  ${EXT_OPT} \
  hub \
  --session-request-timeout ${SE_SESSION_REQUEST_TIMEOUT} \
  --session-retry-interval ${SE_SESSION_RETRY_INTERVAL} \
  --relax-checks ${SE_RELAX_CHECKS} \
  --bind-host ${SE_BIND_HOST} \
  ${HOST_CONFIG} \
  ${PORT_CONFIG} \
  ${SE_OPTS}

In my local version of the images, I also added this code above the call to java. I found it more convenient to generate the classpath data to a file in the Dockerfile, then use the content of that file here instead of trying to set the environment variable in the Dockerfile. There might be a cleaner way to do this part.

if [ ! -z "$SE_EXT_OPT" ]; then
  EXT_OPT="--ext $(cat ${SE_EXT_OPT})"
  echo "Using Selenium EXT option: ${EXT_OPT})"
fi

Once the new --ext option is available in the script, the jar files are needed in the image to support Jaeger and OpenTelemetry. I extended the existing Dockerfiles using the instructions from running java -jar selenium-server.jar info tracing. Interestingly, I found that I had to use newer versions of the dependencies in order for it to work than the ones listed in the info output.

FROM selenium/hub:4.4.0-20220831

RUN sudo curl --fail --location -o /usr/local/bin/coursier https://github.com/coursier/launchers/raw/master/coursier \
    && sudo chmod +x /usr/local/bin/coursier \
    && coursier fetch -p \
        io.opentelemetry:opentelemetry-exporter-jaeger:1.18.0 \
        io.grpc:grpc-netty:1.49.0 >/opt/selenium/jaeger.classpath

COPY start-selenium-grid-hub.sh /opt/bin/

Finally, when deploying using the helm charts, I overrode the image and set the extra environment variables like so.

---
hub:
  imageName: custom-selenium/hub-with-jaeger
  extraEnvironmentVariables:
    - name: SE_JAVA_OPTS
      value: "-Dotel.traces.exporter=jaeger -Dotel.exporter.jaeger.endpoint=http://jaeger-collector:14250 -Dotel.resource.attributes=service.name=selenium-hub"
    - name: SE_EXT_OPT
      value: /opt/selenium/jaeger.classpath
chromeNode:
  imageName: custom-selenium/node-chrome-with-jaeger
  extraEnvironmentVariables:
    - name: SE_JAVA_OPTS
      value: "-Dotel.traces.exporter=jaeger -Dotel.exporter.jaeger.endpoint=http://jaeger-collector:14250 -Dotel.resource.attributes=service.name=selenium-node-chrome"
    - name: SE_EXT_OPT
      value: /opt/selenium/jaeger.classpath

The above configurations work when Jaeger is deployed to the same namespace. Change the configs if you do something different there.