e2e-runner

Docker image for running workflow-e2e

Overview of use

E2e-runner coordinates the entire e2e run against a Workflow chart, including:

1. Coordinating the leasing of a GKE k8s cluster via k8s-claimer,
2. Cleaning up the leased cluster if need be (primarily deleting the deis namespace if still exists),
3. Setting up the local helm install on the leased cluster (deleting the tiller deployment if necessary),
4. Installing the Workflow and Workflow-e2e charts (setting particular values if necessary),
5. Monitoring to see when these charts are up and running,
6. Following and capturing chart logs and placing them where Jenkins/others can find them before deleting the cluster lease and exiting.

See the main run script for the basic outline of actions presented above. It is a good entry point into the finer details of e2e-runner functionality.

Running the tests on CI

To run the tests for a jenkins job you should have a docker run command that looks like the following:

env > /home/jenkins/workspace/${JOB_NAME}/${BUILD_NUMBER}/env.file
docker run \
  --env-file=/home/jenkins/workspace/${JOB_NAME}/${BUILD_NUMBER}/env.file \
  -u jenkins:jenkins \
  -v /home/jenkins/workspace/${JOB_NAME}/${BUILD_NUMBER}:/home/jenkins/logs:rw \
  quay.io/deisci/e2e-runner

A few different things are happening here. First we export all the local environment variables to a file so we can pass those into the container at runtime. This allows us to use secret text values for things like $AUTH_TOKEN. When we call docker run we make sure we specify the user jenkins:jenkins. This way the container has write permissions into the volume mount.

Running the tests locally

$ docker run -e AUTH_TOKEN=$AUTH_TOKEN quay.io/deisci/e2e-runner

Environment Variables

• AUTH_TOKEN - Token needed to talk to k8s claimer
• CLUSTER_DURATION - How long to lease the k8s cluster (default: 800 seconds)
• GINKGO_NODES - How many nodes to use when running e2e tests in parallel (default: 30)