CBI/Jenkins Migration FAQ
- 1 Jenkins Migration FAQ
- 1.1 What's happening?
- 1.2 Why are you doing this?
- 1.3 What does that mean for Eclipse projects using the CI infrastructure?
- 1.4 What does [OI2JIRO] means in bugs like 547135?
- 1.5 What does JIRO means?
- 1.6 What needs to be changed after the migration?
- 1.7 How do I run UI-tests on the new infra?
- 1.8 What’s the plan?
- 1.9 What is happening to instances already hosted on CloudBees Core / Jenkins Enterprise?
- 1.10 Will the new CI infrastructure also support different platforms (e.g. Windows or Mac)?
- 1.11 Will external/remote slaves still be supported?
- 1.12 Will it be possible to trigger jobs on events (like GitHub pull requests)?
- 1.13 Will builds be faster?
- 1.14 Will long-running UI tests be faster?
Jenkins Migration FAQ
Projects hosted at the Eclipse Foundation will soon benefit from a brand new continuous integration (CI) infrastructure. Expected improvements are: resiliency, scalability and nimbleness. We are doing this move with tremendous support from our friends at RedHat with their product OpenShift Container Platform.
Why are you doing this?
We need to
- scale up our current JIPP infrastructure, and make it more efficient
- utilize our hardware better and be able to add interim cloud resources when needed
- be able to limit a project's resource consumption and isolate it from other projects
- be able to update Jenkins masters and to install/update Jenkins plugins in batch
- provide more flexibility to projects to let them build their code in containers so that they control the build environment
- have built-in resilience. Jenkins on OpenShift provide this, by leveraging technologies such as Docker and Kubernetes.
What does that mean for Eclipse projects using the CI infrastructure?
We don’t expect much disruption, and most projects will only need to change minor things in their build settings.
What does [OI2JIRO] means in bugs like 547135?
It means Old Infrastructure 2 JIRO.
What does JIRO means?
See https://github.com/eclipse-cbi/jiro#why-is-it-called-jiro. TL;DR Jenkins Instances Running [on] OpenShift.
What needs to be changed after the migration?
- For build jobs without specific dependencies and UI tests (e.g. a standard Maven/Tycho build) the default jnlp-agent image can and should be used with both freestyle jobs (no special configuration required) or Jenkins pipelines. Basic tools like Ant, Maven, JDK are provided (see https://wiki.eclipse.org/Jenkins#Tools_.28and_locations_on_the_default_JNLP_agent_container.29).
- Build jobs with specific (tool) dependencies (e.g. Node.js, CMake, etc) should try to use corresponding Jenkins plugins first. Please open a Bugzilla issue to request a plugin installation.
- If a custom tailored build environment is required, you can use a custom docker image and Jenkins pipelines, see https://wiki.eclipse.org/Jenkins#How_do_I_run_my_build_in_a_custom_container.3F for more info. This gives you the most flexible environment, but it requires some knowledge about docker images and Jenkins pipelines.
- For UI tests, see below.
- Publishing to download.eclipse.org requires access via SSH/SCP. We will add the necessary credentials to your JIPP. Please see https://wiki.eclipse.org/Jenkins#How_do_I_deploy_artifacts_to_download.eclipse.org.3F for more info.
- If your project used a RAM disk on the old infra (when in doubt, you probably did not), you will need to adjust your build config accordingly. RAM disks might be added again later on, but this has a very low priority.
How do I run UI-tests on the new infra?
For UI tests, there are two scenarios:
- In general, you can use a custom docker image and Jenkins pipelines, see https://wiki.eclipse.org/Jenkins#How_do_I_run_my_build_in_a_custom_container.3F.
- To ease the migration/transition, we provide a basic UI-test pod template that can be used with freestyle jobs. This image only provides a minimal UI test environment. If your project requires specific dependencies, in most cases, you will need to roll your own.
The basic UI-test pod template has the label "ui-test" that can be specified in the job configuration under "Restrict where this project can be run".
What’s the plan?
Starting in February 2019, all new projects will get an instance on the new infrastructure instead of the old infrastructure. In the meantime, we will start migrating existing JIPPs by reaching out to projects to tell them when the migration will happen and see if it fits with their release agenda. There is no set timeline, but we aim to ramp up pretty quickly and do the migration as fast as we can.
The migration itself will happen in two steps. First, we will create an instance on the new infrastructure with the URL https://ci-staging.eclipse.org/<project_name> and migrate existing build jobs to it. The current instance on the old infra will stay as is at the URL https://ci.eclipse.org/<project_name>. Once the project will have confirmed that the new instance behavior is correct, we will remove the old instance and will make the new instance available at the original URL https://ci.eclipse.org/<project_name>. This process should cause only slight disruptions for contributors and projects users.
What is happening to instances already hosted on CloudBees Core / Jenkins Enterprise?
Those instances (especially EE4J instances) already run inside the cluster and will stay as is at least until we finish the migration of Jenkins instances from the current infrastructure to the new one.
Will the new CI infrastructure also support different platforms (e.g. Windows or Mac)?
Yes, we will try to integrate existing and new (cloud hosted) Windows and Mac machines.
Will external/remote slaves still be supported?
Yes, they will be supported.
Will it be possible to trigger jobs on events (like GitHub pull requests)?
This is already working in the current environment with the help of the GitHub pull request plugin and will continue to be working on the new infrastructure.
Will builds be faster?
As mentioned above, we are aiming to better utilize our existing hardware, add additional (cloud based) resources and isolate resource consumption. This should improve the build times in general.
Will long-running UI tests be faster?
More flexible build agents should speed things up. We will also have better monitoring to identify potential bottle-necks. Jenkins Pipelines also make it easier to set up tests in parallel.