openshift-pipelines / plumbing Goto Github PK

Related to tektoncd/plumbing#102, we should try to define, share and
used our way of doing pipeline-as-code here.

If we go the monorepo route, it will be simpler than upstream, as we
will have everything at the same place (so no need for tektoncd/plumbing#103)

This is an issue to discuss and bootstrap the work on building
openshift-pipelines using openshift-pipelines 👼

In a gist, we want to handle everything using a permanent openshift
instance where openshift-pipelines is installed (let's say
cluster-devint
for now). This means:

• nightly build
• nightly sync with upstream master
• cron jobs to build released version (for flakyness, on different
  platform)
• one shot execution (using some chatops in github/slack/…)
• release

The component "upstream" to use (so far) are:

• pipeline
• triggers

We will probably need some code to glue things up, some of our own
services and some shared with upstream (from
plumbing).

Use flag --pull-always when building images. Otherwise a cached version of base image will be used (it could be very old and contain CVEs or other bugs).

It will slow down the builds unpredictably (when a new version of base image needs to be downloaded) but IMHO it's an acceptable trade-off.

The only place building images that I found is https://github.com/openshift-pipelines/plumbing/blob/master/CI/tasks/bootstrap/build-tektoncd-pipeline/Makefile#L73

As tetkoncd/plumbing#137, but based on openshift-install devcluster,
we should aim to have a Task (or a set of Task), to get an openshift
cluster with the specified requirements (like where it runs "gcp, aws,
…", which version, …).

This could be one-shot clusters or getting it from a pool of
clusters that would be pre-populated and managed ahead of time.

This also contains anything required for garbage collection of those
test clusters.