camel-k
camel-k copied to clipboard
apache
Drop support for Openshift 3 and related E2E test
I think we should seriously drop the support for OC3. We're going to have more and more situations like this in the future, limiting our capacity to innovate because we need to find a workaround to run on OC3.
Originally posted by @squakez in https://github.com/apache/camel-k/pull/3419#pullrequestreview-1031127840
We're recently having discussions on the opportunity to drop the flaky test we have for OC3. The reason why we're still having it is because it is easy enough to have a self contained Openshift cluster within Github Actions. However, the differences between the latest version of Openshift 4 (and probably more in general from Kubernetes model) is getting bigger and bigger with the time. The maintenance burden and the flakiness we are experiencing is becoming unsustainable IMO.
@tadayosi @oscerd @astefanutti @lburgazzoli @phantomjinx @christophd WDYT?
+1. For reference, duplicates #1550. What we would lose is immediate feedback for some RBAC and s2i coverage mainly. The pros would be obvious (more stable, faster, cheaper, e2e test runs) and probably now overweight the cons.
+1 too. Let's close the old #1550 and use this to track the issue. We'd lose the coverages Antonin listed + route test, but so long as we have some feedback channel for OCP4 testing we could catch up any regressions on those OCP features.
@tadayosi following that issue, I've seen you made experiments with Microshift. Do you think it is a worth continuing them? or we should just abandon that path?
@squakez Yes, I'll revisit the MicroShift workflow later. Right now, however, it should be able to cover only route tests and not for S2I and RBAC tests.