-
Story
-
Resolution: Unresolved
-
Critical
-
None
-
None
-
None
-
False
-
None
-
False
-
SDN Sprint 225
This is hitting quite a few periodics. Non-AMD is getting hit particularly hard, arm, s390x, ppc64le, also hypershift, okd.
Pressing "See Job Runs" will give you prow links to see the specific failures.
How do we want to handle?
- links to
[TRT-487] Multiple jobs lacking network plugin in name or mismatching with name
there are a few of these left to clean up. This is a good link to use now. We are only concerned with jobs
that have this test case " prow job name should match network type" as a flake. If it's a failure, which
there are, it's because the test case really went for a toss.
I'll get back to fixing some of these from time to time and keep checking back in sippy until we don't
have any more flakes. I'll do it in the back ground. I just checked and the first group (just making
this a note for myself) that I came across are these jobs:
- periodic-ci-shiftstack-shiftstack-ci-main-periodic-4.15-e2e-openstack-dualstack-v6primary
- periodic-ci-openshift-hypershift-release-4.15-periodics-e2e-mce-agent-connected-ipv4-metal3-compact-conformance
- periodic-ci-openshift-release-master-nightly-4.15-e2e-vsphere-upi-zones
- periodic-ci-openshift-hypershift-release-4.15-periodics-e2e-kubevirt-metal-mce
- periodic-ci-openshift-hypershift-release-4.15-periodics-e2e-mce-agent-connected-ipv4-metal3-compact-conformance
Devan Goodwin
added a comment - Yeah we could special case okd, but your plan does sound better so lets go for that.
Devan Goodwin
added a comment - Yeah we could special case okd, but your plan does sound better so lets go for that. Thanks Jamo, we can special case okd if you want in the test, just say the word if you'd rather that approach and I'll do it.
what do you mean? we can let OKD alone and add some logic in origin to not flake those?
either way, I don't mind getting the okd jobs renamed as needed. This is going to be a longer activity I think. deads@redhat.com mentioned that
as long as we can get this done before 4.12 is done then we'd be good. I'm planning to spend 1-2 days each week on this.
Devan Goodwin
added a comment - Thanks Jamo, we can special case okd if you want in the test, just say the word if you'd rather that approach and I'll do it.
Devan Goodwin
added a comment - Thanks Jamo, we can special case okd if you want in the test, just say the word if you'd rather that approach and I'll do it. Sure, I did fixed up everything from Multiarch perspective (s390x/ppc64le/arm64/heterogeneous). I'll wait for a day and see if I covered everything or something missed out and confirm with you.
I can own this ticket. rh-ee-dmistry, please CC me on any PRs you have. I have this one from last week
which fixes some jobs.
My end goal is that this flake about job name not matching the underlying CNI will not appear
in any jobs.
s390x, ppc64le are easier to make the change as all of them only uses SDN.
Arm64 is little bit tricky as it has mix of sdn, ovn. I'll have a PR up in few minutes to fix all the multiarch jobs. I did fix few of the arm jobs yesterday
https://github.com/openshift/release/pull/52251
https://github.com/openshift/release/pull/52252
https://github.com/openshift/release/pull/52253
https://github.com/openshift/release/pull/52254
https://github.com/openshift/release/pull/52255
when these are merged, I'll check back in the sippy search and file more PRs