Mandate e2e-metal-ipi on cluster-network-operator and ovn-kubernetes #12374
Conversation
Multiple times now, pull requests have merged with e2e-metal-ipi failing with explicitly clear messages in the CI summary that the failure was CNO and/or OVN related. We are working on renaming these jobs starting with 4.7 to indicate they are ovn-ipv6, but it's been proven now that optional jobs are meaningless. This sets e2e-metal-ipi to be required to pass on both CNO and OVN master. These jobs are now as stable as any other platform, so there's no reason to leave them as optional. It's also the only IPv6 coverage that exists in OpenShift today.
|
Note the e2e-metal-ipi needs openshift/cluster-network-operator#817 and openshift/ironic-image#107 to merge before it passes |
|
/retest |
|
/test pj-rehearse |
trozet
left a comment
trozet
left a comment
There was a problem hiding this comment.
I wouldn't use build02 for your jobs, we found that a bunch of flakes in CI are due to using that build server:
#12386
Also don't put this in a commit message:
"but it's been proven now that
optional jobs are meaningless."
Additionally, I would make sure your job shows must gather logs extracted in the web browser, rather than making the user download the entire must gather. That would be a good way to enable people to easily check your logs. Here's an example:
|
Also, before we require this. In case we switch back to ipv4 again, please change: to be ovn-kubernetes as well. |
We haven't seen that.
Well, it's true.
You can file an issue in our jira project to take care of it, but this was not the reason no one looked. The multiple breakages including the first breakage at the end of 4.6 FF showed the failure reason clearly to be OVN highlighted in yellow in spyglass without having to dig into a must-gather. |
I already proposed that in openshift-metal3/dev-scripts#1108 and based on discussion there it probably makes sense to set NETWORK_TYPE explicitly for each CI job vs relying on any dev-scripts defaults, that way we can be sure that the configuration used matches the revised naming. |
There are three parts to this - the error pulling the must-gather which you noticed in the failures is a dev-scripts but, that should be fixed via openshift-metal3/dev-scripts#1125 Then there is the problem (as mentioned in that PR) that when the serviceNetwork doesn't come up, must-gather doesn't actually work (arguably a design fault with must-gather, but we could potentially work around it by using scp instead?) Finally there is the issue that we don't unpack the tarball, which would be a nice improvement, I'll add it to our Jira backlog. |
|
/lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: danwinship, stbenjam The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
openshift-ci-robot
added
the
approved
|
@stbenjam: Updated the following 2 configmaps:
DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
6 participants
Multiple times now, pull requests have merged with e2e-metal-ipi failing
with explicitly clear messages in the CI summary that the failure was
CNO and/or OVN related. We are working on renaming these jobs starting
with 4.7 to indicate they are ovn-ipv6, but it's been proven now that
optional jobs are meaningless.
This sets e2e-metal-ipi to be required to pass on both CNO and OVN
master. These jobs are now as stable as any other platform, so there's
no reason to leave them as optional. It's also the only IPv6 coverage
that exists in OpenShift today.