[Failing Job] post-kubernetes-push-image-etcd

[spiffxp]

Which jobs are failing:

• post-kubernetes-push-image-etcd

Which test(s) are failing:

n/a

Since when has it been failing:

2020-08-11

Testgrid link:

https://testgrid.k8s.io/sig-k8s-infra-gcb#post-kubernetes-push-image-etcd

Reason for failure:

docker buildx version
docker: 'buildx' is not a docker command.
See 'docker --help'

Anything else we need to know:

Started failing when #104116 merged

No new versions of etcd are getting pushed until this gets fixed, when means the following PRs haven't fully taken effect:

• Use docker buildx for etcd image #104116
• Update golang used in etcd image to 1.16.3 #104301
• golang 1.17 fails to parse IPs with leading zeros #104368
• [go1.17] Update to go1.17 #103692
• [go1.17] Update to go1.17.1 #104904

Relevant SIG

/sig api-machinery

/priority critical-urgent
/milestone v1.23
/assign @saschagrunert @jpbetz
FYI @kubernetes/release-engineering

[spiffxp]

/sig release
/area release-eng

[spiffxp]

I've opened #105285 to rollback the offending commit, would like to keep this open until we can figure out the correct fix-forward

[fedebongio]

/triage accepted

[spiffxp]

https://prow.k8s.io/view/gs/kubernetes-jenkins/logs/post-kubernetes-push-image-etcd/1442594901840629760 was triggered by the merge of #105285, which pushed gcr.io/k8s-staging-etcd/etcd:3.5.0-3

So that's now current / includes whatever PRs linked in the description may have tried to change (excluding #104116)

[spiffxp]

@pacoxu does this need promotion at all or are we good with the existing k8s.gcr.io/etcd:3.5.0-0 image?

[spiffxp]

/remove-priority critical-urgent
/priority important-soon
The job is back to passing

[pacoxu]

@pacoxu does this need promotion at all or are we good with the existing k8s.gcr.io/etcd:3.5.0-0 image?

Recent updates are related to Debian base image update and golang upgrade.
Nice to get a promotion. Good enough to use k8s.gcr.io/etcd:3.5.0-0.

[saschagrunert]

Yeah, looks like that the env in the build job does not search for docker CLI plugins in the /root/.docker/cli-plugins path, which makes it unavailable during the run.

[saschagrunert]

I think we have to set the $HOME environment variable like we already do in other build jobs, proposing the fix in #105484

[saschagrunert]

Latest test run is green: https://prow.k8s.io/view/gs/kubernetes-jenkins/logs/post-kubernetes-push-image-etcd/1445454283720888320