Start using gangplank to build cloud artifacts
/hold Checking:
14:42:03 time="2021-05-11T17:42:03Z" level=fatal msg="failed to define builder pod: failed setting in-cluster options: pods \"pod-9f110ac1-fc9c-438f-ad93-a5a7589b7020-b6pd3-lfmj6\" is forbidden: User \"system:serviceaccount:coreos-ci:default\" cannot get resource \"pods\" in API group \"\" in the namespace \"coreos-ci\""
/hold Checking: `14:42:03 time="2021-05-11T17:42:03Z" level=fatal msg="failed to define builder pod: failed setting in-cluster options: pods "pod-9f110ac1-fc9c-438f-ad93-a5a7589b7020-b6pd3-lfmj6" is forbidden: User "system:serviceaccount:coreos-ci:default" cannot get resource "pods" in API group "" in the namespace "coreos-ci""
I'd recommend that we use the jenkins service account for agent pods (it has sufficient perms). I think adding serviceAccount: "jenkins" near https://github.com/coreos/coreos-ci-lib/blob/main/resources/com/github/coreos/pod.json#L8 would do the trick.
The reason -a jenkins isn't working is because the default service account in the pod doesn't have the ability to create pods. We have a chicken-and-egg problem: we need a user that has the ability to create a pod and then -a jenkins would work.
I'd recommend that we use the
jenkinsservice account for agent pods (it has sufficient perms). I think addingserviceAccount: "jenkins"near https://github.com/coreos/coreos-ci-lib/blob/main/resources/com/github/coreos/pod.json#L8 would do the trick.
Indeed I had to use that for testing as well. I just create this PR to allow it
/unhold
Gangplank is failing now.
The failure is not in Gangplank so much as in the tar command. I spend hours debugging this, and AFAIK this is a issue with NFS mount/umount changing under the pod which triggers a "file changed" error under Tar.
/retest
@ravanelli: PR needs rebase.
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.