Test-infra: gke-ubuntustable1-k8sstable2-ingress misbehaving?

[MrHohn]

From http://k8s-testgrid.appspot.com/sig-network-gke#gke-ubuntustable1-k8sstable2-ingress&width=20, frequently one run stuck forever and led to all following runs failed while deleting network (in used by a firewall rule created by the stuck run).

Also curious why different runs in this CI use the same network name if they create&delete network anyway...

BTW I never know who owns this CI :) Probably @yguo0905 knows?

[yguo0905]

I can change the tests to use the project pool to mitigate the issue caused by the stuck run. @krzyzacy

GKE tests should be owned by sig-gcp now.

[krzyzacy]

sure, we have enough resource for gke pool

[MrHohn]

Thanks for the quick replies.

Another thought is I don't actually see much value running these ubuntu-ingress CIs as we already have equivalent gke-ingress CIs AND ingress tests are really not OS sensitive...I'm leaning toward just nuke these CIs to save resource and reduce noise.

[MrHohn]

I'm leaning toward just nuke these CIs to save resource and reduce noise.

Discussed offline and seems better to keep these CIs.

I can change the tests to use the project pool to mitigate the issue caused by the stuck run.

@yguo0905 Let me know how I can help :)