Dynamic PV's not being attached to nodes on GCE

[soltysh]

I'm seeing following StatefulSet extended failures (usually due to timeouts) both in master and my rebase PR (https://ci.openshift.redhat.com/jenkins/job/test_pr_origin_extended/984/consoleFull).

[Fail] [k8s.io] StatefulSet [k8s.io] Basic StatefulSet functionality [It] should allow template updates 
/data/src/github.com/openshift/origin/vendor/k8s.io/kubernetes/test/e2e/petset.go:923

[Fail] [k8s.io] StatefulSet [k8s.io] Basic StatefulSet functionality [It] should provide basic identity 
/data/src/github.com/openshift/origin/vendor/k8s.io/kubernetes/test/e2e/petset.go:923

[Fail] [k8s.io] StatefulSet [k8s.io] Basic StatefulSet functionality [It] should handle healthy pet restarts during scale 
/data/src/github.com/openshift/origin/vendor/k8s.io/kubernetes/test/e2e/petset.go:923

@smarterclayton fyi

[smarterclayton]

Stateful set looks like it's failing because of:

  FirstSeen	LastSeen	Count	From			SubObjectPath	Type		Reason			Message
  ---------	--------	-----	----			-------------	--------	------			-------
  10m		45s		36	{default-scheduler }			Warning		FailedScheduling	SchedulerPredicates failed due to PersistentVolumeClaim is not bound: "datadir-pet-0", which is unexpected.

?

[soltysh]

This should be fixed in #12642.

[smarterclayton]

Failing on GCE now https://ci.openshift.redhat.com/jenkins/job/test_pull_requests_origin_gce/188/testReport/junit/(root)/Extended/_k8s_io__StatefulSet__k8s_io__Basic_StatefulSet_functionality_should_provide_basic_identity/

[soltysh]

Looking...

[soltysh]

So the PR of mine changed setup.sh specifically to allow hostPath mounts, but gce tests is starting openshift differently. I need to update appropriate component for that. I'll send a PR shortly, I first need to figure out how it starts.

[smarterclayton]

You need to fix the tests. They are not allowed to access hostpaths.

[smarterclayton]

probably by setting privileged like @pmorie and @ncdc did for the other tests.

[smarterclayton]

This looks like PVs are not being mounted correctly to nodes on 1.5 on GCE. I see node status attached, the PVC is bound, but the actual disk in GCE is not being attached to the instance. I'll need some help debugging from sos-storage.

[smarterclayton]

Disabling the tests while this is investigated https://ci.openshift.redhat.com/jenkins/job/test_pull_requests_origin_gce/216/#showFailuresLink was the last failing test.

[smarterclayton]

This is from post 1.5.2 update.

[childsb]

@rootfs

[rootfs]

upstream pr created kubernetes/kubernetes#41034

[rootfs]

@soltysh @smarterclayton fix in #12835