test: Detect flakes caused by container teardown races on CRI

CRI returns a different reason than the dockershim, and we must also
catch the symptoms of https://github.com/kubernetes/kubernetes/issues/88766
in that environment.
This commit is contained in:
Clayton Coleman
2020-03-08 16:21:57 -04:00
parent 672aa55ee4
commit d80981b657

View File

@@ -350,7 +350,7 @@ var _ = SIGDescribe("Pods Extended", func() {
switch {
case t.ExitCode == 1:
// expected
case t.ExitCode == 128 && t.Reason == "ContainerCannotRun" && reBug88766.MatchString(t.Message):
case t.ExitCode == 128 && (t.Reason == "StartError" || t.Reason == "ContainerCannotRun") && reBug88766.MatchString(t.Message):
// pod volume teardown races with container start in CRI, which reports a failure
framework.Logf("pod %s on node %s failed with the symptoms of https://github.com/kubernetes/kubernetes/issues/88766")
default: