Test flake in cmd/quota.sh: secrets not showing up

[miminar]

FAILURE after 60.119s: test/cmd/quota.sh:25: executing 'oc describe appliedclusterresourcequota/for-deads-by-annotation -n bar --as deads' expecting any result and text 'secrets.*18'; re-trying every 0.2s until completion or 60.000s: the command timed out
Standard output from the command:
Name:       for-deads-by-annotation
Namespace:  <none>
Created:    2 seconds ago
Labels:     <none>
Annotations:    <none>
Label Selector: <null>
AnnotationSelector: map[openshift.io/requester:deads]
...

... repeated 2 times
Name:       for-deads-by-annotation
Namespace:  <none>
Created:    About a minute ago
Labels:     <none>
Annotations:    <none>
Label Selector: <null>
AnnotationSelector: map[openshift.io/requester:deads]
Resource    Used    Hard
--------    ----    ----
secrets     17  50
... repeated 3 times
There was no error output from the command.
[ERROR] jUnit test marker could not be placed, expected one test in flight, got 0
[ERROR] PID 2608: hack/lib/cmd.sh:95: `return 1` exited with status 1.
[INFO]      Stack Trace: 
[INFO]        1: hack/lib/cmd.sh:95: `return 1`
[INFO]        2: hack/lib/cmd.sh:139: os::cmd::internal::run_until_text
[INFO]        3: test/cmd/quota.sh:25: os::cmd::try_until_text
[INFO]   Exiting with code 1.

See job 3506 for details.

[deads2k]

@stevekuznetsov The etcd dump for the linked build doesn't have any data in it. Any idea where it went?

[stevekuznetsov]

Looking ... the jUnit stuff failed as well. Weird.

[deads2k]

Looking ... the jUnit stuff failed as well. Weird.

Thanks. I'd like to know why my secrets aren't getting created. Getting less than 18 secrets means that something won't work.

[stevekuznetsov]

I mean let's be honest here, 18 is a lot of secrets. Can't we just live with, 15 or so?

[stevekuznetsov]

As an aside, can you do a jsonpath into the object instead of a grep on the output of describe?

[stevekuznetsov]

Fixed the jUnit bug. but since the cleanup for test-cmd is still a) special and b) inside set +o errexit we really can't tell if that curl succeeded or not, sorry. There's no logs and I guess it just failed silently?

[deads2k]

This is the "off by one" (off by X since we started batching the updates) that is caused by a reconciliation check that doesn't get a complete list that includes secrets about to be created. The design here: kubernetes/kubernetes#20113 ought to resolve it.

I'll address this by loosening the regex to eliminate the flake.