Flake in test/cmd/builds.sh: oc cancel-build build/ruby-sample-build-1

[liggitt]

seen in https://ci.openshift.redhat.com/jenkins/job/test_pull_requests_origin/7860 run from #6330:

Running hack/../test/cmd/builds.sh:113: executing 'oc cancel-build build/ruby-sample-build-1' expecting any result and text 'A cancellation event was already triggered for the build ruby-sample-build-1.'; re-trying every 0.2s until completion or 60.000s...
FAILURE after 59.535s: hack/../test/cmd/builds.sh:113: executing 'oc cancel-build build/ruby-sample-build-1' expecting any result and text 'A cancellation event was already triggered for the build ruby-sample-build-1.'; re-trying every 0.2s until completion or 60.000s: the output content test failed
Standard output from the command:
A build can be cancelled only if it has new/pending/running status.
A build can be cancelled only if it has new/pending/running status.
A build can be cancelled only if it has new/pending/running status.
A build can be cancelled only if it has new/pending/running status.
A build can be cancelled only if it has new/pending/running status.
A build can be cancelled only if it has new/pending/running status.

[smarterclayton]

I thought we had a known flake here @bparees on cancelling builds too fast.

[bparees]

doesn't ring a bell, and the above code is retrying for 60s, so it seems like the build never got created or got created and immediately failed, making it non-cancellable.

oc cancel-build should probably dump the current state of the build as part of the error message, that would tell us more about why cancel is failing.

[smarterclayton]

The error message is also scarier than it should be.

On Wed, Dec 16, 2015 at 1:47 PM, Ben Parees [email protected]
wrote:

doesn't ring a bell, and the above code is retrying for 60s, so it seems
like the build never got created or got created and immediately failed,
making it non-cancellable.

oc cancel-build should probably dump the current state of the build as
part of the error message, that would tell us more about why cancel is
failing.

—
Reply to this email directly or view it on GitHub
#6339 (comment).

[bparees]

The error message is also scarier than it should be.

how so?

[smarterclayton]

"A build can be cancelled only if it has new/pending/running status." ->
"Build %s is complete and cannot be cancelled". Also, I'm not sure that's
an error - that a no-op (like deleting something that doesn't exist).

On Wed, Dec 16, 2015 at 2:03 PM, Ben Parees [email protected]
wrote:

The error message is also scarier than it should be.

how so?

—
Reply to this email directly or view it on GitHub
#6339 (comment).

[liggitt]

was a race condition... if the cancellation was acted on quickly enough, the message would change to "must be in pending/running state"