Jobs should not be restarted indefinitely
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Zuul |
New
|
Undecided
|
Unassigned |
Bug Description
Zuul is willing to restart a job if it ends without Jenkins providing a status. In some cases Jenkins jobs can fail consistently in such a way that Zuul willingly loops them over and over forever. An example of a job where this behavior was seen--a test hangs and when Jenkins tries to kill it for the configured timeout, an exception is thrown communicating with the slave agent...
test_
test_leak_1 Build timed out (after 60 minutes). Marking the build as failed.
++ rm -rf /tmp/tmp.ojOMmvvlVW
[31mERROR[0m[31m 3438.30[0m
test_
hudson.
at hudson.
at hudson.
at hudson.
at hudson.
at hudson.
at hudson.
at hudson.
at hudson.
at hudson.
at hudson.
Caused by: hudson.
at hudson.
at hudson.
at hudson.
Caused by: java.io.
at hudson.
Caused by: java.io.
at java.io.
at java.io.
at java.io.
at hudson.
at hudson.
at hudson.
At first I thought the test loop was due to this patchset getting switched to a draft after testing began, but were that the case switching it back to non-draft should have allowed it to complete. Investigating logs on zuul.openstack.org it looks like Jenkins does not report a completion status for this job.
summary: |
- Switching patchset to draft causes indefinite job loop + Jobs should not be resubmitted indefinitely |
summary: |
- Jobs should not be resubmitted indefinitely + Jobs should not be restarted indefinitely |
description: | updated |