1543935 - failures while doing retriggers from push-health

Status: RESOLVED FIXED

(Tree Management :: Push Health, defect)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

in bug 1541612, the ability to retrigger a job from inside push health was added. In doing this, we find that the action results in failure.

What happens with a retrigger is we execute a taskcluster action add-new. The add-new looks at the taskgraph and adds another job to it with what you specified. In this case the add-new is failing.

For example look at this try push:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=2ad35432151aa18a59d29b594d98906dd724fffd&searchStr=add-new

there are many traditionally done retriggers and one done via push-health.

Here is the failure from the push health failure:
https://taskcluster-artifacts.net/Wj7AFJnEQ9qvQUNXZX5xOQ/0/public/logs/live_backing.log

[task 2019-04-11T15:21:39.236Z] No label-to-taskid.json found for QEowgTYRQNeT3m2qRLkNdQ: 404 Client Error: Not Found for url: https://queue.taskcluster.net/v1/task/QEowgTYRQNeT3m2qRLkNdQ/artifacts/public/label-to-taskid.json
[task 2019-04-11T15:21:39.237Z] Traceback (most recent call last):
[task 2019-04-11T15:21:39.237Z] File "/builds/worker/checkouts/gecko/taskcluster/mach_commands.py", line 272, in action_callback
[task 2019-04-11T15:21:39.237Z] test=False)
[task 2019-04-11T15:21:39.237Z] File "/builds/worker/checkouts/gecko/taskcluster/taskgraph/actions/registry.py", line 364, in trigger_action_callback
[task 2019-04-11T15:21:39.237Z] cb(Parameters(**parameters), graph_config, input, task_group_id, task_id)
[task 2019-04-11T15:21:39.237Z] File "/builds/worker/checkouts/gecko/taskcluster/taskgraph/actions/add_new_jobs.py", line 56, in add_new_jobs_action
[task 2019-04-11T15:21:39.237Z] raise Exception('{} was not found in the task-graph'.format(elem))

It looks like we just don't have the proper information sent to the action task. Here is the code for retriggerJob() that works for a single task from actionbar:
https://github.com/mozilla/treeherder/blob/e5fb5d61af317e2173c015f47be25d544d03e8c5/ui/job-view/details/summary/ActionBar.jsx#L337

and here is the code that is similar but inside of push health:
https://github.com/mozilla/treeherder/blob/e5fb5d61af317e2173c015f47be25d544d03e8c5/ui/push-health/TestFailure.jsx#L77

both of these end up calling jobmodel.retrigger, and from inspecting the task, it seems like the input.tasks is null:
https://github.com/mozilla/treeherder/blob/e29e4a4076f920b979eb3e9bab821edaf566f50c/ui/models/job.js#L153

my best guess is that from push health we are not passing in the same job object as we do from the action bar.

Comment 1

[GitHub Bugzilla PR Linker]

Attachment: Link to GitHub pull-request: https://github.com/mozilla/treeherder/pull/4913