1167419 - UI for detail panel / failure summary tab on pending or running jobs keeps reloading constantly

Status: RESOLVED FIXED

(Tree Management :: Treeherder, defect, P1)

Description

[Cameron Dawson [:camd]]

It should not try to reload that panel for pending or running jobs because it's looking for a ``Bug suggestions`` artifact that won't exist till the job is complete.

It should only do the check once the state is complete for the job.

Comment 1

[Cameron Dawson [:camd]]

This worked normally on the first job you select.  But as you click different jobs, clear selected job and reselected new ones, it was possible to leave timeout loops around still looking for bug suggestions.

This PR stops the loops when you switch jobs and clear the selected job.

Comment 2

[Cameron Dawson [:camd]]

Attachment: fix failures summary tab reloading

Comment 3

[Treeherder GitHub Bugbot]

Commit pushed to master at https://github.com/mozilla/treeherder

https://github.com/mozilla/treeherder/commit/4096a1d1d99fb668275f65047006ceb3148240bc
Bug 1167419 - fix constant reloads on failure summary tab

Comment 4

[Jonathan French (:jfrench)]

When I run latest master w/vagrant using ./bin/run_gunicorn and local ingestion, I still see completed jobs reloading in the job panel tabs every 5 seconds.

Steps to reproduce:
o pull latest master
o re-initialize your db, do all the other server setup steps (I did this just to be sure..)
o (venv) ./bin/run_gunicorn
o (venv) celery -A treeherder worker -B
o http://local.treeherder.mozilla.org
o give it some time for jobs to ingest..
o select a completed Successful job
o select say, the Annotations tab

Expected:
It sits quietly on that tab.

Observed:
It refreshes after 5 seconds and jumps to the Job details tab.

If I point my local.conf.js to production or stage and restart gunicorn only, it behaves correctly and doesn't reload. I'm curious if you can reproduce that camd?

Comment 5

[Cameron Dawson [:camd]]

I actually CAN reproduce that, yeah. 

One of the differences I noticed between service on prod vs. service locally is that the job details panel is blank locally and the log parsing status is "pending."  The panel is supposed to refresh when the log is not yet parsed (to look for new info that arrives once it is.)

Maybe there's a way to detect that the job has not changed on the refresh and prevent changing the tab to the default?

Comment 6

[Jonathan French (:jfrench)]

Thanks camd. This seems to be a change in behavior of our current venv `celery -A treeherder worker -B` which at least for me doesn't appear to be parsing log files anymore. That behavior started roughly 1-1/2 weeks ago. All my log parsing for every job remains 'pending' even when I click on the job to put it higher in the queue, so the Tabs continually reload.

I switched to using supervisord and the log parser started and my local server behaved normally (ie. like `celery..` above used to).