1600150 - new browsertime tasks don't get SETA optimized 24h after they got added

Status: RESOLVED FIXED

(Tree Management Graveyard :: Treeherder: SETA, defect)

Comment 1

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

Bug 1595789 added new browsertime tests and 24h after they got added, they still don't get SETA optimized.

Run https://sql.telemetry.mozilla.org/queries/65428/source?p_tree_name_65428_65428=mozilla-central&p_tree_name_65428=autoland#166400 for "autoland" and zoom in on the SETA optimized pushes (blue) on the 27th and 28th November.

The count of tasks per push has increased from ~260 to ~340.

The lists of scheduled tasks for pushes can be compared with https://sql.telemetry.mozilla.org/queries/65429?p_revision_old_65429=651a75&p_revision_new_65429=f182de&p_repository_65429=autoland (entering the first characters of a revision is sufficient).

Andrew, what would be the expected behavior here - should the tasks be SETA optimized now?

Comment 2

[Andrew Halberstadt [:ahal]]

I'm not sure. I know the data collection period was reduced, but not sure if 24hrs is the expected turn around or not. Maybe it was 48hrs? I'd needinfo Joel, but I guess we'll know whether it works or not by the time he's back.

Comment 3

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

The task names start with vismet and thus are not SETA optimized .

Tarek, is it intended that the vismet tasks run on every push? The browsertime tasks are getting optimized and most of them don't run on every push.

Comment 4

[Tarek Ziadé (:tarek)]

The vismet task are supposed to run only when the browsertime one gets triggers. e.g. it's a "subtask"
So we should not see any vismet- task triggered by itself

Comment 5

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

Because the vismet tasks don't get optimized out (because of the task name), the 'normal' browsertime also runs (because it's a dependency). Can you alter the task name to have it match with test- and not get scheduled for every push?

Comment 6

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

Tarek: Any ETA for a fix?

Bug 1590561 added more vismet tasks yesterday, all with short runtimes of ~1 minutes. Can they be run as one task? The risk of overhead between tasks is concerning.

Comment 7

[Tarek Ziadé (:tarek)]

Attachment: Bug 1600150 - new browsertime tasks don't get SETA optimized 24h after they got added

Comment 8

[Tarek Ziadé (:tarek)]

Try run https://treeherder.mozilla.org/#/jobs?repo=try&revision=537b979b4366d46a7476d5ccd10c8f52f973d52e

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/fdd283de605b
new browsertime tasks don't get SETA optimized 24h after they got added r=aryx

Comment 10

[Raul Gurzau (:RaulG)]

https://hg.mozilla.org/mozilla-central/rev/fdd283de605b