Platform Microbencharmark alerts are still generated for Linux platforms

NEW
Assigned to

Status

enhancement
2 years ago
2 years ago

People

(Reporter: igoldan, Assigned: wlach)

Tracking

57 Branch
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

We still get alerts like https://treeherder.mozilla.org/perf.html#/alerts?id=8597 or https://treeherder.mozilla.org/perf.html#/alerts?id=8594 even though bug 1369807 silenced them explicitly.

Log file on Linux's GTest further indicate that no alerts should be triggered:
[task 2017-08-04T17:41:53.405710Z] 17:41:53     INFO -  TEST-START | Strings.PerfStripCharsWhitespace
[task 2017-08-04T17:41:58.438820Z] 17:41:58     INFO -  PERFHERDER_DATA: {"framework": {"name": "platform_microbench"}, "suites": [{"name": "Strings", "subtests": [{"name": "PerfStripCharsWhitespace", "value": 1007364, "replicates": [1007364,1006496,1008299,1006889,1008311], "lowerIsBetter": true, "shouldAlert": false}]}]}
So we have to make sure we completely disable alerts from Linux.
Will take a look at this tomorrow.
Assignee: nobody → wlachance
I don't understand how this can be happening: I looked at the database and should_alert for the signature is false as we expect, so the alert very clearly shouldn't be generated:

https://github.com/mozilla/treeherder/blob/ed1ec920516aea3e5e1f6df7946e514b3d08bf92/treeherder/etl/perf.py#L177

I changed this logic recently in bug 1387562 to solve an issue related to alerts being generated on wrong branches, but I don't see how that would make a difference in this case. Let's leave this for a week or so and see if the problem reoccurs.
You need to log in before you can comment on or make changes to this bug.