1683894 - Large increase in nursery collections on 2020-12-16

Status: RESOLVED FIXED

(Core :: JavaScript: GC, defect, P1)

Description

[Jon Coppeard (:jonco)]

There was a large spike in the number of nursery collections performed on the 16th December:

https://telemetry.mozilla.org/new-pipeline/evo.html#!aggregates=Median!Mean!5th%2520percentile!25th%2520percentile!75th%2520percentile!95th%2520percentile&cumulative=0&end_date=null&include_spill=0&keys=&max_channel_version=nightly%252F86&measure=GC_MINOR_US&min_channel_version=nightly%252F85&processType=*&product=Firefox&sanitize=1&sort_keys=submissions&start_date=null&trim=1&use_submission_date=0

This is most likely a result of landing bug 1677765.

Comment 1

[Jon Coppeard (:jonco)]

Attachment: Bug 1683894 - Don't trigger GC slices while waiting for unmarking to finish r?sfink

The problem here is that when we're unmarking on a background thread there's
nothing for the main thread GC to do, but the browser GC scheduling doesn't
know this and will happily trigger GC slices based on its timer / idle time.
This has the unfortunate effect that these repeated nursery collection will
shrink the nursery down to its minimum size pretty quickly.

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/ed0b7d9cd487
Don't trigger GC slices while waiting for unmarking to finish r=sfink

Comment 3

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/ed0b7d9cd487

Comment 4

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1677765