Closed Bug 1541544 Opened 6 years ago Closed 5 years ago

Random freezes in foreground tab, caused by treeherder work in background tab (which may be sharing the same content process)

Categories

(Tree Management :: Treeherder: Frontend, defect, P3)

Desktop
Linux
defect

Tracking

(Performance Impact:none)

RESOLVED FIXED
Performance Impact none

People

(Reporter: sefeng, Assigned: camd)

References

(Blocks 1 open bug)

Details

(Keywords: hang, perf)

I am on 20190402083512 nightly, and the web version of slack starts to randomly freeze very frequently. The tab is pinned.

about:support: https://pastebin.com/RTzZGf9U

https://perfht.ml/2YLuvNQ, https://perfht.ml/2YQeN4b

Whiteboard: qf → [qf]

Looks like treeherder is causing this. handleApplyNewJobs is taking a long time (5-8 seconds), in both profiles, and monopolizes the main thread in that process, which presumably slack is sharing.

Probably best to move this to treeherder?

Component: General → Treeherder: Frontend
Keywords: hang, perf
Product: Core → Tree Management
Summary: Radom freezes on slack → Random freezes on slack caused by treeherder
Version: unspecified → ---

Yikes! Yeah, here's the first profile showing handleApplyNewJobs taking ~8 seconds:
https://perfht.ml/2U1ZnpG

Second profile:
https://perfht.ml/2U1ZnpG

So yeah, treeherder folks should look into breaking this work up if they can.

Whiteboard: [qf] → [qf-]

(Denis notes that Fission would help here, though)

This is definitely in my wheel-house to fix.

Assignee: nobody → cdawson
Status: NEW → ASSIGNED
Priority: -- → P1

[dropping "slack" from bug title, since there's nothing Slack-specific here; this will just manifest as jank in whatever tabs are sharing a process with treeherder tabs.]

Summary: Random freezes on slack caused by treeherder → Random freezes in other tabs caused by treeherder work in background tab
Summary: Random freezes in other tabs caused by treeherder work in background tab → Random freezes in foreground tab, caused by treeherder work in background tab (which may be sharing the same content process)

There is certainly room for improvement still, but this should be MOSTLY remedied by recent changes.

Blocks: 1499551
Priority: P1 → P3

Has anyone seen this recently? Our recent deploys should have helped with this. I'm going to close for now. If you see this some more, please reopen. Thanks!

Status: ASSIGNED → RESOLVED
Closed: 5 years ago
Resolution: --- → FIXED
Performance Impact: --- → -
Whiteboard: [qf-]
You need to log in before you can comment on or make changes to this bug.