Open Bug 1403884 Opened 2 years ago Updated Last month

Netmonitor causing long hangs?

Categories

(DevTools :: Netmonitor, defect, P3)

defect

Tracking

(firefox57 fix-optional)

Tracking Status
firefox57 --- fix-optional

People

(Reporter: smaug, Unassigned)

References

(Blocks 2 open bugs)

Details

(Whiteboard: [bhr-html])

Attachments

(1 file)

Attached image netmonitor_hangs.png
I'm looking at http://arewesmoothyet.com.
I don't know how to share links to that, so here is a screenshot.
Whiteboard: [bhr-html]
It looks like bug 1404917 finally fixed the very long hangs, over 1s.

https://arewesmoothyet.com/?mode=track&trackedStat=Devtools%20Hangs
(you have to click on thresholds label <1s to really see that the slowest hangs are very rare)

But some work is still pending in bug 1404913, which should continue removing hangs related to netmonitor.

Keeping open to still track the other smaller hangs.
Depends on: 1404913
Just looked at the data and netmonitor is still there.
Bug 1404913 only had an impact on hangs >=1s.
I'm a bit surprised, I thought it would lead to a much more important improvement.
We used to copy all requests response content. Always, for all requests, no matter if we actually opened the response content panel. We no longer do that since bug 1404917 landed.
It looks like BHR hangs aren't necessarely reflecting all slownesses equally. It seems like slow reflows have a very big impact on the number of BHR hangs being reported. I understood that while writing Q4 report [1].
A very naive fix on the console (bug 1400176) reduced the number of hangs associated to the console by 2x or 4x!
I'm currently investigating slow reflows related to netmonitor in bug 1425978.

[1] https://docs.google.com/document/d/e/2PACX-1vQ3ZxZR8muRVND9ghZD8myEYRR0k2IEg7Md_cC4r9G8WXtUoTkbzOuz-ZD-U3VwxWSnyzRt0MmvhL_G/pub
Product: Firefox → DevTools
You need to log in before you can comment on or make changes to this bug.