Closed Bug 1862597 Opened 1 year ago Closed 1 year ago

Build profiles still have some spikes in their CPU use charts

Categories

(Testing :: Mozbase, defect, P3)

defect

Tracking

(firefox123 fixed)

RESOLVED FIXED
123 Branch
Tracking Status
firefox123 --- fixed

People

(Reporter: florian, Assigned: florian)

References

Details

Attachments

(2 files)

Example profile: https://share.firefox.dev/3Qpn2PU

When zooming on a spike (eg. https://share.firefox.dev/3Mqn4pK) we see:

  • There's a sample that's much longer than expected
  • There's a short ~50ms sample.
  • In the tooltips, the "CPU Percent" value is 100% for both samples.
  • User+nice+system has a smaller value than expected (total isn't 100%) in the first long sample, and much larger than expected (total is more than 400%) in the second shorter sample.

There's likely an issue in the order of these lines: https://searchfox.org/mozilla-central/rev/11d085b63cf74b35737d9c036be80434883dd3f6/testing/mozbase/mozsystemmonitor/mozsystemmonitor/resourcemonitor.py#139-143

I think we would need to take the timestamp right after (or before?) the call to psutil.cpu_times.

The severity field is not set for this bug.
:gbrown, could you have a look please?

For more information, please visit BugBot documentation.

Flags: needinfo?(gbrown)
Severity: -- → S3
Flags: needinfo?(gbrown)
Priority: -- → P3
Assignee: nobody → florian
Status: NEW → ASSIGNED
Pushed by fqueze@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/bdb5fb154d34 in resourcemonitor.py, take the timestamps right after the call to psutil.cpu_times, r=jcristau. https://hg.mozilla.org/integration/autoland/rev/82efe64cc658 when generating a profile from resourcemonitor.py, ensure the CPU use values used to display the chart are never above 100%, r=jcristau.
Status: ASSIGNED → RESOLVED
Closed: 1 year ago
Resolution: --- → FIXED
Target Milestone: --- → 123 Branch
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: