New issue
Advanced search Search tips
Note: Color blocks (like █ or █) mean that a user may not be available. Tooltip shows the reason.

Issue 878940 link

Starred by 1 user
Status: Started
Owner:
chanli@chromium.org
Cc:
█ robert...@chromium.org
chanli@chromium.org
lijeffrey@chromium.org
liaoyuke@chromium.org
st...@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

[Findit] Metrics: gather duration metrics for each analysis

Project Member Reported by chanli@chromium.org, Aug 29 
Add some CumulativeDistributionMetric to track
- pending time of heuristic analyses for consistent failures
- running time of heuristic analyses for consistent failures
- pending time of swarming task
- running time of swarming task
- pending time of try job
- running time of try job

Add graphs accordingly to display 95% and 99% durations and add alerts for spikes of durations.

Comment 1 by st...@chromium.org, Aug 29 

I think we have pending time for try-jobs already. You may talk to robertocn@ on these metric set up.

A follow-up question is: if those alerts are raised, what action are expected? If we can't do anything, the alert might not be useful.

Comment 2 by chanli@chromium.org, Aug 29 

The existing pending try job metric is from buildbucket, so we may be able to see a change of the pending time but not sure what causes it easily.

For the try job pending metric, the new one I plan to add will include more info, such as master/builder etc, so it might be easier to spot on the causes. 

And I also think alerts need to be actionable. So I'll think more on that part.

Comment 3 by st...@chromium.org, Aug 29 

Roberto, what do you say about the try-job pending alert? We used to have that, and then deleted that to be in favor of the buildbucket one IIRC.
►

Sign in to add a comment