`glean.database.write_time` metric can cause empty metrics pings to be sent
Categories
(Data Platform and Tools :: Glean: SDK, defect, P1)
Tracking
(Not tracked)
People
(Reporter: travis_, Assigned: travis_)
References
Details
Attachments
(1 file)
Since the addition of the glean.database.write_time metric to glean-core, we have seen an increase in metrics pings that contain only this metric and no others, i.e. "empty" metrics pings.
On mobile, these pings are overwhelmingly ping_info.reason=upgrade, which very well could be "empty" if there were no activity between the last metrics ping sent and the current run that includes an upgrade. This would typically not be sent as there would be no data to send, but because of how the write_time metric is instrumented, it is guaranteed to be there, which forces an otherwise "empty" metrics ping to be sent.
We see this across all glean consumers, but Desktop is slightly different in that we see the spike with reason=overdue instead of upgrade. This may just be due to the nuance in how desktop sends pings on upgrade because it otherwise lines up with the other platforms.
It also appears that Fenix's earlier spike in August may be unrelated since it was showing up in pings with reason=overdue, leading me to believe this might be something separate.
Refs:
Fenix: https://sql.telemetry.mozilla.org/queries/102427/source#252873
Firefox iOS: https://sql.telemetry.mozilla.org/queries/102607/source#252872
Desktop: https://sql.telemetry.mozilla.org/queries/102430/source#252416
In order to fix this, I propose we set the metric as disabled but leave it instrumented. This way it can be enabled again via Server Knobs, if needed, but should eliminate the mostly empty metrics pings we are seeing caused by it.
|
||
Comment 1•17 days ago
|
||
|
Assignee | |
Updated•17 days ago
|
Description
•