`glean.error.preinit_tasks_overflow` includes the count of non-overflown tasks as well as the number of tasks we dropped
Categories
(Data Platform and Tools :: Glean: SDK, defect, P2)
Tracking
(Not tracked)
People
(Reporter: chutten, Assigned: janerik)
Details
Attachments
(1 file)
To be useful and less confusing, I think glean.error.preinit_tasks_overflow should only count the number of tasks we didn't enqueue in the preinit queue (ie, the count of tasks that overflowed out of the queue) and not include (as it does now) the number of tasks in the queue.
If we really want to know the size of the queue, we can record that using a quantity (or we can read it out of the code by cross-referencing the sdk version).
This might be the perfect time to develop some best practices around the use of the version field in metrics.yaml.
In the meantime, I'll add a Gleannotation to highlight that the Description isn't kidding.
|
Assignee | |
Updated•2 years ago
|
|
|
Assignee | |
Comment 1•2 years ago
|
||
So I quickly looked into the history of this metric, but couldn't find any definitive answer why we report the total number of tasks and not just the number of overflowing ones.
The only justification we got for the metric itself is "We want to know not just "did it overflow" but "by how much".
But only reporting the number of overflows would explain that excatly.
I wouldn't mind us changing that metric now. It's not used externally anyway (or at least no one but us should ever look at it).
We could increase the version, though none of our tooling handles that field afaik, it also shouldn't break.
|
|
Assignee | |
Comment 2•2 years ago
|
||
|
|
Assignee | |
Comment 3•2 years ago
|
||
badboy merged PR #2026: "Bug 1764573 - Only report the number of overflowing tasks in preinit_tasks_overflow metric" in d0f7836.
|
|
Assignee | |
Comment 4•2 years ago
|
||
Note: https://dictionary.telemetry.mozilla.org/apps/fenix/metrics/glean_error_preinit_tasks_overflow will hopefully show version: 1 in a day or 2.
Description
•