RLB's dispatcher's preinit queue size is too small for FOG
Categories
(Data Platform and Tools :: Glean: SDK, defect, P1)
Tracking
(Not tracked)
People
(Reporter: chutten, Assigned: chutten)
References
Details
Attachments
(3 files)
The preinit buffer in the Glean SDKs' dispatchers is 100. I believe I remember this is to limit the max RAM impact of the SDK not being init until later in startup.
In FOG's case, Glean's operating in the same environment where Legacy Firefox Telemetry keeps everything in RAM, so a low limit isn't necessary. We might not even want any limit.
This recently came to a head where a lot of metrics tasks were being queued in a test situation, causing data loss. And there's plenty of evidence that the limit's too low in the real world, too.
Ideally we'd make this limit configurable with an option to make it limitless, but for now this bug is about bumping GLOBAL_DISPATCHER_LIMIT
from 100 to 1000.
Comment 1•3 years ago
|
||
Assignee | ||
Comment 2•3 years ago
|
||
Assignee | ||
Comment 3•3 years ago
|
||
Not closing this off as we still need
- A Glean SDK release
- To vendor in the updated Glean SDK release
Shouldn't be too long.
Assignee | ||
Comment 4•3 years ago
|
||
Assignee | ||
Comment 5•3 years ago
|
||
Depends on D143749
Comment 7•3 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/a53c3c4105b7
https://hg.mozilla.org/mozilla-central/rev/63bd3eac0849
Description
•