-
Type: Task
-
Resolution: Fixed
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
Atlas Streams
-
Fully Compatible
-
Sprint 32, Sprint 33, Sprint 34, Sprint 35, Sprint 36, Sprint 37
$tumblingWindow, $group, $sort, $lookup can have arbitrarily high memory usage. We need to detect all such high memory usage scenarios and throw necessary exceptions to prevent the pod from OOMing.
Filing this ticket to for tracking purposes. We should break up this ticket into multiple tickets eventually. Given how important it is to get this right, we would need a design doc for this.
Our behavior here also needs to be clearly documented for our users.
- related to
-
SERVER-84005 Streams: prevent SP process from being unschedulable after encountering the first OOM signal
- Closed