-
Type: Task
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
-
None
-
Storage Execution
-
Fully Compatible
-
Execution Team 2024-01-22, Execution Team 2024-02-05
-
166
Right now we're maintaining the current behaviour where if bucket compression fails, we leave the bucket uncompressed. When bucket compression fails, chances are the bucket is corrupt in some way, and further writes to it could result in data loss. Instead, we should fail the insert, close the bucket, and retry the insert without re-opening buckets. We cannot re-open buckets as we can end up in a scenario where we keep retrying the insert in the same bucket. We'll also need to limit the number of retries.
- is related to
-
SERVER-86072 Handle time-series bucket compression failures consistently between ordered and unordered inserts
- Closed
-
SERVER-86073 Add a way to track whether the number of frozen time-series buckets is increasing
- Closed
-
SERVER-86087 Prevent frozen time-series buckets from being tracked in the registry forever
- Closed
-
SERVER-86085 Improve naming/semantics around frozen time-series buckets and reopening
- Closed
- related to
-
SERVER-84234 BSONColumnBuilder throwing should be handled gracefully
- Closed