Currently, when we fill a time-series bucket and roll over to a new one, we can in some cases generate the same bucket OID. This is because we round the timestamp down based on the collection granularity. If we end up with the same rounded timestamp portion of the OID, the rest of the OID may be the same as the previous one.
- causes
-
SERVER-61412 Avoid time-series bucket OID collisions after election
- Backlog
- is duplicated by
-
SERVER-61151 batch->getResult().getStatus() == ErrorCodes::TimeseriesBucketCleared
- Closed