Details
-
Bug
-
Resolution: Unresolved
-
Major - P3
-
None
-
None
-
None
-
None
-
Query Integration
-
ALL
Description
We migrate the whole buckets during chunk migration, so the chance of collision is higher for time-series collection since we prefix the first 4 bytes of _id with timestamp. So there are fewer bytes to generate a unique id across the cluster.
We can consider a workaround to regenerate the bucket (with a new _id) when a chunk migration for time-series collection fails because of _id collision