Details
Description
In HBASE-28210, we need to synchronized on RootProcedureState when updating procedure store, which will decrease the performance a lot when there are lots of sub procedures for a procedure, such as SCP, so it will impact MTTR.
But actually, SCP does not support rollback, so I do not think we need to maintain this information. If so, I think we do not need the synchronization added in HBASE-28210, at least for SCP then. This could get the performance back.
Attachments
Issue Links
- breaks
-
HBASE-28248 Race between RegionRemoteProcedureBase and rollback operation could lead to ROLLEDBACK state be persisent to procedure store
- Resolved
- is related to
-
HBASE-28210 There could be holes in stack ids when loading procedures
- Resolved
- links to