-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: Index Maintenance
-
None
-
Fully Compatible
-
ALL
-
Execution Team 2020-08-24
When trying to turn on the resumable index build behavior by default, we found that it's possible for a node to be hung during startup recovery after an unclean shutdown while applying a commitIndexBuild oplog entry. The node is stuck waiting for the last optime before it established the interceptor to be majority committed even though other nodes in the replica set have applied entries from a later term than the optime the node is waiting on.
- is related to
-
SERVER-48419 Extend rollback to recover resumable index builds efficiently
- Closed
-
SERVER-48476 resumable index build should use majority read cursor for collection scan
- Closed
-
SERVER-49847 Turn on resumable index build feature by default
- Closed
- related to
-
SERVER-50525 Resumable index build hangs waiting for majority commit point in backup_restore_backup_cursor.js
- Closed