-
Type: Bug
-
Resolution: Unresolved
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
Storage Execution
-
ALL
-
Execution Team 2024-10-28, Execution Team 2024-11-25
-
(copied to CRM)
In the following series of events:
- Index build starts. It is journaled and written to the oplog.
- Node is shutdown uncleanly
- A checkpoint was not taken
The node can be left in a state at startup where the index build does not exist on the catalog but exists in the oplog.
This can be a problem when a problematic index build is the cause of unclean shutdown as we'll likely end up in a restart loop (node crashes during startup from the index build, restarts, crashes again). We are unable to drop the index build in standalone because it is not reconciled to the catalog. recoverFromOplogAsStandalone=true is also unhelpful in this case as it is a read-only mode and thus unable to drop the index while the problematic index build is in progress.
- is related to
-
SERVER-68982 Key generation on single document can exceed index build sorter limit
- Backlog
-
SERVER-38705 Remove the --noIndexBuildRetry server parameter flag and the storage.indexBuildRetry config file option
- Closed