-
Type: Bug
-
Resolution: Duplicate
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
ALL
-
Execution Team 2022-09-05, Execution Team 2022-09-19, Execution Team 2022-10-03
-
5
Similar to SERVER-63865, replica set nodes can fail to startup after an unclean shutdown if we didn't finish an index build (two phase).
After a node crashes during an index build, we can startup and have the correct catalog entry but not have the corresponding index table. During index build recovery, we check that the collection is initialized. This code path leads us to initialize the index catalog and eventually we try to access the index table.
We likely need to generalize the solution in SERVER-63865 to apply to nodes in startup recovery in general, regardless of whether they are standalone nodes, to prevent opening index tables that may not exist.
- depends on
-
SERVER-69877 Remove untimestamped writes to the catalog when restarting unfinished index builds during startup recovery
- Closed
- duplicates
-
SERVER-69877 Remove untimestamped writes to the catalog when restarting unfinished index builds during startup recovery
- Closed
- related to
-
SERVER-63865 Handle missing index idents during standalone startup recovery after unclean shutdown
- Closed