-
Type:
Bug
-
Resolution: Fixed
-
Priority:
Minor - P4
-
Affects Version/s: None
-
Component/s: None
-
Cluster Scalability
-
Fully Compatible
-
ALL
-
200
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Summary
The test jstests/multiVersion/resharding_version_pin_with_setfcv.js (added in SERVER-127574) fails intermittently on the multiversion_auth suite with:
timed out waiting for rangeDeletions on test.user0 to be empty @ config
Root Cause
The test creates its own ShardingTest without setting orphanCleanupDelaySecs. This parameter defaults to 3600 seconds in production. The moveChunk in each test iteration creates a CleanWhenEnum::kDelayed range deletion task on the config shard. The range deleter will not process it for 3600 seconds. When st.stop() calls checkOrphansAreDeleted, the assert.soon check (90-second default timeout) times out waiting for config.rangeDeletions to empty.
Note: resmoke's standalone.py sets orphanCleanupDelaySecs to 1 for fixtures it manages directly, but multiversion_auth tests use nodb: "" and create their own ShardingTest instances, bypassing that injection.
Fix
Add orphanCleanupDelaySecs: 0 to the ShardingTest setParam, consistent with:
- jstests/noPassthrough/catalog/secondary_read_killed_by_range_deletion.js
- jstests/sharding/transition_to_dedicated_config_server.js
References
- BF-43933
SERVER-127574(introduced the test)
- is related to
-
SERVER-127574 Write setFCV upgrade/downgrade test with pinned resharding version
-
- Closed
-