[SOLR-16473] Race condition in shard split when a sub-shard is put in recovery state

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 9.2
• Assignee: Bruno Roustant
• Reporter: Bruno Roustant
• Created: 19/Oct/22 09:42
• Updated: 27/Mar/23 18:04
• Resolved: 06/Dec/22 13:17

Description

There is a race condition in shard splits code when a sub-shard is put in RECOVERY state and the remaining replicas are added. The problem is that the code does not wait to ensure the sub-shard state is indeed changed to RECOVERY.

andyvuong reproduced the bug by adding some artificial delay, and this race condition seems to be the root cause of indexing holes.

The fix is to wait for the RECOVERY state.

Issue Links

• links to: GitHub Pull Request #1088