[SOLR-16783] Race condition deleting clusterstate.json can prevent Solr from starting

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 9.0, 9.1, 9.2, 9.1.1, 9.2.1
• Fix Version/s: main (10.0), 9.3, 9.2.2
• Assignee: Tomas Eduardo Fernandez Lobbe
• Reporter: Tomas Eduardo Fernandez Lobbe
• Created: 02/May/23 18:30
• Updated: 21/Jul/23 20:42
• Resolved: 04/May/23 17:10

Description

ZkController attempts to delete the deprecated clusterstate.json file from ZooKeeper if empty on startup. If multiple instances attempt to delete it at the same time one of them will succeed and the others will fail with `NoNodeException`, which bubbles up and causes the instance to fail startup, requiring another restart

Issue Links

• links to: GitHub Pull Request #1624