[SOLR-12433] Recovering flag of a replica is set equals to leader even it failed to receive update on recovering

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 7.4
• Assignee: Cao Manh Dat
• Reporter: Cao Manh Dat
• Created: 01/Jun/18 01:56
• Updated: 08/Jun/19 15:11
• Resolved: 01/Jun/18 02:24

Description

When digging into RestartWhileUpdatingTest failure, I see that term of replicas is kinda mess up. 

[junit4] 1> /collections/collection1/terms/shard1 (0)
[junit4] 1> DATA:
[junit4] 1>

Unknown macro: { [junit4] 1> "core_node24_recovering"}

By design, the core_node24_recovering and core_node24 should be always equals to each other. The reason here is ZkShardTerms.ensureTermsIsHigher also increase the core_node24_recovering is a higher number. This will lead to a case when a replica finished recovering but it won't be able to become active.