Description
Follow-up to SOLR-10704, a similar scenario occurs in MoveReplicaCmd when replication factor is 1 - the only copy of the source replica may be deleted while the target replica is still recovering.
Also, even when replicationFactor > 1 but the only remaining replicas are of the PULL type then leader election won't be able to find any replica to become a leader for this shard, which will result in effective data loss for that shard.
Attachments
Issue Links
- is broken by
-
SOLR-10914 RecoveryStrategy's sendPrepRecoveryCmd can get stuck for 5 minutes if leader is unloaded
- Resolved
-
SOLR-9440 ZkStateReader on a client can cache collection state and never refresh it
- Closed