Details
-
Bug
-
Resolution: Won't Fix
-
Major
-
Cheshire-Cat
-
1
Description
When GSI indexes are being restore, index service will decide optimal placement for indexes being restored. This step of finding optimal placement for indexes can fail if the cluster topology changes.
i.e If an indexer node is being failed over or being removed from the cluster, while restore is happening, the two operations (restore and topology change) can race with each other and restore can fail.
As per the documentation:
https://docs.couchbase.com/server/current/backup-restore/cbbackupmgr-restore.html#description
The cbbackupmgr is expected to work during failover/rebalance. But I think this statement was intended to be true only for data service. For indexing service, this doesn't hold true.
So, there is a need to change the documentation.
Please note that the GSI restore was never intended to work concurrently with failover/rebalance.
Also note that subsequent attempt restore should be successful as long as it is not affected by any other concurrent topology change. The GSI restore operation is idempotent.
Attachments
Issue Links
- relates to
-
MB-43276 Restore using rest api is failing during the rebalance
- Open