Details
-
Bug
-
Resolution: Unresolved
-
Major
-
None
-
None
-
None
-
None
-
0
Description
If an unhandled exception such as SocketException or IOException happens when attempting to create and initialize new nodes while processing a new BucketConfig in ClusterContext.ProcessClusterMapAsync, there is no retry mechanism.
Nothing will retry creating the one failed node, and nothing will retry processing of that particular BucketConfig.
Therefore, if a node failed to get created during a server scale up, it will not be attempted again until the next BucketConfig is requested to be processed.
In the case of 7.6, since the server does not re-send the body of config that it has already sent in response to a ClusterMapChangeNotification or NotMyVBucket/VBucketBelongsToAnotherServer, the client may be in an inconsistent state until the server generates a new config version.