Details
-
Bug
-
Resolution: Unresolved
-
Major
-
4.1.0
-
Security Level: Public
-
None
-
Untriaged
-
Ubuntu 64-bit
-
Unknown
Description
Steps to reproduce:
- Create a 4 node cluster (I used Vagrants)
- Delete the default bucket and add a new memcached bucket
- Start load (50/50 - R/W) on the cluster (log the cluster maps)
- From the mgmt console click Remove (I did the last node)
- From the mgmt console click Rebalance
- Once completed, compare the cluster maps that are returned. Note that during the rebalance, the count of Nodes and NodesExt are unequal, but recover at the end.
In my files uploaded, the revision number is in sequence and you can see that rev 1069 has 4 nodes and 4 nodesext, that rev 1075 has 3 nodes and 4 nodesext (10.141.110.104 has been removed) and that rev 1080 has 3 nodes and 3 nodesext (with 10.141.110.104 no longer there as expected).
Questions/Comments:
- Is this expected behavior?
- If so, it makes it very difficult for the clients to know which array to depend on.