Details
-
Bug
-
Resolution: Won't Fix
-
Critical
-
2.0-beta-2
-
Security Level: Public
-
None
-
2.0
Description
Consider, 2 cluster each containing 3 nodes each (c1-n1, c1-n2, c1-n3 and c2-n1, c2-n2, c2-n3).
Now we want to set up replication between these 2 clusters:
Create replication reference from c1 to say c2-n1.
Now, if we try creating another replication reference from c1, but this time to c2-n2, there's going to be an error message saying that c2-n2 already belongs to the cluster containing c2-n1, this is expected!
The work around:
Consider I have a 2 node cluster (c1-n1, c1-n2) and a 4 single node clusters (c2-n1, c3-n1, c4-n1, c5-n1).
Set up 4 replications from c1 to c2, c3, c4, c5.
This is now a star topology.
Once this is set up:
1> add node c2-n1 to c1 (Permitted)
2> add nodes c4-n1 and c5-n1 to c2-n1 (Permitted)
3> Rebalance the above 2 operations.
Once rebalance is complete, it is seen that c1 (which now has 3 nodes), has 4 replications, three of which are going to a one cluster (c2-n1, c4-n1, c5-n1) and one replication to itself.
Also seen are 4 outgoing replications on cluster1 showing different stats when load is added on it, although 3 of these replications are to the same cluster (now) and one to itself.
I think there should at least be a prompt asking to delete such a replication reference before adding single nodes (another cluster altogether) to a current cluster.
However, once the load completes, the item count is the same on the now chain linked 3:3 cluster set up.
c1-n1: 10.1.3.235
c1-n2: 10.1.3.236
c3-n1: 10.3.2.54
c2-n1: 10.1.3.237
c4-n1: 10.1.3.238
c5-n1: 10.3.2.55
Attached are the grabbed diags for the nodes involved in the above scenario.