Details
-
Bug
-
Resolution: Fixed
-
Critical
-
6.6.6
-
0
Description
Currently we make 3 updates to bucketconfig during failover of a bucket as here
1. Update fast forwayd map
2. Update current vbucket map
3. remove failed nodes from server list
We should make them as a part of one transaction to reduce the number of updates to bucketconfig which in turn affects the replication of the config properties. The buckets key in ns_config is dependent on the number of bucket and the effect of replicating multiple updates to the same large key(when using large number of buckets) is wasteful.
The end result is that failover processing time should show improvement due to less config replications.