When a server is failed over (manually/automatic), what are the step I have to take?
Is a rebalance required to make replicated data active on the rest of nodes? In case of no, how long does it take to the rest of the nodes to serve the data that recovered from the failed node?
Yes, a rebalance is required to get back to the replication factor for which you’ve configured the cluster. See the section in the docs covering failover for more information on this.
Thanks,
The question is more like will the data that was on the failed over node should be available from the other nodes before the rebalance?
@ohassidi yes, if you have replicas configured. The failover (see the docs that @ingenthr linked) promotes the replica partition to active instantly and the SDKs are aware of the change. The rebalance part then will take longer, but makes sure that you end up with the right amount of replicas again. Because, if you think about it, once you’ve failed over you are -1 replicas in total until the rebalance corrects it.