Hi Robert, Your action plan looks good.
You can think of the *cassandra-topology.properties* file as a map for the cluster. The map between the nodes must be consistent because each node uses it to determine where it is meant to be located logically. It is good hygiene to maintain the *cassandra-topology.properties* so it contains only the IPs (broadcast addresses) currently used in the cluster. Technically, you cloud leave the entry for the decommissioned node in there. The problem is later on if that IP address is used by a node it will be placed back in DC1 and this could be the wrong logical placement for it. So I would advise removing the address if the node is inactive. For step 3, there is no need to restart all the nodes. However if you do want them to reload the configuration you will need to perform a rolling restart on the cluster (i.e. restart one node at a time). Regards, Anthony On Thu, 11 Apr 2019 at 03:38, rastrent <rastr...@protonmail.com.invalid> wrote: > Hi there, > > I am running a cassandra cluster (v3.0.9) with 2 DCs (4/3 nodes > respectively) using endpoint_snitch: PropertyFileSnitch and I would like to > decommission one node in DC1 but I wonder about what kind of actions I need > to take related with the the topology settings. > My cassandra-topology.properties has the those simple settings below: > > x.x.x.x=DC1:RAC1 > x.x.x.x=DC1:RAC1 > x.x.x.x=DC1:RAC1 > x.x.x.x=DC1:RAC1 > x.x.x.x=DC2:RAC1 > x.x.x.x=DC2:RAC1 > x.x.x.x=DC2:RAC1 > > default=DC1:r1 > > My action plan is to: > > 1)Decomission a node in DC1 > 2) After node leaves cluster, edit cassandra-topology.properties in every > node in the cluster > 3) Question: No I need to restart all nodes in cluster? (one each time of > course) > > Bonus question: Do I need to change the cassandra-topology.properties > before move/remove nodes? > > Cheers, > > Robert, > > > Sent with ProtonMail <https://protonmail.com> Secure Email. > >
