I am trying to build a Apache cassandra(3.11.4) cluster with 10 nodes. I wanted to understand,what is difference between keeping all nodes as seeds nodes than keeping only few node as seeds.
I didn't find much input on Apache cassandra about seed nodes.
I have only one datacenter and in the future I may get the requirement to increase or decrease number of nodes in the cluster considering these factors, Please suggest me if how many seed nodes I need to keep for cluster.
You have to understand that seed nodes are nothing special except for you to expect one of them to always be reachable when a nodes comes up after being added to the cluster so they can propagate the cluster topology to the newly joining node.
So if you have a cluster of 10 nodes it's not necessary to have all 10 nodes in the seed node table, but rather you can only have 3 or 4 configured as seed nodes.
When a new node wants to join the cluster it will try to reach the seed nodes from it's cassandra.yaml to discover if there is a already running cluster. If one seed node is reachable it will retrieve the cluster topology from it and join the the cluster.
Subsequent restarts use the previous topology information so restarting/operating a cluster is not impacted even when all seed nodes are down.
You just want to have more than one seed-node up at all times, otherwise if all seed nodes are down you can't add new nodes to the cluster. So make sure to spread your seed nodes across availability zones/racks.
