I have a cassandra cluster which is 21 nodes large and the seed node is one of the node whicg is seeing high disk usage, high memory usage and also services lot of connection. Currently the num_tokens=256 and across all node.
My question is since seed nodes have the responsibility of processing the gossip info, do that need to take lesser data than the other nodes. Eg: keep num_tokens=64 for seed nodes and num_tokens=256 for other nodes. I could not find any information related to this. Does anyone have information.