There are 5 cluster(A, B, C, D, E), and I also build the xdr from A to others. The data size which need to be synced from A to others is about 600G (about 600M records).Then I find a lots of tcp connection in cluster A. So how to control the connection number and increase packet size to avoid a lots of small packet when rsyncing lots of data from A to others.
Due to each cluster has two nodes(A, B) and replication-factor is 2, is there any way to uniform a ip address for each cluster, Because if our client just connect node A, if node A failed, it seems our clients can’t get data.
It is maximum number of connection XDR will open with remote DCs.
Client connection works as -
With provided node ip (seed node ip), say node A, it tries to connect to that node.
On successful connection, it understands cluster view of that node. By communicating with other nodes in that view, client discovers entire cluster and forms its cluster view. Establish connection with all nodes of cluster.
Later if node A fails, it does not affect client. With connections with other nodes, client will update its cluster view.