Answers inline. From: Lu, Boying [mailto:boying...@emc.com] Sent: Wednesday, April 30, 2014 12:20 AM To: user@cassandra.apache.org Subject: Some questions to adding a new datacenter into cassandra cluster.
Hi, All, I need to add a Cassandra (running on a three nodes) fresh installed in a datacenter say DC3 to existing Cassandra datacenters say DC1 and DC2. I read the document about it at http://www.datastax.com/documentation/cassandra/2.0/cassandra/operations/ops_add_dc_to_cluster_t.html But I have several questions: 1. Does the operations in the step 7 (last step) only run on the new datacenter i.e. DC3? Do I need to do something on DC1 and DC2? 7a - altering the keyspace, is not DC specific. 7b - the rebuild - needs to be run only in DC3. When we did this in 1.2, I believe the documentation wasn't this clear, and we did not specify the existing data center with the rebuild command, and as mentioned the new nodes did not contain any data. We had to run a repair after that which ended up moving the data over. 2. If we use LOCAL_QUORUM (for read) and EACH_QUORUM (for writing) and the keyspaces/column families in DC1 and DC2 needs to propagate to DC3, do I need to restart Cassandra services on existing DCs i.e. DC1 and DC2? No 3. Since we don't change the seeds configuration in the yaml files of DC1 and DC2, how do DC1 and DC2 know the nodes in the DC3 if they reboot for some reason later? The purpose of the seed nodes are for newly joining nodes to know which nodes to initially contact to be able to join the ring. Once the new nodes in DC3 join the ring, all the nodes will gossip and know about all the other nodes. Thanks a lot. Boying
