Known limitation : Join flags when restoring a backup onto new machines

From the Know Limitation page, Join flags section. It stats that “when starting all subsequent nodes, --join should be set to the address of node 1”. Is that only for restoring a backup? Or if that’s a best practice when we want to start a new node or join a node back to the cluster? If I am restarting node 1, should I make it join node 2? Or any node?


Hi Leo,

Apologies for that paragraph being a little confusing! When you’re starting up a cluster (not restoring just restoring from backup), it’s standard/best practice to have all nodes other than the first one join the address of the first one. If you’re restarting node 1, you don’t have to make it join another node, but it would be a good practice to make it join one of the other nodes. That way, if for some reason node 1’s data gets deleted and it restarts, it will rejoin the cluster properly.

The second paragraph of that known limitation only applies to the case of restoring from backup onto machines that have different IP addresses from the original cluster.


The join flag accepts multiple values (comma-separated), and I’d say the best practice is to always specify more than once. Every node should have the --join flag pointing to the first two or three nodes (including the first node, except for the first time it is run with no --join flag)