Rolling Restart in a Cluster

You can do a rolling restart of your cluster, where you keep your cluster up and running throughout the process via HA. To avoid loss of data, actions like rolling restart or removal of a leaf require more than just stopping and restarting the nodes. The following commands should be used.

To shut down a node:

Note: The sdb-admin stop-node command will display this query:

Continue to stop the node(s) without taking a database snapshot?

Enter 'n' at the prompt.

To restart a node:

Restart the nodes in any order and combination, ensuring all the time, you have enough nodes up to keep the cluster running. This implies only shutting down nodes in one availability group at a time, and keeping enough availability groups up to serve the workload.

If you frequently need to stop all nodes on one host it is advisable to use the load-balanced mode of HA.

In case you do not plan to attach a leaf back again, then the REBALANCE PARTITIONS command should be used.