The basic flow is to upgrade each yb-master and yb-tserver one at a time, verifying after each step from the yb-master Admin UI that the cluster is healthy and the upgraded process is back online.
If you plan to script this in a loop, then a pause of about 60 secs is recommended before moving from one process/node to another.
Preserving data and cluster configuration across upgradesYour data/log/conf directories are generally stored in a separate location which stays the same across the upgrade so that the cluster data, its configuration settings are retained across the upgrade.
1. pkill yb-master 2. switch binaries to new release 3. start the yb-master process 4. verify in http://<any-yb-master>:7000/ that all masters are alive 5. pause ~60 secs before upgrading next yb-master
1. pkill yb-tserver 2. switch binaries to new release 3. start yb-tserver process 4. verify in http://<any-yb-master>:7000/tablet-servers to see if the new TServer is alive and heart beating 5. pause ~60 secs before upgrading next yb-tserver