Once you have the cluster up and running, you should perform a trial failover using
PlannedReparentShard to make sure it works as expected.
A typical use case for
PlannedReparentShard is to use it during software updates. The command has a convenient
avoid_master flag that allows you to specify the current vttablet you are going to perform maintenance on. If that is a primary, then it performs a failover to another eligible replica. Otherwise, it is a no-op.
You can also perform a “Planned Reparent” through the browser from the
vtorcis also deployed, it will detect this and failover to another eligible replica as the primary.
vtorcwill not intervene during a graceful shut down.
If you are running
vtorc, you can perform the same function from the dashboard by doing a drag-and-drop operation of the nodes.
For more information, please refer to the Reparenting section.