Overview of the Issue

When cancelling a reshard workflow via the Reshard Cancel command, Vitess currently "cleans up" the target shards of the reshard workflow by deleting them from the topology.

This seems unintended, as it causes two problems:

1. The new shards (and their tablet records) in the topology are gone, but the tablets are left running. That's unexpected. If we want to keep this behaviour, it should be an optional flag and not enabled by default.
2. Because the new shards and their tablets get deleted from the topology, the workflow cleanup step doesn't actually clean up anything because at that point Vitess doesn't even know about the existance of the tablets anymore. This means the vttablets for the deleted shards just keep executing the workflow.

This seems to be happening here:

And here:

I don't think we should be dropping the target shards here. Instead, we should only drop the tables on the target shards, to restore them to the state they had before resharding was started.

Reproduction Steps

Create new shards and start a Reshard workflow. Cancel the Reshard workflow via Reshard Cancel command.

Notice that the shards and tablets have been deleted from the topology. Check the tablet status pages to see that the tablets are still running the reshard workflow.

Binary Version

v19 and later

Operating System and Environment details

N/A

Log Fragments

N/A