Buffering related backports

[arthurschreiber]

Description

This backports various fixes around buffering that I think are important to have in our 18.x branch.

First, this backports the changes from vitessio#14968, which is a test only change but reduces the numbers of conflicts that needed to be resolved in some of the other backported changes.

The next change this backports is vitessio#15701. This fixes interaction between buffering and MoveTables operations. As we want to run some MoveTables operations in the very near future, I think it's important we backport this change to not run into any issues when we perform traffic switching.

Then I also included vitessio#15347. This change itself isn't really related to buffering, but reduces the conflicts for backporting and is a simple performance improvement / reduces load on etcd, so seems like a good change to have in 18.x.

The changes from vitessio#16655 fix a race condition where buffering would not kick in right after starting a vtgate process. We've not encountered this so far, but with more focus on safe automation, we should pull in this change as well to ensure we don't see sporadic errors when vtgate restarts happen at the same time as primary failovers.

Finally, this pulls in the set of commits from vitessio#17013. This changeset is going to fix the buffering issues we've seen where performing multiple failovers via PlannedReparentShard on different shards in the same keyspace could lead to a premature buffering stop and a spike of errors.

Related Issue(s)

Checklist

• "Backport to:" labels have been added if this change should be back-ported
• Tests were added or are not required
• Documentation was added or is not required

Deployment Notes