Overview of the Issue

If PRS fails during the initialisation of a shard, and the failure happens while promoting the primary while it is writing to the topo-server such that the write succeeds, but fails with a timeout, then the tablet won't change its internal display state to a primary tablet. VTOrc sees this failure and tries to fix this by calling UndoDemotePrimary, but that doesn't change the type of the tablet to PRIMARY. It only fixes the mysql level settings and this causes the cluster to not have a primary at all.

Reproduction Steps

1. Run PRS, and simulate a failure that happens before new primary tablet has promoted itself.

Binary Version

main

Operating System and Environment details

-

Log Fragments