This code leads to weird issues in production when slots are rebalanced: https://github.com/antirez/redis/blob/e6b0e8d9ec4561a07864358af8d2d4e81ac413fc/src/cluster.c#L1592-L1606 I reduced cluster size. When excess master become empty, slaves were automatically reassinged, and fork happends on "new" master. That leads to timeouts and 500 http requests in our application. It is really weird "feature", and I do not understand why it exists.
Comment From: funny-falcon
It hit me again. Now it hit me more painfully.
Again at cluster size reduction it caused timeout when my tool tried to send CLUSTER SETSLOT command, and cluster configuration become inconsistent. It leads to 15 minutes of unavailability of our production.
Comment From: funny-falcon
Look like it could be count as duplicate for #4592
Comment From: funny-falcon
Please, reconsider to fix this issue. I've updated PR #5285
Comment From: yossigo
@funny-falcon thank you for raising this, let's continue the discussion on the newer PR you have opened.
Duplicate of #5285