Re: Issue with marking replicas down at startup

[Mark Miller]

Yes, you are correct. It doesn’t really work. Depending on the distributed
mode you are running in, it may still publish the cores as down, in one of
the modes it sends a down node cmd to the Overseer which should do it based
on what cores are in the cluster state. In that case it should still
publish as down, but in both cases it doesn’t wait for the down state state
anyway, so you can see active replicas before they are active. Before the
two modes it always published, but it never waited.