First we delete bucket in config and then we wait for bucket events until we see bucket is deleted on all nodes. If bucket event(s) fire before we subscribe we'll timeout.
I think new implementation should use active polling for pre-2.0 compat mode and new janitor_agent call for 2.0+ mode.
|For Gerrit Dashboard: &For+MB-6203=message:MB-6203|
|28869,3||MB-6203 Fix race in waiting for bucket deletion on nodes.||ns_server||Status: MERGED||+2||+1|