Yes, detaching lb member from loadbalancer before corresponding node is deleted is the correct logic. The problem is in some cases, we don't know which node(s) will be deleted before the action is executed, e.g. resize cluster with random victim choosing. So I guess we may need to eliminate those random deleting logic before fix this problem. Will think about it.
Yes, detaching lb member from loadbalancer before corresponding node is deleted is the correct logic. The problem is in some cases, we don't know which node(s) will be deleted before the action is executed, e.g. resize cluster with random victim choosing. So I guess we may need to eliminate those random deleting logic before fix this problem. Will think about it.