Clients fail if first node in node list has been removed from cluster
I'm seeing a problem with both the Perl and Ruby clients, against Couchbase 2.0.0.
If I remove one node from the cluster via the GUI (Remove, then Rebalance), then all clients who had listed that node as the first in their node list will now fail to connect.
Note that clients which *are already connected* are fine. The problem is when a new worker process fires up and tries to connect. I am guessing it hits the removed node, who accepts the connection on port 8091 but reports that it isn't responsible for any buckets in that cluster. It sounds like the clients should really be reading that error as a sign to try another host in the cluster, rather than a total failure.
The Perl client reports "Temporary failure. Try again later".
The Ruby client reports:
/var/lib/gems/1.9.1/gems/couchbase-1.2.2/lib/couchbase.rb:59:in `initialize': HTTP/1.1 404 Obj (Couchbase::Error::BucketNotFound)