Anton, I have also seen this issue with decommissioned nodes remaining in the system.peers table.
On the bright side, they can be safely removed from the system.peers table without issue. You will have to check every node in the cluster since this is a local setting per node. Jeff On 6 July 2015 at 22:45, nowarry <nowa...@gmail.com> wrote: > Hey guys, > > I'm using Ruby driver( http://datastax.github.io/ruby-driver/ ) for > backup scripts. I tried to discover all peers and got wrong peers that are > different with nodetool status. > > ================= > Status=Up/Down > |/ State=Normal/Leaving/Joining/Moving > -- Address Load Tokens Owns Host ID > Rack > UN 10.40.231.53 1.18 TB 256 ? > b2d877d7-f031-4190-8569-976bb0ce034f RACK01 > UN 10.40.231.11 1.24 TB 256 ? > e15cda1c-65cc-40cb-b85c-c4bd665d02d7 RACK01 > > cqlsh> use system; > cqlsh:system> select peer from system.peers; > > peer > -------------- > 10.40.231.31 > 10.40.231.53 > > (2 rows) > > What to do with these old peers, whether they can be removed without > consequences since they are not in production cluster? And how to keep up > to date the peers? > > -------------------------------------- > Anton Koshevoy > >