0

When I run nodetool -host ip.ip.ip.ip ring on all of the nodes in my 3 node cluster, the output always shows only one of the nodes..

I ran a netstat and I am not sure how to interpret this. I reviewed all my config files and it looks fine. Do I need to make changes to the /etc/hosts file? I double checked my security group rules in the console and things look fine. 7000 is open inbound and outbound..

netstat -ant | grep 7000
tcp        0      0 10.160.13.61:7000           0.0.0.0:*                   LISTEN
tcp        0      0 10.160.13.61:48403          10.160.13.37:7000           ESTABLISHED
tcp        0      0 10.160.13.61:7000           10.160.13.62:60845          ESTABLISHED
tcp        0      0 10.160.13.61:7000           10.160.13.37:56271          ESTABLISHED
tcp        0      0 10.160.13.61:7000           10.160.13.37:60656          ESTABLISHED
tcp        0      0 10.160.13.61:38436          10.160.13.62:7000           ESTABLISHED
tcp        0      0 10.160.13.61:52762          10.160.13.62:7000           ESTABLISHED
user836087
  • 2,271
  • 8
  • 23
  • 33

1 Answers1

1

Most likely you don't have your seeds configured correctly in cassandra.yaml. Any new node should point to one or more existing nodes in the seeds config; this tells the new nodes how to join the ring. Keep in mind also that you should specify your tokens manually so you get even key distribution.

rs_atl
  • 8,935
  • 1
  • 23
  • 28
  • I can't wrap my head around key distributions and how to do it manually. I'm a total noob intern. can you explain? The web guide doesnt help. In the cassandra.yaml file on each of the 3 nodes, I have one seed of ip 10.160.13.37. – user836087 Dec 14 '12 at 20:51
  • Regarding token assignment, see [my answer to this question](http://stackoverflow.com/questions/13654286/how-do-you-fix-node-token-collision-issues-when-starting-up-cassandra-nodes-in-a/13726022#13726022). If you point a new node to an existing node (using the seed property), and the two nodes can talk to each other, the new node will join the cluster automatically. Feel free to post your yaml if you can't get it going. – rs_atl Dec 14 '12 at 21:16
  • You can get things into a weird state if your nodes aren't able to talk to each other on initial startup. If this has happened, you'll need to delete all the LocationInfo sstables in the data dir. – rs_atl Dec 14 '12 at 21:17