You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@cassandra.apache.org by Augusto Uehara <gu...@gmail.com> on 2014/02/03 13:39:29 UTC

Balance Virtual Nodes

Dear Cassandra list,

We are experiencing the following distibution of our tokens:

State    Address                 Load             Tokens   Owns         Host ID                                                            Rack
UN       10.180.199.188     108.72 GB     263         0.0%         f2c84a20-a4b7-4db1-b5bd-95eaf231ac76     RAC1
UN       10.180.206.15         71.86 GB     254         25.0%       a13afce3-7e98-4a3b-98d1-5bf6742435a4     RAC1
UN       10.45.163.69         152.23 GB     258         25.0%       90247170-e2bc-43ae-a827-2dcb082e43fc    RAC1
UN       10.235.36.125       103.91 GB     248         50.0%       89f362f5-e844-470d-a303-f54e46abc1cb      RAC1

our num_tokens is set to 256 and this node started as a regular node and later this was migrated to a virtual node.

We are using the org.apache.cassandra.dht.Murmur3Partitioner.

I've already tried the move and the shuffle nodetool commands after reading a few other posts about this somewhere else, but with no luck.

Can anyone please advice? At this point, our servers don't have enough data that I wouldn't consider creating another cluster, swap the apps to the new cluster and migrate the data using snapshots or scripts, but I think this is a dumb way to do this...


Kind regards,
Augusto