cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Edward Capriolo <edlinuxg...@gmail.com>
Subject Re: balancing load
Date Sun, 16 Jan 2011 20:27:07 GMT
On Sun, Jan 16, 2011 at 11:45 AM, Karl Hiramoto <karl@hiramoto.org> wrote:
> Hi,
>
> I have a keyspace with  Replication Factor: 2
> and it seems though that most of my data goes to one node.
>
>
> What am I missing to have Cassandra balance more evenly?
>
> ./nodetool  -h host1 ring
> Address         Status State   Load            Owns
> Token
>
> 82740373310283352874863875878673027619
> 10.1.4.14     Up     Normal  17.45 GB        77.48%
> 44427918469925720421829352515848570517
> 10.1.4.12     Up     Normal  8.1 GB          8.12%
> 58247356085106932369828800153350419939
> 10.1.4.13     Up     Normal  49.51 KB        1.66%
> 61078635599166706937511052402724559481
> 10.1.4.15     Up     Normal  54.48 KB        6.37%
> 71909504454725029906187464140698793550
> 10.1.4.10     Up     Normal  44.38 KB        6.37%
> 82740373310283352874863875878673027619
>
>
> I use phpcasa as a client and it should randomly choose a host to
> connect to.
>
> --
> Karl
>

For a 5 node cluster your initial Tokens should be:

tokens=5 ant -DclassToRun=hpcas.c01.InitialTokens run
run:
     [java] 0
     [java] 34028236692093846346337460743176821145
     [java] 68056473384187692692674921486353642290
     [java] 102084710076281539039012382229530463435
     [java] 136112946768375385385349842972707284580

To see how these numbers were calculated :
http://wiki.apache.org/cassandra/Operations#Token_selection

Use nodetool move and nodetool cleanup to correct the imbalance of your cluster.

Mime
View raw message