Cluster Setup Suggestions

[Chad Russell]

I am working on a project to move an existing MySQL database to a MySQL cluster setup.

Here is what I have:

1 Management node (will be two when moved into production)

2 quad processor Xeon with 8GB ram each for the SQL/Data nodes.

Within this I am running three data node processes on each box to get around the ~2GB process limit and to make use of all the processors. My database was about 2.2GB on disk while it was a complete MyISAM only database. Now that all tables are in the cluster it appears it is using about 3.6GB of ram in the cluster (data and index) after issuing the "all dump 1000" command.

However on both the SQL/Data node machines I am seeing a lot of ram being cached on the server and the total available ram is hanging around in the 17MB free range.

This seems to cause the performance of this cluster to perform very badly. In fact worse than it did as a MyISAM only database with only 1GB of ram.

Here is an output from running the "free" command:
total used free shared buffers cached
Mem: 8312720 8295312 17408 0 17900 4215420
-/+ buffers/cache: 4061992 4250728
Swap: 4096564 0 4096564


Are there any configuration params I am over looking here? Any suggestions?

Also Just one of the tables in the cluster has a little over 23 million rows. The second biggest table has about 3.5 million rows.

Any help would be greatly appreciated.

Thanks,
Chad



Edited 1 time(s). Last edit at 02/27/2006 11:39PM by Chad Russell.