On Mon, Feb 27, 2012 at 9:48 PM, Antonio Modesto <[email protected]> wrote: > I didn't know about the memory limit. So if I have a 10GB database, Must > I have, at least, 10GB configured for each cluster node?
David posted a link which can help assign memory on data nodes. Short version is: - while mysql cluster can store some data on disk, you really should put it all on memory. Otherwise there wouldn't be much speed improvement - only a portion of memory on data node can be used as data memory - multiple data nodes can add the available data memory, but replicas trade memory for availability (thus reducing the number of memory for data) so if you have (for example) 4 datanodes, 16GB total memory each, and configure them as 2 node group @2 replicas, don't be surprised if total usable memory for data is around 10-20 GB. When your memory is limited and radacct is huge, better keep them separated: - put tables OTHER than radacct on cluster - put radacct on normal mysql instance - store "live" accouting data on cluster The last part is kinda hard, since you need to create your own queries. The default sql queries doesn't split "live" (e.g. records that don't have acct-stop yet) and "archive" data. -- Fajar - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
