Hello William, Only 1GB RAM on a 64-bit platform?
We run essentially the same configuration you do, but we have clusters of 6, 16, 8, and 8 nodes respectively, and all of the nodes have 8GB RAM. They also have 8 hardware threads each. So each daemon essentially gets its own 1GB in-core heap and hardware thread. I think too much with too little is right... - Andy --- On Wed, 7/23/08, William Clay Moody <[EMAIL PROTECTED]> wrote: > From: William Clay Moody <[EMAIL PROTECTED]> > Subject: Re: HRegionservers aborting after rolling logs > To: [email protected] > Date: Wednesday, July 23, 2008, 5:43 PM [...] > My 5 regionservers are also the 5 TaskTrackers and > Datanodes for the cluster. A separate node is the NameNode, > HMaster and JobTracker. [...] > The machines that make up the cluster are: AMD Athlon64 3000+ > 2GHZ, with 1GB PC3200 DDR running Ubuntu Server 64-bit. Hadoop > 0.16.4 and HBase 0.1.3. [...] > I have a feeling I might be trying to do too much with too > little, especially considering data input jobs create very > hot regions due to the nature of the data.
