Hi Jason, If you are familiar with ROCKS clusters, they also do this, so if you're using Linux, you can just grab the appropriate RPM.
I pulled the parts I need from the ganglia-hpc-4.3-2 RPM that comes with the ROCKS 4.3 distro and installed it on a vanilla RHEL4 system and it worked great. The tmpfs approach for Ganglia is also particularly effective, and on a cluster that size, definitely recommended. cheers, Klaus On 3/14/08 9:41 AM, "Martin Hicks" <[EMAIL PROTECTED]>did etch on stone tablets: > > sending to the list. > > On Fri, Mar 14, 2008 at 09:17:23AM -0700, Jason Venner wrote: >> We have been thinking about having our RRD's in a ramfs, and have the >> shutdown script copy the ramfs to a persistent store, and the startup >> script copy it back, and have a 1ce per 30 minute copy to persistent >> store to cover the inevitable crashes. >> >> We are building 100+ machine clusters (for hadoop) and the disk IO for >> the rrd files are trashing the seek queues on our disks. >> >> Has anyone implemented this or something similar recently? > > SGI has scripts to copy to disk before shutdown, and restore to tmpfs at > startup. We currently don't do periodic syncs to disk. > > mh > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Ganglia-general mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/ganglia-general ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Ganglia-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ganglia-general
