To test the network side of it, you should be able to run ttcp or use netcat dd'ing from /dev/zero on a source box to /dev/null on the destination box. That would eliminate disk IO as a potential bottleneck. If you get decent performance there, then the issue is somewhere else.
On 11/23/2009 02:13 PM, John P. Rouillard wrote: > > In message<[email protected]>, > "Richard \"Doc\" Kinne" writes: > >> Harmonia's specs: >> Dual AMD opteron 3.2Ghz duo core (4 processors) >> 6 x 2GB ECC DDR2 PC5300 RAM (12 GB) >> 8 x 500GB Seagate SATA300 32MB cache 7200RPM drives setup as RAID 5 (3.5 = >> TB) >> Running Fedora 10 > > Are all 8 disks on the same disk controller? > >> 1) I can't seem to copy anything with regard to this machine without its >> load average going through the roof. > > What does your i/o wait look like when this happens? > >> An scp, even a cp will drive the >> computer's load average to between 12 and 18. Copying a large file, or >> doing a mysqlhotcopy, will make the load average slowly climb, with some >> spikes up to that level. I can't think that's right. Not for something >> with 4 processors. >> >> 2) Whenever the load average goes above, say, 5 NFS starts seriously >> flaking out. It goes "bye-bye," stops responding, and starts >> disconnecting disks at the clients. After the load average goes down, it >> will come back. >> >> NFS exports is here: >> >> /BigBang *(rw,insecure,sync,nohide) >> /mnt/data *(rw,insecure_locks,nohide,insecure) >> >> /mnt/data is our RAID drive. /BigBang is an alias to /mnt/data/BigBang > > What raid controller are you using? > > Are you running software raid or hardware raid? If the former are the > os/mysql disks sharing disk controllers? > >> 3) Finally, network performance is completely in the toilet. Copying >> files via scp, or rsync, in addition to driving the load average up, >> seems to go at a rate of 2.5MB/s. I have two gigabit ethernet cards in >> this machine. These are wired to a gigabit switch. The switch is wired >> to the rest of the office via a 100Mbps hub, but the issue is we got far >> better transfer performance from the old server (Occam) than we are from >> the new, faster server (Harmonia) and I can't figure out why. > > What does your interrupt level look like, and are any of the raid > card, network cards sharing interrupts? > > -- > -- rouilj > John Rouillard > =========================================================================== > My employers don't acknowledge my existence much less my opinions. > > _______________________________________________ > bblisa mailing list > [email protected] > http://www.bblisa.org/mailman/listinfo/bblisa _______________________________________________ bblisa mailing list [email protected] http://www.bblisa.org/mailman/listinfo/bblisa
