On Fri, 28 Jul 2006 11:04:22 -0400 "Justin S. Leitgeb" <[EMAIL PROTECTED]> wrote:
> Hi all, > > I looked at fyodor a little more closely today when it was doing an rsnapshot > backup. I know that we're getting off of this machine soon, but it has come > to a point where performance is really bad during the backup interval > (moinmoin pages don't load, and even typing "w" take a good 10 seconds to > return). Yes, it's pretty bad during backup. > I started looking at the problem by checking vmstat 1. This showed up to 18 > blocked processes during one snapshot. Then, I looked at iostat to see where > the highest activity was. Tons of reads and writes to /dev/hda9 and > /dev/hdb9, which are the disk partitions that rsnapshot points to, through > the RAID 1 configuration. I am presuming that this is the disk activity that > is blocking other processes. Yes. The crucial problem here is that both disks are on the same controller, as installed by InterServer staff. > If we are going to be on fyodor for even a few weeks, I would like to try and > resolve this if it is easy enough to do. One thing that I was thinking of > trying that may lessen the side-effects of this backup processes is to > somehow get it to not write through the software RAID setup. Basically, we > would have to get an area of space that is sufficient for our backups, then > point rsnapshot to this partition. There's no disk space we could use in this way without significant trouble. (One which is, imho, not worth it at the moment -- shortly before switching to a new setup anyway). Nice day everyone! -doc _______________________________________________ HCoop-SysAdmin mailing list [EMAIL PROTECTED] http://hcoop.net/cgi-bin/mailman/listinfo/hcoop-sysadmin
