Software RAID resync overruns other partitions
OK,
So I have set up my software RAID like this. There are only 2 disks here so I created three partitions.
/dev/md0 - raid1 - /boot
/dev/md1 - raid0 - lvm - root fs and swap
/dev/md2 - raid1 - lvm - /home
My problem is that I copied over many gigabytes of data to /dev/md2. However, after completing the copy the 'device' went idle. So md started syncing the drives at the fastest possible rate. When I started trying to compile a program the system was completely unresponsive. Why? Because all my compiling activities where isolated to /dev/md1 - and /dev/md2 was 'idle,' so md kept resyncing as fast as the disks would go.
I've already implented a reasonable solution by setting the following via sysctl:
dev.raid.speed_limit_max=25000
However, why wouldn't there be a way to tell md that /dev/md0,1, and 2 are the same physical disks and that access to any one of those devices should be considered activity on all of them? I can't find it in all my searching so I was hoping someone else might have run into this.
Last edited by uselesslogin; 02-23-2011 at 11:46 PM.
|