The latest kernel which was released ended up with a "bug" in the RAID rebuild code, where it doesn't give proper priority to the process. RAID rebuilds now take about 2 days to complete (they do complete and everything is fine, it just runs a LONG time!)
A new kernel is in QA at Sun; until it's released, one of the engineers passed me a "hack" to fix the problem:
Shell into the server as root and do
echo 5000 > /proc/sys/dev/md/speed-limit
You may want to add a similar line to the end of /etc/rc.d/rc.local so each time the server reboots this command will be executed (remember to remove it when the new fixed kernel is released and installed).
Side effect is a slight increase in server load while the RAID is rebuilding, but I think it's a small price to pay for completion in a "normal" amount of time (2 hrs for my 2x40GB 4r).
I guess by putting an even bigger number in, it would speed up even more (to a point), and the server load would go up even more...
This info is also in the cobalt-users list archives -- if you're not on the list, I highly recommend it. There is a lot of good info that gets handed out there by a lot of very knowledgeable people. Go to
http://list.cobalt.com/mailman/listinfo/cobalt-users to sign up, or to search the archives.
Linear Mode
