The scenario is this: I have a system with several external USB hard drives that are part of a RAID5 array. (It’s a work in progress, I’m simply out of SATA controllers right now.) The problem is that sometimes (about once per month) two or more of the drives will have USB bus errors simultaneously and go offline. This causes the RAID array to stop in its tracks.
The solution is simple, but it’s taken a bit of testing and research to make sure it works. So here goes:
mdadm --assemble --force /dev/md1
This causes mdadm to update the event count on the drives which simultaneously disconnected to match the drives that were fine all along. After a quick resync, everything is happy again.
Here’s where I found this tidbit: http://www.linux.org/docs/ldp/howto/Software-RAID-HOWTO-8.html