KVM Host Raid Consistency Checking

The Problem

Eventually a disk will fail, but in some cases a sector on a disk just goes bad without being used, and noone knows...

Consider the following.

• A sector on disk A goes bad. The sector is not used yet.
• A sector on disk B goed bad. The sector on disk B is used.
• Raid software will take disk B out of the array.
• A new disk is added to replace disk B.
• Resyncing from disk A to disk B is started.
• Resyncing reaches bad sector on disk A,
• Resyncing is now impossible.

The solution

Run regular raid consistency checks. This will detect the bad sectors before they become a problem. I run this on tuesday night. Edit your crontab.

crontab -e

Add the following to your crontab

0 4 * * 2 echo "check" > /sys/block/md0/md/sync_action
0 4 * * 2 echo "check" > /sys/block/md1/md/sync_action