[DRBD-user] interaction with md and rebuilds

James Harper james.harper at bendigoit.com.au
Sun May 5 02:14:45 CEST 2013

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


I've had a 2TB disk fail, which is partitioned like:

part1 - GPT bios boot
part2 - md0 on / - RAID1 sda2/sdb2
part3 - md1 for drbd - RAID0 sda2/sdb2

The disk has been replaced, I rebuilt md0 first, and drbd is rebuilding now and is 85% complete. What's happened this morning though is that being the first Sunday of the month, Debian has decided to do a check of md0, which is hurting performance even more than any one of those operations on its own. Backups are running too so the whole thing is working pretty hard.

What would be the correct approach to prevent this situation from arising again? md by itself is smart enough to know that if (say) a RAID1 md0 and a RAID1 md1 both share a physical disk it should run any rebuilds sequentially not concurrently.

It would be easy enough to put a check in the scheduled cron check to see if drbd is running, but it would be nicer if it happened in the driver and md and drbd could coordinate their checks, putting one on hold if another was in progress.

James



More information about the drbd-user mailing list