Has anybody implemented DRL for RAID 1?
After a crash, my RAID 1 tries to resync the entire mirror:
kris@valiant:~> cat /proc/mdstat
Personalities : [raid1]
read_ahead 1024 sectors
md0 : active raid1 sdc8[1] sda8[0]
12241408 blocks [2/2] [UU]
[==>..................] resync = 11.0% (1358400/12241408)
finish=124.3min speed=1457K/sec
unused devices: <none>
With a region bitmap and dirty region logging, the recovery time
would be much lower, since the RAID subsystem does not have to
resync all of the disk, but only those regions marked as dirty
during a crash.
Kristian