I just set up a largish (~100 GB) RAID1 array under Linux 2.6.0. Now,
/proc/mdstat is happily telling me that the resync will be completed
in 3700 minutes. This seems terribly slow to me. At first, it
wouldn't work at all, complaining about "bio too big", so I changed
RESYNC_BLOCK_SIZE to 32k.
Is there a way to mark the superblocks as up to date manually? I
don't really care what a read will return for parts of the array that
I haven't written to yet.
--
M?ns Rullg?rd
[email protected]
[email protected] (M?ns Rullg?rd) writes:
> I just set up a largish (~100 GB) RAID1 array under Linux 2.6.0. Now,
> /proc/mdstat is happily telling me that the resync will be completed
> in 3700 minutes. This seems terribly slow to me. At first, it
> wouldn't work at all, complaining about "bio too big", so I changed
> RESYNC_BLOCK_SIZE to 32k.
I found it. It's a raid1 on top of raid0. Apparently the raid1
resync considered the I/O from the raid0 devices to the physical disks
as normal I/O and reduced the speed. I increased the min rate in
/proc/... and now it's resyncing at 40 MB/s. Much better.
--
M?ns Rullg?rd
[email protected]