there have been many reports of this. supposedly andrea arcangeli has patches
which fix this. i suggest you look in the archives, cause there was a patch that
fixes this for 2.2.2 posted last week or the week before.
it is a buffering problem with the newer kernels. you could make your arrays
under a 2.0.36 patched kernel then upgrade, this is what i had to do.
al
> Hi raidpeople. I have a redhat 5.2 machine with all updates applied
> (regular updates and kernel-2.2 update stuff).
> I have been having problems using mkraid to make a raid5 array.
> I installed the 2.2.3 kernel and applied the latest patch.
> I also built and installed an rpm for raidtools 19990309.
> I have 4 scsi disks on a scsi bus connected to a symbios ncr53c8xx
> controller, each disk is an identical 4.3GB quantum viking2 or some such
> drive.
> The machine is a dual p2-450 w/ 256MB ram
> The first drive (sda) is dedicated to system stuff, the other three
> will be the raid array.
> I made an /etc/raidtab that looks something like this:
> raiddev /dev/md0
> raid-level 5
> nr-raid-disks 3
> nr-spare-disks 0
> chunk-size 128
>
> device /dev/sdb1
> raid-disk 0
> device /dev/sdc1
> raid-disk 1
> device /dev/sdd1
> raid-disk 2
>
> I then run mkraid and it aborts. I have also tried many options with mkraid
> such as --force, -c /etc/raidtab, etc.
>
> The raid5 kernel module is loaded and registered as shown in /proc/mdstat
> The drives are formatted as one large partition each, but that should be
> wiped out when I create and format the raid device if I am correct.
> The /dev/md* devices do exist and the drives seem to be working fine.
> But when I run mkraid, mkraid /dev/md0, and other variations, it starts to
> go and then aborts with no explination.
>
> Another thing I wanted to ask which altho in the FAQ and other docs, seems
> to be under contention (different docs say different things, also, the web
> raid howto is not up to date with the howto distributed with raidtools):
> chunk size/stride size.
> For general purpose file serving of mostly files under 100k (which are
> reasonably quick anyway) and doing database stuff,
> What chunk size should I use?
> The docs said 128 is a reasonable value for raid5.
> I am planning on making the fs with a 4k block size, then I should use
> 16 as the -R stride option to mke2fs, correct?
> Just checking, any input is appreciated.
> I\'m mainly concerned with just figuring out how to get the mkraid to work,
> the rest I can fiddle with.
>
> Thanks,
> Chris
>
>
>
