On 08/29/2012 12:06 PM, Brian Foster wrote:
> On 08/29/2012 10:26 AM, Brian Candler wrote:
>> On Wed, Aug 29, 2012 at 08:47:22AM -0400, Brian Foster wrote:
> ...
>>
>> Running a couple of concurrent instances of
>>
>> while [ 1 ]; do bonnie++ -d /mnt/point -s 16384k -n 98:800k:500k:1000; done
>>
>> was enough to make it fall over for me, when the underlying filesystem was
>> XFS, but not with ext4 or btrfs. This was on a system with 24 disks: 16 on
>> an LSI 2116 controller and 8 on an LSI 2008. It was MD RAID0:
>>
>> mdadm --create /dev/md/scratch -n 24 -c 1024 -l raid0 /dev/sd{b..y}
>> mkfs.xfs -n size=16384 /dev/md/scratch
>> mount -o inode64 /dev/md/scratch /mnt/point
>>
>
> Thanks. I didn't see an obvious way to pass through physical disks in
> the interface I have, but I set up a hardware raid0 and a couple
> instances of bonnie. This may not be close enough to your workload, but
> can't hurt to try.
>
To follow up on this, I ran this workload for a couple days without a
problem. I was able to configure a bunch of single disk raid0 volumes to
put into an md raid0, so I'm testing that next.
If you do happen to reproduce the problem again, I would reiterate the
suggestion to append that blocked task data to the thread over on the
xfs list (re: my last post, it looks like some data was missing..?), as
we might get more conclusive analysis on the state of the filesystem at
the point of the hang.
Brian
