[stgt] iSER data corruption

[Matthew Chan]

  Sure, I have two machines that haven't had production services 
deployed on it yet.

I'm still working on deploying a bit of other stuff and the academic 
year is about to start, so it may be a week or two before I can get 
around to testing the new code. I can push it earlier if you're in a 
rush to merge the code though.

I've never tested the data integrity of a link before, and the previous 
issue was only detected because it was serious enough to cause 
widespread filesystem corruption. Do you have any suggestions on how to 
test this more robustly? I'm thinking of doing a dd followed by a sha256sum.

Matt

On 08/26/2010 08:41 PM, FUJITA Tomonori wrote:
> On Sun, 22 Aug 2010 03:35:39 -0400
> Matthew Chan<talcite at gmail.com>  wrote:
>
>>    Hi,
>>
>> I'm experiencing what I think is data corruption issues when using the
>> iSER driver in CERN SLC5.5 (which is based directly off RHEL 5.5).
>>
>> I just set up stgt with the iSER driver on a mellanox infiniband mesh
>> with 7 nodes in it. I was having problems with my OCFS2 cluster crashing
>> randomly on large data writes, so I simplified it down to 1 initiator
>> and 1 target and an ext4 partition.
>>
>> The backing store is a 5 TB linux raid 6 softraid, running on SLC 5.5.
>> The initiators are running open-iscsi on Ubuntu Server 10.04. I'm using
>> the OFED packages from each respective distro.
>>
>> On my test setup with 1 target and 1 initiator, copying files with
>>
>> 'dd if=/dev/zero of=/<ext4 array>/zeroes bs=64k count=10000'
>>
>> generated a whole slew of ext4 errors on the initiator. A subsequent
>> fsck.ext4 showed thousands of inode errors. Trying to transfer a file
>> with cp generated similar errors.
>>
>> Are there any known quirks with the iSER driver, or am I misconfiguring
>> something? My infiniband connection seems quite stable, and I'm using
>> ipoib quite heavily right now.
> Some people have reported problems with the iSER driver.
>
> Alexander submitted the completely new implementation:
>
> http://lists.wpkg.org/pipermail/stgt/2010-July/003868.html
>
> Can you try it?
>
> git://git.kernel.org/pub/scm/linux/kernel/git/tomo/tgt.git iser
>
>
> I've not merged it yet but probably I'll do. It would be greatly
> appreciated if you could test the new driver.
>
>
> Thanks,
--
To unsubscribe from this list: send the line "unsubscribe stgt" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html