[lustre-discuss] Odd behavior with Matlab.

[Patrick Farrell]

Agreed, strongly - The root cause is known, and it's a patch which was 
added in Lustre 1.8.2.  There's no reason to think any current or recent 
version does not have this problem.

All the info I'm referencing is in LU-4185.
https://jira.hpdd.intel.com/browse/LU-4185

- Patrick
On 06/02/2016 10:54 AM, Laifer, Roland (SCC) wrote:
> Phill,
>
> I fear you are searching in the wrong direction. As I said before the
> return code of Lustre for mkdir on an existing directory is pretty
> arbitrary. (As Patrick pointed out this might change if you do a stat
> before.) We had seen that the Matlab problem disappeared for days and
> weeks and afterwards it appeared again quite frequently. This happened
> while there was no change in the installed Lustre versions.
>
> Regards,
>    Roland
>
>
> Am 02.06.2016 um 17:39 schrieb Phill Harvey-Smith:
>> On 02/06/2016 14:32, E.S. Rosenberg wrote:
>>> What is the output of:
>>> cat /sys/fs/lustre/version
>> Ok /proc/fs/lustre/version :
>>
>> On the 16.04 system which barfs on matlab :
>>
>> lustre: 2.8.53_51_g3680fa1
>> kernel: patchless_client
>> build:  2.8.53_51_g3680fa1
>>
>> On the 12.04 system where matlab works :
>>
>> lustre: 2.4.53
>> kernel: patchless_client
>> build:  v2_4_52_0-g55605c6-CHANGED-3.8.0-27-generic
>>
>> Cheers.
>>
>> Phill.
>>
>>
>> _______________________________________________
>> lustre-discuss mailing list
>> lustre-discuss at lists.lustre.org
>> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
> _______________________________________________
> lustre-discuss mailing list
> lustre-discuss at lists.lustre.org
> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org