[torquedev] [patch] Fix -W regression from changeset 2367

Ken Nielson knielson at adaptivecomputing.com
Thu Aug 5 09:38:39 MDT 2010


On 08/05/2010 12:34 AM, Eygene Ryabinkin wrote:
> Ken, good day.
>
> Wed, Aug 04, 2010 at 09:53:53AM -0600, Ken Nielson wrote:
>    
>> On 07/29/2010 01:58 PM, Eygene Ryabinkin wrote:
>>      
>>> Today I was messing with the errors that were coming from our Grid
>>> gatekeeper that whined that -W value for qsub is incorrect.  It turned
>>> out that something like "-W stagein=file1 at host1:src1,file2 at host2:src2"
>>> was used and qsub rejected such specification.
>>>
>>> Digging the sources I had found that Subversion's r2367 was the point
>>> where the regression was coined in.  I had created a quick patch that
>>> will eliminate this regression, the patch is attached.
>>>
>>> It works for me already for some hours and some hundreds of jobs, but
>>> more eyes looking at the patches is always a good thing.
>>>        
>> There was a fix for this put in a while ago in the 2.4 branch. In order
>> to get the syntax to work you need to put single quotes around the value
>> for stagin like the following:
>>
>> -W stagein=\'file1 at host1:src1,file2 at host2:src2\'
>>      
> Will it work for the older releases?  For something like 2.4.2?  I am
> asking, because we are using this functionality in the EGEE gLite
> middleware stack and some sites run rather old Torque, but some are
> trying to be on the cutting edge.  But submission scripts are meant to
> be universal.
>
> I understand that we can just use multiple '-W stagein ...' directives,
> one for each file, but still -- is the cited syntax supported for old
> releases?  I see no signs of it in manual pages, but they can be
> incomplete.
>
> Thanks!
>    
This broke in 2.4.6. So 2.4.2 would not have the problem.

Ken


More information about the torquedev mailing list