[torqueusers] osc mpiexec and torque4

Brock Palen brockp at umich.edu
Wed Jul 25 12:15:29 MDT 2012


I actually tried with PBS_SERVER defined and not defined.  In both cases the same problem.

I modified mpiexec to change

fd = pbs_connect("our server host") ;

And this fixes the problem for our immediate needs (matlab+mvapich+tm support),

Obviously this is not acceptable as a long term solution.
I will look at filing a bug with adaptive about this behavior change.

Brock Palen
www.umich.edu/~brockp
CAEN Advanced Computing
brockp at umich.edu
(734)936-1985



On Jul 25, 2012, at 1:19 PM, Lloyd Brown wrote:

> I've heard through the grapevine that there's a torque release with a
> number of bugfixes, due out at the end of July.  Maybe if you hurry, you
> can get this in.
> 
> My guess is that, with less than a week 'till then, that'll be hard, but
> you can try, right?
> 
> Lloyd Brown
> Systems Administrator
> Fulton Supercomputing Lab
> Brigham Young University
> http://marylou.byu.edu
> 
> On 07/25/2012 11:01 AM, Michael Jennings wrote:
>> On Wednesday, 25 July 2012, at 12:48:54 (-0400),
>> Doug Johnson wrote:
>> 
>>> Has pbs_connect changed in torque 4?
>> 
>> It has, indeed.  :-)
>> 
>>> From the man page,
>>> 
>>>       If the parameter, server, is  either  the  null  string  or  a  null
>>>       pointer,  a  connection  will  be opened to the default server.  The
>>>       default server is defined by (a)  the  setting  of  the  environment
>>>       variable PBS_DEFAULT which contains a destination, or (b) the desti-
>>>       nation    in    the    batch    administrator    established    file
>>>       {PBS_DIR}/default_destn.
>> 
>> Ah, okay, sounds like mpiexec is relying on defined behavior that
>> TORQUE is no longer properly handling.  I wrongly assumed passing NULL
>> was an error; sorry!  :-]
>> 
>> Comparing the code from the 2.5 branch and the 4.x branch, it looks
>> like someone was trying to clean up/consolidate some code and
>> overlooked the NULL/empty string case.
>> 
>> The change I recommended in my previous e-mail should handle this just
>> fine.  Brock, you may want to open a bug for this against TORQUE 4.
>> 
>> Michael
>> 
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers



More information about the torqueusers mailing list