More reasoning is that read access should not make a resource
unavailable, moreover your odds of consistently hitting the few
milliseconds wherein writing takes place are slight.
Unless you have changed the code, it is unlikely the program's fault, so
the next culprit looks like the file system not keeping up with closes,
or the batch system somehow meddling in i/o.
Bill
On 4/21/16 11:10 AM, Bill Ross wrote:
> > At line 810 of file runfiles.F90 (unit = 7, file = 'mdinfo')> [on
> mdinfo] Resource temporarily unavailable
>
> I suspect the program is not closing files successfully, probably this
> one as it repeatedly opens and reopens it, so it runs out of file
> descriptors. It would be nice if the prog could specify whether it was
> on open or write.
>
> Bill
>
>
> On 4/21/16 3:43 AM, Karolina Markowska wrote:
>> Hi,
>>
>> I'm using a simple command like:
>>
>> pmemd.cuda -O -i prod1.in -o prod1.out -p *.prmtop -c emd2.crd -x prod1.nc
>> -r prod1.crd
>>
>> I'm running my jobs on a cluster with 4GPUs and the queueing system I use
>> is Torque.
>>
>> And in the prod.in file looks like that:
>>
>> 50 ns accelerated MD (aMD) Langevin dynamics
>> &cntrl
>> imin=0,
>> irest=1,
>> ntx=5,
>> nstlim=25000000,
>> dt=0.002,
>> ntc=2,
>> ntf=2,
>> nscm=0,
>> ig=-1,
>> cut=8.0,
>> ntb=2,
>> ntp=1,
>> ntpr=1000,
>> ntwx=1000,
>> ntt=3,
>> gamma_ln=1.0,
>> temp0=300.0,
>> ioutfm=1,
>> iwrap=1,
>> iamd=3,
>> ethreshd=5545, alphad=239,
>> ethreshp=-174476, alphap=12195,
>> /
>>
>> Is it OK?
>>
>> Best regards.
>>
>>
>> 2016-04-21 12:13 GMT+02:00 Hannes Loeffler <Hannes.Loeffler.stfc.ac.uk>:
>>
>>> On Thu, 21 Apr 2016 11:01:03 +0200
>>> Karolina Markowska <markowska.kar.gmail.com> wrote:
>>>
>>>> At line 810 of file runfiles.F90 (unit = 7, file = 'mdinfo')
>>>> Fortran runtime error: Resource temporarily unavailable
>>> This error message tells you that the data has not been finished
>>> writing yet. So you obviously can't access it until it has. The pmemd
>>> code does no guard against errors here and terminates.
>>>
>>> The interesting question is now why the file is not ready yet. This
>>> may be related to the setup of your system. How are you running your
>>> jobs?
>>>
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Apr 21 2016 - 11:30:04 PDT
