AMBER Archive (2006)Subject: Re: AMBER: job crashes
From: Xiaowei (David) Li (xl3a_at_virginia.edu)
Date: Tue Sep 26 2006 - 22:57:18 CDT
Hi, Junjun:
Thanks a lot for your help. I certainly will try "cleanipcs".
Best,
Xiaowei
JunJun Liu wrote:
> Hi Xiaowei,
>
> Google it with "semop lock failed" and you will know it's related to
> your MPI shared memory. Try using "cleanipcs" to clean them up in
> each computation node.
>
> Good luck!
>
> Liu
>
> On Tue, 26 Sep 2006 23:09:04 -0400, Xiaowei (David) Li
> <xl3a_at_virginia.edu> wrote:
>
>> Dear all:
>> I have met intensive (almost every job I submitted) MD job
>> crashes during recent simulation work. The job crashes always
>> happend upon the completion points of simulations (for example, the
>> crash happens around 950 ps for a 1 ns simulation). All of the
>> errors messages have the "semop lock failed" information as following.
>> Job is running on node(s):
>> ------------------------
>> compute-2-5 compute-2-6 compute-2-7 compute-2-9
>> ------------------------
>> p4_error: latest msg from perror: Invalid argument
>> p0_9469: p4_error: OOPS: semop lock failed: -1
>> forrtl: error (69): process interrupted (SIGINT)
>> forrtl: error (69): process interrupted (SIGINT)
>> forrtl: error (69): process interrupted (SIGINT)
>> forrtl: error (69): process interrupted (SIGINT)
>> p3_28418: (207846.623829) net_send: could not write to fd=5, errno = 32
>> forrtl: error (69): process interrupted (SIGINT)
>> p0_9469: (207848.918605) net_send: could not write to fd=4, errno = 32:
>>
>> I was running the parallel simulation with MPI on a linux cluster
>> with Athlon Opteron 244 processors. The input file is:
>> &cntrl
>> imin = 0,
>> irest = 1,
>> ntx = 5,
>> ntb = 2,
>> pres0 = 1.0,
>> ntp = 1,
>> tautp=5
>> taup =5,
>> cut = 10,
>> ntr = 0,
>> ntc = 2,
>> ntf = 2,
>> tempi = 300.0,
>> temp0 = 300.0,
>> ntt = 1,
>> nstlim =500000,
>> dt = 0.002,
>> ntpr = 100,
>> ntwx = 100,
>> ntwr = 1000,
>> nscm=1,
>> &end
>> Any help or suggestion will be deeply appreciated. Thanks.
>>
>> Best,
>> Xiaowei Li
>> University of Virginia
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber_at_scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
>
>
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
|