I noticed that there were some mismatched fonts in the command after my
msg was sent through the server.
mpirun -npernode 4 … -j 2 -memory_per_thread 16
-dont_combine_weights_via_disc
The … stands for your relion mpi command.
On 10/17/14, 8:58 AM, "Qiu-Xing Jiang" <[log in to unmask]>
wrote:
>Hi,
>I have had segment faults in a few occasions in the past. My sys admin set
>up a monitor to examine the memory addressing errors and confirmed that
>these memory issues were due to not enough RAM assignment. Based on what
>we have done, I would suggest you to try ³ mpirun -npernode 4 Š -j 2
>--memory_per_thread 16 ‹dont_combine_weights_via_disc². If this still
>crashes, try ³ -j 1 ‹memory_per_thread 32² or ³ -npernode 2 -j 2
>‹memory_per_thread 21², etc. This strategy has worked for solving my
>problems.
>I would also confirm that the first node of your nodelist has 128 GB RAM.
>Hope this would help.
>Best,
>Qiu-Xing
>
>
>
>
>On 10/17/14, 4:50 AM, "Sjors Scheres" <[log in to unmask]> wrote:
>
>>How big are the images? How many movie frames does each image have? What
>>are the command line arguments of the job that crashes?
>>S
>>
>>On 10/17/2014 08:53 AM, SUBSCRIBE CCPEM SIddhanta wrote:
>>> Hi all,
>>>
>>> I am new to Relion. I have been following the tutorial closely to
>>>process around 12000 particles. Processing is done on a cluster (details
>>>given below). At almost every job (mainly for movie processing) we have
>>>faced memory issues or other problems which are causing the program to
>>>crash. However, after extensive trial and error with the cluster admin,
>>>we managed to find optimum settings for MPI processes and threads for
>>>each job.
>>>
>>> Unfortunately, now in particle polishing none of the configurations are
>>>working. I've tried many. It mostly exits with the error 'segmentation
>>>fault' at the particle writing stage or some times with
>>>'std:bad_alloc()' right at the beginning.
>>>
>>> Cluster details: 8 nodes, each with 64 cores and 128 GB ram.
>>>
>>> I'm giving a little extra info which might help in trouble shooting.
>>>(these are all for movie processing)
>>> Optimal settings determined for
>>>
>>> 3D Classification: 128 MPI processes, 4 threads, 2 GB per thread
>>> 3D Auto refine: 32 MPI processes, 16 threads, 2 GB per thread
>>>
>>> For particle polishing, I've tried everything, from 8 MPI and 32
>>>threads to 32 MPI and 4 threads. They all return with segmentation
>>>fault. And higher settings than 32 MPI or lower than 8 MPI cause the
>>>program to crash immediately.
>>>
>>> I hope I can get some help on this. Thanks in advance.
>>
>>--
>>Sjors Scheres
>>MRC Laboratory of Molecular Biology
>>Francis Crick Avenue, Cambridge Biomedical Campus
>>Cambridge CB2 0QH, U.K.
>>tel: +44 (0)1223 267061
>>http://www2.mrc-lmb.cam.ac.uk/groups/scheres
>
>
>________________________________
>
>UT Southwestern Medical Center
>The future of medicine, today.
|