Hi,
I don't know how alias interacts with RELION.
You can create a job submission script by saying 'submit to queue: Yes'
and edit it before running it locally.
Best regards,
Takanori Nakane
On 2019/07/02 23:41, Joel Meyerson wrote:
> Hi Takanori,
> Thank you for the suggestion, and I'm on one of our lab workstations not a cluster. As a first step, I tested mpirun without and with the hostfile as follows, and succeeded with the hostfile. Note that the hostfile contains "localhost slots=8 max-slots=8".
>
> [joel@localhost ~]$ mpirun -np 8 python -c "print 'hey'"
> --------------------------------------------------------------------------
> There are not enough slots available in the system to satisfy the 8 slots
> that were requested by the application:
> python
>
> Either request fewer slots for your application, or make more slots available
> for use.
> --------------------------------------------------------------------------
> [joel@localhost ~]$ mpirun --hostfile hostfile -np 8 python -c "print 'hey'"
> hey
> hey
> hey
> hey
> hey
> hey
> hey
> hey
>
> Next, I set mpirun to have a new alias in my bashrc, so it would always default to running with the hostfile. Note, I confirmed that this alias works with the above Python example.
>
> alias mpirun='mpirun --hostfile ~/hostfile'
>
> However, when I try to run Relion it throws the same error about not enough slots being available. Is it possible that Relion is somehow calling mpirun through some other means, so does not see the alias I'm using?
>
> Thanks again,
> Joel
>
> ________________________________________
> From: Takanori Nakane <[log in to unmask]>
> Sent: Wednesday, June 26, 2019 1:39 AM
> To: Joel Meyerson; [log in to unmask]
> Subject: [EXTERNAL] Re: [ccpem] problem running MPI on all threads
>
> Hi,
>
> This is probably a problem with your MPI runtime, not RELION.
> Your MPI runtime is not aware that you may use 8 processes.
>
> If this is on a cluster, you should ask your cluster administrator
> how to properly submit a job. If this is your own machine,
> you can edit your 'machinefile' or force 'oversubscribing'.
> See https://urldefense.proofpoint.com/v2/url?u=https-3A__www.open-2Dmpi.org_faq_-3Fcategory-3Drunning-23oversubscribing&d=DwIC-g&c=lb62iw4YL4RFalcE2hQUQealT9-RXrryqt9KZX2qu2s&r=JnYzuUa3E5p4u7qrfPzeZV8LIgMpwauz4WIvX7NJIi8&m=UT-q1l7h2_A-a0JAG2sSZFrg4sKeBkG4vHunt7NSnFM&s=j-AkTS6z3MMfP4bymtK9dWPI1r-If-9TeAVXOz1j7hw&e= .
> (Don't do this on a cluster, because it might affect other users!)
>
> Best regards,
>
> Takanori Nakane
>
>
> On 2019/06/26 5:22, Joel Meyerson wrote:
>> I am unable to run Relion jobs with MPI procs on all available threads. For example, if running CTFFIND4 on our 4 core / 8 hyperthread machine, the job runs successfully with 4x MPI procs, but fails with 8x MPI procs. The error message reports: "Either request fewer slots for your application, or make more slots available for use".
>>
>> One suspicion is that this problem may have arisen during compiling, so I've provided commands for "cmake" and "make" below, used when building Relion. I've also attached the output for the two commands. The system has Relion 3, Centos 7, openmpi3.
>>
>> sudo cmake -DCUDA_ARCH=61 -DMPI_CXX_COMPILER=/usr/lib64/openmpi3/bin/mpicxx -DMPI_C_COMPILER=/usr/lib64/openmpi3/bin/mpicc ..
>>
>> sudo make -j 8
>>
>> Any suggestions on how to solve this?
>>
>> Thanks,
>> Joel
>> ########################################################################
>>
>> To unsubscribe from the CCPEM list, click the following link:
>> https://urldefense.proofpoint.com/v2/url?u=https-3A__www.jiscmail.ac.uk_cgi-2Dbin_webadmin-3FSUBED1-3DCCPEM-26A-3D1&d=DwIC-g&c=lb62iw4YL4RFalcE2hQUQealT9-RXrryqt9KZX2qu2s&r=JnYzuUa3E5p4u7qrfPzeZV8LIgMpwauz4WIvX7NJIi8&m=UT-q1l7h2_A-a0JAG2sSZFrg4sKeBkG4vHunt7NSnFM&s=rWu1hnkFR-fItsTTPh2OzsXzBvTrRvFp4fkSGVmaWdQ&e=
>>
>
> ########################################################################
>
> To unsubscribe from the CCPEM list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|