[gmx-users] gromacs on glacier

Justin A. Lemkul jalemkul at vt.edu
Mon Jun 8 21:53:23 CEST 2009



Payman Pirzadeh wrote:
> Dear Justin,
> Here is the mpich version:
> MPICH Version:          1.2.7p1
> MPICH Release date:     $Date: 2005/11/04 11:54:51$
> MPICH Patches applied:  none
> MPICH configure:        -prefix=/share/apps/intel/mpich/
> MPICH Device:           ch_p4
> 

Well, I think that MPICH is your problem then.  There are several reports of 
sporadic bugs (the one that you're seeing!) with that version.  Since it's 
several years old, getting a fix is probably a bit unlikely :)

Perhaps you can sort out with the sysadmins what you can do.  Like I said 
before, probably OpenMPI is a better bet - we've never had a problem with it. 
You can probably install it yourself in your home directory, and point to mpicc 
with environment variables during the Gromacs installation.

> Also the gcc compiler:
> gcc (GCC) 3.4.6 20060404 (Red Hat 3.4.6-9)
> Copyright (C) 2006 Free Software Foundation, Inc.
> 
> About the system, I have water model with 3 atom sites and 3 virtual sites.
> 

Probably not a very good test case for diagnosing problems, but I think it's 
unrelated in this case.  When testing, keep it simple - run a simulation with 
spc216.gro from the Gromacs installation, it should pretty much always work :)

-Justin

> Payman
> 
> -----Original Message-----
> From: gmx-users-bounces at gromacs.org [mailto:gmx-users-bounces at gromacs.org]
> On Behalf Of Justin A. Lemkul
> Sent: June 8, 2009 11:56 AM
> To: Gromacs Users' List
> Subject: Re: [gmx-users] gromacs on glacier
> 
> 
> 
> Payman Pirzadeh wrote:
>> Hi Justin,
>> Since the manual itself was not sufficient, I asked some other people who
>> are running GROMACS in our group (but they run only on 2 CPUs). Here are
> the
>> steps I took to compile the parallel version (I have included my notes
> that
>> they told me as well):
>>
>> Installation
>> 1.	./configure --prefix=/global/home/pirzadeh/gromacs-4.0.4
>> This line specifies the installation path
>>
>> 2.	Make
>> 3.	Make install
>> These two commands will make 'grompp' and analysis functions of GROMACS.
>>
>> 4.	Make clean
>> This command will clean some files generated during installation which are
>> not needed anymore.
>>
>> 5.	./configure --enable-mpi --disable-nice
>> --prefix=/global/home/pirzadeh/gromacs-4.0.4
>> Here we compile the code for parallel version of GROMACS
>>
>> 6.	make mdrun
>> 7.	 make install-mdrun
>> Now the parallel version of mdrun is built. The analysis functions are
> found
>> in 'bin' folder accompanied by 'GMXRC'.
>>
>> 8.	Before running 'grompp' to produce the topology file for simulation,
>> we should use the command  source
>> /global/home/pirzadeh/gromacs-4.0.4/bin/GMXRC to specify the path for the
>> current code.
>>
> 
> This is pretty much the standard installation procedure.  What we also need
> to 
> know are the compilers used, etc. for the installation.  The error you're 
> getting is from MPICH.  Which version is installed on the cluster?  It may
> be 
> old and buggy.  In any case, you can try to install something newer, like a 
> recent version of OpenMPI (which we have on our cluster); it may be more 
> reliable.  Only a random thought, means nothing unless we know what you have
> 
> installed :)
> 
> What about the contents of your system, as I asked before?  Do you really
> have a 
> 100% virtual site system?
> 
> -Justin
> 
>> Sorry for tons of e-mails.
>>
>> Payman
>>
>> -----Original Message-----
>> From: gmx-users-bounces at gromacs.org [mailto:gmx-users-bounces at gromacs.org]
>> On Behalf Of Justin A. Lemkul
>> Sent: June 8, 2009 11:12 AM
>> To: Discussion list for GROMACS users
>> Subject: Re: [gmx-users] gromacs on glacier
>>
>>
>>
>> Payman Pirzadeh wrote:
>>> Hi,
>>>
>>> I had the chance to run the GROMACS 4.0.4 on another cluster. Same 
>>> problem still persists. But what I found is that it can be run on a node 
>>> with 2 CPUs, but as soon as the number of nodes are increased to 2, 3, . 
>>> it will crash. Following are the last lines reported in different files:
>>>
>>> "In the log file of the code":
>>>
>>>  
>>>
>>> There are: 1611 Atoms
>>>
>>> There are: 1611 VSites
>> All of your atoms are virtual sites?  If so, I would try a simpler test
>> case, to 
>> rule out stumbling across some obscure bug.
>>
>> Also:
>>
>>> p2_22627:  p4_error: Timeout in establishing connection to remote
> process:
>> 0
>>
>> This is an error message from MPICH, not Gromacs.  See, for example:
>>
>> http://www.mail-archive.com/gmx-users@gromacs.org/msg10968.html
>>
>> <snip>
>>
>>> To me, it seems that code can not communicate through more than one 
>>> node. I am suspicious of doing sth wrong during installation! I tried 
>>> wiki, but I can not find the documents as before, and I eally do not 
>>> know in which step I might have gone wrong.
>>>
>> If you suspect you have done something wrong, then post the details of the
> 
>> system configuration (hardware, compilers, OS, etc.) as well as a
>> step-by-step 
>> record of what you did to compile the software.  If your procedure is
> sound,
>> then it helps rule out the possibility that you messed something up.
>>
>> -Justin
>>
>>>  
>>>
>>> Payman
>>>
>>>  
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> gmx-users mailing list    gmx-users at gromacs.org
>>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>>> Please search the archive at http://www.gromacs.org/search before
> posting!
>>> Please don't post (un)subscribe requests to the list. Use the 
>>> www interface or send it to gmx-users-request at gromacs.org.
>>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================



More information about the gromacs.org_gmx-users mailing list