[mpich-discuss] hydra_pmi_proxy: No such file

Dave Goodell goodell at mcs.anl.gov
Fri Oct 5 17:25:29 CDT 2012 

MPICH2 does not support running in a heterogeneous configuration (some 32-bit binaries and some 64-bit binaries).  However, on most Linux distributions you can run 32-bit binaries on 64-bit platforms.  Just configure your MPICH2 with "-m32" in your CFLAGS/FFLAGS/FCFLAGS/CXXFLAGS and you should get a working installation for both machines that can be shared over NFS.

-Dave

On Oct 5, 2012, at 4:47 PM CDT, Martin Gilchrist wrote:

> I did a uname -a on the two nodes and one is 64 and the other is 32 bit. 
>  
> I guess I'll have to install and build mpich2 on the nodes separately, rather than have it on an nfs shared directory.
>  
> If anyone has a simpler solution please let me know.
>  
> Thanks all.
>  
> Martin.
> 
> From: Dave Goodell <goodell at mcs.anl.gov>
> To: Martin Gilchrist <gilchrist.martin at yahoo.com>; mpich-discuss at mcs.anl.gov 
> Cc: Pavan Balaji <balaji at mcs.anl.gov> 
> Sent: Friday, October 5, 2012 2:34 PM
> Subject: Re: [mpich-discuss] hydra_pmi_proxy: No such file
> 
> What is the output of "uname -a" on each host?
> 
> -Dave
> 
> On Oct 5, 2012, at 4:04 PM CDT, Martin Gilchrist wrote:
> 
> > Hi Pavan.
> >  
> > I've got some more information that might give some insight into what is happening on my system.
> >  
> > When I do ldd ./cpi on the base node where nfs is installed and mpich was compiled I get  some paths to libraries, but when I do it on the other node I get the message that ./cpi is not a dynamic executable.
> >  
> > When I do mpicc ./child.c, an a.out is produced, when I do the same thing on the other node I get messages like:
> >  
> > /usr/bin/ld: skipping incompatible /mirror/mpich2/lib/libmpich.a when searching for -lmpich.
> >  
> > Also, a simple invocation of ./cpi on the main node works, but gives the error message: bash: ./cpi: No such file or directory when run on the other node.
> >  
> > I have mpich and the user mpiu installed on nfs.  Could it be that the binaries compiled on one node are not compatible with the processor on the other node?
> >  
> > Martin. 
> > 
> > From: Pavan Balaji <balaji at mcs.anl.gov>
> > To: Martin Gilchrist <gilchrist.martin at yahoo.com> 
> > Cc: mpich-discuss at mcs.anl.gov 
> > Sent: Wednesday, October 3, 2012 8:29 PM
> > Subject: Re: [mpich-discuss] hydra_pmi_proxy: No such file
> > 
> > 
> > This is weird.  Can you run mpiexec with the -verbose option and send 
> > the output?
> > 
> >  -- Pavan
> > 
> > On 10/03/2012 05:09 PM, Martin Gilchrist wrote:
> > > Yes.
> > >
> > > The first is the machine where nfs host machine and the second is the
> > > other node.
> > >
> > >
> > > mpiu at martinlaptop-HP-Pavilion-dm4-Notebook-PC:~$ hostname
> > > martinlaptop-HP-Pavilion-dm4-Notebook-PC
> > > mpiu at martinlaptop-HP-Pavilion-dm4-Notebook-PC:~$ ls -l
> > > /mirror/mpich2/bin/mpiexec
> > > lrwxrwxrwx 1 mpiu mpiu 13 2012-10-03 14:32 /mirror/mpich2/bin/mpiexec ->
> > > mpiexec.hydra
> > > mpiu at martinlaptop-HP-Pavilion-dm4-Notebook-PC:~$ ls -l
> > > /mirror/mpich2/bin/mpiexec.hydra
> > > -rwxr-xr-x 1 mpiu mpiu 1065207 2012-10-03 14:32
> > > /mirror/mpich2/bin/mpiexec.hydra
> > > mpiu at martinlaptop-HP-Pavilion-dm4-Notebook-PC:~$ ls -l
> > > /mirror/mpich2/bin/hydra_pmi_proxy
> > > -rwxr-xr-x 1 mpiu mpiu 851803 2012-10-03 14:32
> > > /mirror/mpich2/bin/hydra_pmi_proxy
> > > mpiu at martinlaptop-HP-Pavilion-dm4-Notebook-PC:~$
> > >
> > >
> > >
> > >
> > > mpiu at martin-desktop:~$ ls -l /mirror/mpich2/bin/mpiexec
> > > lrwxrwxrwx 1 mpiu mpiu 13 2012-10-03 14:32 /mirror/mpich2/bin/mpiexec ->
> > > mpiexec.hydra
> > > mpiu at martin-desktop:~$ ls -lat /mirror/mpich2/bin/mpiexec.hydra
> > > -rwxr-xr-x 1 mpiu mpiu 1065207 2012-10-03 14:32
> > > /mirror/mpich2/bin/mpiexec.hydra
> > > mpiu at martin-desktop:~$ ls -l /mirror/mpich2/bin/hydra_pmi_proxy
> > > -rwxr-xr-x 1 mpiu mpiu 851803 2012-10-03 14:32
> > > /mirror/mpich2/bin/hydra_pmi_proxy
> > > mpiu at martin-desktop:~$
> > >
> > >
> > > ------------------------------------------------------------------------
> > > *From:* Pavan Balaji <balaji at mcs.anl.gov>
> > > *To:* Martin Gilchrist <gilchrist.martin at yahoo.com>
> > > *Cc:* "mpich-discuss at mcs.anl.gov" <mpich-discuss at mcs.anl.gov>
> > > *Sent:* Wednesday, October 3, 2012 3:46 PM
> > > *Subject:* Re: [mpich-discuss] hydra_pmi_proxy: No such file
> > >
> > > Hi Martin,
> > >
> > > Can you please stick to one configuration?  It's hard for me to tell
> > > when you are running in which configuration.
> > >
> > > Can you please run the following on each of the two machines:
> > >
> > > % hostname
> > >
> > > % ls -l /mirror/mpich2/bin/mpiexec
> > >
> > > % ls -l /mirror/mpich2/bin/hydra_pmi_proxy
> > >
> > >    -- Pavan
> > >
> > > On 10/03/2012 04:44 PM, Martin Gilchrist wrote:
> > >  > Yes the two are different.
> > >  > The one where the proxy is not found is produced when mpiexec is invoked
> > >  > on the main computer where I have installed nfs and done all the make
> > >  > and make install of mpich2.  The one with the mpiexec not found is
> > >  > produced when mpiexec is invoked on the node.
> > >  > Martin.
> > >  >
> > >  > *From:* Pavan Balaji <balaji at mcs.anl.gov <mailto:balaji at mcs.anl.gov>>
> > >  > *To:* Martin Gilchrist <gilchrist.martin at yahoo.com
> > > <mailto:gilchrist.martin at yahoo.com>>
> > >  > *Cc:* mpich-discuss at mcs.anl.gov <mailto:mpich-discuss at mcs.anl.gov>;
> > > Anthony Chan <chan at mcs.anl.gov <mailto:chan at mcs.anl.gov>>
> > >  > *Sent:* Tuesday, October 2, 2012 1:39 PM
> > >  > *Subject:* Re: [mpich-discuss] hydra_pmi_proxy: No such file
> > >  >
> > >  >
> > >  > On 10/02/2012 03:27 PM, Martin Gilchrist wrote:
> > >  >  > I can ssh into the node and hydra_pmi_proxy is visible and executable.
> > >  >  > And when I am logged into the node directly it is visible and
> > > executable.
> > >  >
> > >  > Both those statement sound identical.  Did you mean to add a "not"
> > >  > somewhere there?
> > >  >
> > >  >  > One more thing, when I am directly logged into the node and do
> > >  >  > mpiexec -n 2 -f machinefile ./cpi
> > >  >  > I get the message:
> > >  >  > bash:/mirror/mpich2/bin/mpiexec:No such file or directory.
> > >  >  > It looks like it might be an issue with permissions since bash can't
> > >  >  > find the executable, but it is visible via which and has executable
> > >  >  > permission under ls -lat.
> > >  >
> > >  > This is not an executable permission issue.  It should say "Permission
> > >  > denied" for that, not "No such file or directory".
> > >  >
> > >  > Can you send me the following output?
> > >  >
> > >  > % ls -l /mirror/mpich2/bin/mpiexec
> > >  >
> > >  > % /mirror/mpich2/bin/mpiexec -n 2 -f machinefile ./cpi
> > >  >
> > >  > I'm having trouble believing that you can see the file, but are getting
> > >  > the following error, so I'd like to see that myself :-).  Of course,
> > >  > it's always possible that it's some weird NFS-ness that I'm not aware of.
> > >  >
> > >  > -- Pavan
> > >  >
> > >  > -- Pavan Balaji
> > >  > http://www.mcs.anl.gov/~balaji
> > >  >
> > >  >
> > >
> > > --
> > > Pavan Balaji
> > > http://www.mcs.anl.gov/~balaji <http://www.mcs.anl.gov/~balaji>
> > >
> > >
> > 
> > -- 
> > Pavan Balaji
> > http://www.mcs.anl.gov/~balaji
> > 
> > 
> > _______________________________________________
> > mpich-discuss mailing list    mpich-discuss at mcs.anl.gov
> > To manage subscription options or unsubscribe:
> > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> 
> 
>

More information about the mpich-discuss mailing list