<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=us-ascii">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7036.0">
<TITLE>RE: [mpich-discuss] SMPD, Problem launching when using -host</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=2> Hi,<BR>
Yes, if you need to set the export PATH to all the executables you should use the "-genvlist" (& friends).<BR>
<BR>
Regards,<BR>
Jayesh<BR>
<BR>
-----Original Message-----<BR>
From: James S Perrin [<A HREF="mailto:james.s.perrin@manchester.ac.uk">mailto:james.s.perrin@manchester.ac.uk</A>]<BR>
Sent: Friday, October 17, 2008 12:19 PM<BR>
To: Jayesh Krishna<BR>
Cc: mpich-discuss@mcs.anl.gov<BR>
Subject: Re: [mpich-discuss] SMPD, Problem launching when using -host<BR>
<BR>
Hi,<BR>
<BR>
The -path option is of no use as I think mpiexec only uses this path to find the executable. I don't think that it is then set in the environment that the executable is launched in, hence the executable still can't find DLLs.<BR>
<BR>
Using -genvlist PATH or similar seems to work however.<BR>
<BR>
Regards<BR>
James<BR>
<BR>
Jayesh Krishna wrote:<BR>
> Hi,<BR>
> It is recommended that you use the "-path" option available with<BR>
> mpiexec to specify the path to the executable.<BR>
><BR>
> Regards,<BR>
> Jayesh<BR>
><BR>
> -----Original Message-----<BR>
> From: James S Perrin [<A HREF="mailto:james.s.perrin@manchester.ac.uk">mailto:james.s.perrin@manchester.ac.uk</A>]<BR>
> Sent: Friday, October 10, 2008 4:57 AM<BR>
> To: Jayesh Krishna<BR>
> Cc: mpich-discuss@mcs.anl.gov<BR>
> Subject: Re: [mpich-discuss] SMPD, Problem launching when using -host<BR>
><BR>
> Hi,<BR>
><BR>
> I have found the reason why my executable is failing to start, however I<BR>
> think -host is not behaving as it should or at least the<BR>
> documentation needs clarifying.<BR>
><BR>
> I guessed that using -host was somehow changing the executable's<BR>
> environment and so it is failing to start correctly because it<BR>
> couldn't find a dll.<BR>
><BR>
> On windows the PATH variable should be made up of the system wide<BR>
> settings and the user specific additions:<BR>
><BR>
> ie echo %PATH% => <system settings>;<user settings><BR>
><BR>
> The user settings are required to launch the process. When I launch as<BR>
> follows:<BR>
><BR>
> mpiexec -localroot -n 1 master : -n 1 slave<BR>
><BR>
> both get the path setting as above, however if I use<BR>
><BR>
> mpiexec -localroot -n 1 master : -host roobarb -n 1 slave<BR>
><BR>
> process 1 has PATH=<system settings>;<user settings> but process 2 has<BR>
> PATH=<system settings> only<BR>
><BR>
> I have no idea why the following works but it does, if I add -host<BR>
> roobarb to the process 1 process 2 now gets the full PATH variable<BR>
><BR>
> mpiexec -localroot -host roobarb -n 1 master : -host roobarb -n 1<BR>
> slave<BR>
><BR>
> Final permutation, if I now don't specify -localroot both processes<BR>
> only get the only the system settings for PATH:<BR>
><BR>
> mpiexec -host roobarb -n 1 master : -host roobarb -n 1 slave<BR>
><BR>
> In summary using -host only the system path settings are used and not<BR>
> the user specific settings. Is this a security feature or a<BR>
> non-iteractive login issue c.f bash under linux the .bashrc is not<BR>
> executed for processes started remotely?<BR>
><BR>
> A little extra testing confirmed that when process gets both the<BR>
> system and user path settings it is getting this from the current cmd shell.<BR>
><BR>
> The solution is to either make sure paths are added to the system path<BR>
> variable or launch via a script that sets up the environment for each<BR>
> processes though I would have like to avoid this if possible. The<BR>
> first is a pain for development and the later a pain for user installations.<BR>
><BR>
> FYI I was examining the PATH variable using:<BR>
><BR>
> mpiexec -l -host roobarb -n 1 env : -host roobarb -n 1 env | grep<BR>
> \]PATH=<BR>
><BR>
> I have the UNIX commands env and grep in my PATH.<BR>
><BR>
> Regards<BR>
> James<BR>
><BR>
> Jayesh Krishna wrote:<BR>
> > Hi,<BR>
> > Can you send us the debug output of mpiexec and smpd ? Please follow<BR>
> > the instructions below to send us the debug output, > > # Stop<BR>
> any instances of smpd using the command, smpd -stop # Start > smpd in<BR>
> the debug mode using the command, smpd -d # Run a non-MPI > program<BR>
> with mpiexec in the verbose mode using the command, mpiexec ><BR>
> -verbose -n 1 hostname : -host IPADDRESS_OF_roobarb -n > 1 hostname <BR>
> > > # Run an MPI program (cpi.exe provided with MPICH2) with mpiexec<BR>
> in > the verbose mode using the command, mpiexec -verbose -n 1<BR>
> cpi.exe :<BR>
> > -host IPADDRESS_OF_roobarb -n 1 cpi.exe > > # Send us the<BR>
> debug/verbose outputs of mpiexec and smpd.<BR>
> ><BR>
> > Let us know the results.<BR>
> ><BR>
> > Regards,<BR>
> > Jayesh<BR>
> ><BR>
> > -----Original Message-----<BR>
> > From: owner-mpich-discuss@mcs.anl.gov ><BR>
> [<A HREF="mailto:owner-mpich-discuss@mcs.anl.gov">mailto:owner-mpich-discuss@mcs.anl.gov</A>] On Behalf Of James S Perrin <BR>
> > Sent: Tuesday, October 07, 2008 5:25 AM > Cc:<BR>
> mpich-discuss@mcs.anl.gov > Subject: Re: [mpich-discuss] SMPD,<BR>
> Problem launching when using -host > > Hi, ><BR>
> > No I get the same error if I use the ipaddress.<BR>
> ><BR>
> > Regards<BR>
> > James<BR>
> ><BR>
> ><BR>
> > Jayesh Krishna wrote:<BR>
> > > Hi,<BR>
> > > Does it work if you specify the ipaddress of the machine instead of<BR>
> > > hostname (mpiexec -n 1 master : -host IPADDRESS_OF_roobarb -n 1<BR>
> slave) ?<BR>
> > ><BR>
> > > Regards,<BR>
> > > Jayesh<BR>
> > ><BR>
> > > -----Original Message-----<BR>
> > > From: James S Perrin [<A HREF="mailto:james.s.perrin@manchester.ac.uk">mailto:james.s.perrin@manchester.ac.uk</A>]<BR>
> > > Sent: Monday, October 06, 2008 5:18 AM > To: Jayesh Krishna > Cc:<BR>
> > mpich-discuss@mcs.anl.gov > Subject: Re: [mpich-discuss] SMPD, ><BR>
> Problem launching when using -host > > Hi, > > Jayesh Krishna ><BR>
> wrote:<BR>
> > > > Hi,<BR>
> > > ><BR>
> > > > >> mpiexec -n 1 -host roobarb master : -n 1 slave<BR>
> > > > The command above("-host" option specified for only one<BR>
> > > > executable) works for me. What is the error message that you<BR>
> get > > > (Provide us with the snapshot of your command and the<BR>
> error > output. It > > would also help us if you provide more<BR>
> details - Is > roobarb a remote > > machine ? etc) ?<BR>
> > ><BR>
> > > The error is:<BR>
> > ><BR>
> > > [0] PMI_Init failed: FAIL - init called when another process has <BR>
> > > exited without calling init Fatal error in MPI_Init_thread: Other<BR>
> MPI > > error, error stack:<BR>
> > > MPIR_Init_thread(294): Initialization failed > ><BR>
> MPID_Init(82)........: channel initialization failed > ><BR>
> MPID_Init(333).......: PMI_Init returned -1unable to read the cmd > <BR>
> > header on the pmi context, generic socket failure, error stack:<BR>
> > > MPIDU_Sock_wait(2603): The specified network name is no longer <BR>
> > > available. (errno 64).<BR>
> > ><BR>
> > > job aborted:<BR>
> > > rank: node: exit code[: error message] > 0: ROOBARB: 3: Fatal <BR>
> > error in MPI_Init_thread: Other MPI error, error > stack:<BR>
> > > MPIR_Init_thread(294): Initialization failed > ><BR>
> MPID_Init(82)........: channel initialization failed > ><BR>
> MPID_Init(333).......: PMI_Init returned -1 > 1: roobarb: -1073741515 <BR>
> > > > The second process is not starting for some reason.<BR>
> > ><BR>
> > > roobarb happens to be the local machine in this case but the ><BR>
> problem > also occurs on a cluster.<BR>
> > ><BR>
> > > It will launch correctly if I use:<BR>
> > ><BR>
> > > mpiexec -n 1 master : -n 1 slave - SUCCESS > > which should be<BR>
> no > different from:<BR>
> > ><BR>
> > > mpiexec -n 1 master : -host roobarb -n 1 slave - FAILS > ><BR>
> when > everything is running on roobarb.<BR>
> > ><BR>
> > > > >> mpiexec -localroot -n 1 roobarb master : -host roobarb -n<BR>
> 1 > > slave ><BR>
> > > > When using the "-localroot" option you should not<BR>
> specify the<BR>
> > > > hostname for the 1st executable. The command should be, > > <BR>
> > >> > mpiexec -localroot -n 1 master : -host roobarb -n 1 slave > <BR>
> > > sorry typo I meant if would work I used:<BR>
> > ><BR>
> > > mpiexec -localroot -host roobarb -n 1 master : -host roobarb -n<BR>
> 1 > > slave > > Regards > James > > > > > -----Original ><BR>
> Message----- > > From: owner-mpich-discuss@mcs.anl.gov > > ><BR>
> [<A HREF="mailto:owner-mpich-discuss@mcs.anl.gov">mailto:owner-mpich-discuss@mcs.anl.gov</A>] On Behalf Of James S Perrin <BR>
> > > > Sent: Friday, October 03, 2008 12:13 PM > To: mpich > Subject:<BR>
> > > [mpich-discuss] SMPD, Problem launching when using -host > > Hi,<BR>
> > > > Processes fail to start if -host is used for only some but not<BR>
> > > > all processes when launching. ie the machines that some ><BR>
> processes > > launch on is left up to the smpd to allocate.<BR>
> > > ><BR>
> > > > eg<BR>
> > > ><BR>
> > > > mpiexec -n 1 -host roobarb master : -n 1 slave > > when > <BR>
> > -localroot is used the following fails unless -host is also > > ><BR>
> specified for the master.<BR>
> > > ><BR>
> > > > mpiexc -localroot -n 1 roobarb master : -host roobarb -n 1<BR>
> slave > > > > Using MPICH2 1.0.7 on WinXP ia32.<BR>
> > > ><BR>
> > > > Regards<BR>
> > > > James<BR>
> > > > --<BR>
> > > ><BR>
> > ------------------------------------------------------------------------<BR>
> > > > James S. Perrin<BR>
> > > > Visualization<BR>
> > > ><BR>
> > > > Research Computing Services<BR>
> > > > The University of Manchester<BR>
> > > > Kilburn Building, Oxford Road<BR>
> > > > Manchester, M13 9PL<BR>
> > > ><BR>
> > > > t: +44 (0) 161 275 6945<BR>
> > > > e: james.perrin@manchester.ac.uk<BR>
> > > > w: www.manchester.ac.uk/researchcomputing<BR>
> > > ><BR>
> > ------------------------------------------------------------------------<BR>
> > > > "The test of intellect is the refusal to belabour the obvious"<BR>
> > > > - Alfred Bester<BR>
> > > ><BR>
> > ><BR>
> ><BR>
> ----------------------------------------------------------------------<BR>
> > > > --<BR>
> > > ><BR>
> > ><BR>
> > > --<BR>
> > ><BR>
> ------------------------------------------------------------------------<BR>
> > > James S. Perrin<BR>
> > > Visualization<BR>
> > ><BR>
> > > Research Computing Services<BR>
> > > The University of Manchester<BR>
> > > Kilburn Building, Oxford Road<BR>
> > > Manchester, M13 9PL<BR>
> > ><BR>
> > > t: +44 (0) 161 275 6945<BR>
> > > e: james.perrin@manchester.ac.uk<BR>
> > > w: www.manchester.ac.uk/researchcomputing<BR>
> > ><BR>
> ------------------------------------------------------------------------<BR>
> > > "The test of intellect is the refusal to belabour the obvious"<BR>
> > > - Alfred Bester<BR>
> > ><BR>
> ><BR>
> ----------------------------------------------------------------------<BR>
> > > --<BR>
> > ><BR>
> ><BR>
> > --<BR>
> > ------------------------------------------------------------------------<BR>
> > James S. Perrin<BR>
> > Visualization<BR>
> ><BR>
> > Research Computing Services<BR>
> > The University of Manchester<BR>
> > Kilburn Building, Oxford Road<BR>
> > Manchester, M13 9PL<BR>
> ><BR>
> > t: +44 (0) 161 275 6945<BR>
> > e: james.perrin@manchester.ac.uk<BR>
> > w: www.manchester.ac.uk/researchcomputing<BR>
> > ------------------------------------------------------------------------<BR>
> > "The test of intellect is the refusal to belabour the obvious"<BR>
> > - Alfred Bester<BR>
> ><BR>
> ----------------------------------------------------------------------<BR>
> > --<BR>
> ><BR>
><BR>
> --<BR>
> ------------------------------------------------------------------------<BR>
> James S. Perrin<BR>
> Visualization<BR>
><BR>
> Research Computing Services<BR>
> The University of Manchester<BR>
> Kilburn Building, Oxford Road<BR>
> Manchester, M13 9PL<BR>
><BR>
> t: +44 (0) 161 275 6945<BR>
> e: james.perrin@manchester.ac.uk<BR>
> w: www.manchester.ac.uk/researchcomputing<BR>
> ------------------------------------------------------------------------<BR>
> "The test of intellect is the refusal to belabour the obvious"<BR>
> - Alfred Bester<BR>
> ----------------------------------------------------------------------<BR>
> --<BR>
><BR>
<BR>
--<BR>
------------------------------------------------------------------------<BR>
James S. Perrin<BR>
Visualization<BR>
<BR>
Research Computing Services<BR>
Rm 074B, Devonshire House, University Precinct<BR>
The University of Manchester<BR>
Oxford Road, Manchester, M13 9PL<BR>
<BR>
t: +44 (0) 161 275 6945<BR>
e: james.perrin@manchester.ac.uk<BR>
w: www.manchester.ac.uk/researchcomputing<BR>
------------------------------------------------------------------------<BR>
"The test of intellect is the refusal to belabour the obvious"<BR>
- Alfred Bester<BR>
------------------------------------------------------------------------<BR>
</FONT>
</P>
</BODY>
</HTML>