[mpich-discuss] Unable to connect issue

[Jayesh Krishna]

Hi,
 Do you have the same SMPD passphrase specified (during installation
process) on both the machines ?
 Can you try uninstalling-installing MPICH2 on both machines ? Make sure
that you use the same MPICH2 installable to install MPICH2 on the machines
and that you specify the same passphrase during installation.
 Let us know if that does not work.

Regards,
Jayesh

  _____

From: bob ilgner [mailto:bobilgner at gmail.com]
Sent: Friday, September 19, 2008 10:26 AM
To: Jayesh Krishna
Subject: Re: [mpich-discuss] Unable to connect issue


Hi,

The smpd versions are both the same. 1.0.7

I have attached the logs for the machine master(images16) and the
slave(images17)

Regards, bob






2008/9/19 Jayesh Krishna <jayesh at mcs.anl.gov>


Hi,
 Do you have the same version of MPICH2 installed on both the machines ?
You can check the version of smpd using the "smpd -version" command.
 The error does not seem to be because of a firewall issue. If the above
suggestion does not work please provide me the smpd debug outputs from
BOTH hosts and the mpiexec verbose output.

(PS: I am one of the contributors to the MPICH2 dev manual. Let me know if
you need any info to be added to the manual.)

Regards,
Jayesh

  _____

From: bob ilgner [mailto:bobilgner at gmail.com]
Sent: Friday, September 19, 2008 10:03 AM
To: Jayesh Krishna

Subject: Re: [mpich-discuss] Unable to connect issue


Hi Jayesh,

Your assumption is correct. The command was mpiexec -hosts 2 images16
images17 cpi

mpiexec -n 2 cpi.exe works fine

Same error message if I use IP addresses instead of names for mpiexec
command

I have attached the debug log for smpd and the verbose log for the mpiexec
as requested.

I have just put 2+2 together and noticed that you put together the
development manual for the MPICH2.

Regards, bob








2008/9/19 Jayesh Krishna <jayesh at mcs.anl.gov>


Hi,

# Can you run the MPI program on a single host ? (i.e., mpiexec -n 2
cpi.exe)
# Can you try specifying ipaddresses of the hosts instead of the hostnames
? (i.e., mpiexec -hosts 2 IPADDRESS_OF_IMAGES16 IPADDRESS_OF_IMAGES17
cpi.exe)

    If the above suggestions don't work please provide us the verbose
output of smpd and mpiexec. To get the verbose output of smpd & mpiexec,

# Stop any instances of smpd using, smpd -stop, command
# Run smpd in debug mode using, smpd -d, command
# Run mpiexec in the verbose mode using the "-verbose" option of mpiexec.

   Let us know the results.

(Note: In the command that you listed in your email both hostnames are the
same -- there is no images16 specified with "-hosts" arg of mpiexec. I am
assuming you meant "mpiexec -hosts 2 images16 images17 cpi.exe")

Regards,
Jayesh

  _____

From: owner-mpich-discuss at mcs.anl.gov
[mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of bob ilgner
Sent: Friday, September 19, 2008 8:30 AM
To: mpich-discuss at mcs.anl.gov
Subject: Re: [mpich-discuss] Unable to connect issue


Hi,

I am running XP Pro on the machines and am therefore using smpd. As far as
I know the mpd is used in the unix/linux environment. The smpd is
operational on both PCs. The local check for the status os the smpd on
either machine shows that they are running.  When i do a remote check with
smdp I get the same message.i.e.

smpd -status <remotehostname>
abort: unable to connect to <remotehostname>



Regards, bob


On Fri, Sep 19, 2008 at 1:38 PM, The Source <thesourcehim at gmail.com>
wrote:


do you have mpd ring running and connected on both machines?

bob ilgner пишет:


I have installed mpich2 1.0.7 on 2 XP Pro hosts and am trying to run the
cpi process on both hosts with the command
mpiexec -hosts 2 images17 images17 cpi
where images16 and images17 are the hosts and cpi is the supplied example
application and get the error:
abort: unable to connect to images16
I have checked passwords and regsitration is the same on both hosts.
Firewalls are both down. Host names defined in hosts file under system32.
I have ensured that smpd is operating on both machines. Usernames and
password are the same on both hosts and have both been delegated. I have
gone through the mpich2 user manual and find no way forward with this. Is
there anything else I could try ?
Would it help if I listed the output from the mpiexec with the verbose
switch on.
Regards, bob






-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20080919/c9256648/attachment.htm>