It might be time to have a look at the way in which various SMP and process environment (such as limits) are specified, and come up with a unified approach (including a replacement for the old MPIT proposal, which even google can no longer find; it described a way to define the partitioning of threads/processes between MPI and the application). Bill Begin forwarded message:

From: [email protected] Date: September 26, 2007 9:44:06 AM CDT To: [email protected] Subject: BOUNCE [email protected]: Non-member submission from ["Jose Francisco Nanez Gomez" <[email protected]>]

From [email protected] Wed Sep 26 09:44:04 2007 Received: from mailgw.mcs.anl.gov (mailgw.mcs.anl.gov [140.221.9.4]) by mcs.anl.gov (8.11.6/8.9.3) with ESMTP id l8QEi4Z135500 for <[email protected]>; Wed, 26 Sep 2007 09:44:04 -0500 Received: from localhost (localhost [127.0.0.1]) by mailgw.mcs.anl.gov (Postfix) with ESMTP id 6EA54348003 for <[email protected]>; Wed, 26 Sep 2007 09:44:04 -0500 (CDT) Received: from fk-out-0910.google.com (fk-out-0910.google.com [209.85.128.188]) by mailgw.mcs.anl.gov (Postfix) with ESMTP id 3EAFB348002 for <[email protected]>; Wed, 26 Sep 2007 09:44:01 -0500 (CDT) Received: by fk-out-0910.google.com with SMTP id f40so2267494fka for <[email protected]>; Wed, 26 Sep 2007 07:44:00 -0700 (PDT) Received: by 10.82.156.12 with SMTP id d12mr1700842bue.1190817830039; Wed, 26 Sep 2007 07:43:50 -0700 (PDT) Received: by 10.82.148.6 with HTTP; Wed, 26 Sep 2007 07:43:49 -0700 (PDT) Message-ID: <[email protected]> Date: Wed, 26 Sep 2007 09:43:49 -0500 From: "Jose Francisco Nanez Gomez" <[email protected]> To: [email protected], [email protected] Subject: mpiexec: failed to obtain sock from manager MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_6861_28773483.1190817830021" X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mailgw.mcs.anl.gov X-Spam-Status: No, hits=0.1 tagged_above=-0.1 required=5.5 tests=BAYES_00, HTML_00_10, HTML_MESSAGE, RCVD_BY_IP, SARE_MSGID_LONG40 X-Spam-Level: X-MCS-Mail-Loop: mpi-users

------=_Part_6861_28773483.1190817830021 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline

We are trying to run the WRF model in a cluster of two intel machines (each one with two dual core procesors x86_64) under linux fedora 6. We can run succesfully the model in one machine with 4 nodes (two dual core 2x processor), Using MPICH2 2-1.0.5p4 with PGI version 7.0.7. However, when we run the same model in the cluster= , Mpich 2 only recognizes two processors (one per machine) as it is shown by the comands: mpdtrace, mpdringtest and mpiexec.

We machine file this:

[administrador@cs-room443-d01 ~]$ cat mf cs-room443-d01:4 cs-room443-d02:4

[administrador@cs-room443-d01 Paquetes]$ mpdtrace cs-room443-d01 cs-room443-d02

[administrador@cs-room443-d01 Paquetes]$ mpdringtest time for 1 loops =3D 0.000850915908813 seconds

[administrador@cs-room443-d01 Paquetes]$ mpiexec -l -n 1 /bin/hostname 0: cs-room443-d01

[administrador@cs-room443-d01 Paquetes]$ mpiexec -l -n 2 /bin/hostname 0: cs-room443-d01 1: cs-room443-d02

[administrador@cs-room443-d01 Paquetes]$ mpiexec -l -n 3 /bin/hostname mpiexec_cs-room443-d01 (mpiexec 425): mpiexec: failed to obtain sock from manager

is there any way to configure mpich for our cluster with 8 nodes?. Any suggestion will be helpful.

Thanks.

Joseph =D1a=F1ez Systems Engineer National Unversity of Colombia.

------=_Part_6861_28773483.1190817830021 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline

<br>We are trying to run the WRF model in a cluster of two intel machines (each one with two dual core procesors x86_64) under linux fedora 6.<br>We = can run succesfully the model in one machine with 4 nodes (two dual core 2x= processor), Using MPICH2 2-1.0.5p4<br>with PGI version 7.0.7.  However, when we run the same model in the cluster= , Mpich 2 only recognizes two processors (one per machine) as it is shown by the comands: mpdtrace, mpdringtest and mpiexec.<br><br>We machine file t= his:<br><br>[administrador@cs-room443-d01 ~]$ cat mf<br>cs-room443- d01:4 <b= r>cs-room443-d02:4  <br><br><br>[administrador@cs-room443-d01 Paquetes= ]$ mpdtrace <br>cs-room443-d01<br>cs-room443-d02<br><br>[administrador@cs- room443-d01 P= aquetes]$ mpdringtest<br>time for 1 loops =3D 0.000850915908813 seconds<br>= <br>[administrador@cs-room443-d01 Paquetes]$ mpiexec -l -n 1 /bin/ hostname <br>0: cs-room443-d01<br><br>[administrador@cs-room443-d01 Paquetes] $ mpiex= ec -l -n 2 /bin/hostname<br>0: cs-room443-d01<br>1: cs-room443- d02<br><br>[= administrador@cs-room443-d01 Paquetes]$ mpiexec -l -n 3 /bin/ hostname<br> mpiexec_cs-room443-d01 (mpiexec 425): mpiexec: failed to obtain sock from m= anager<br><br><br>is there any way to configure mpich for our cluster with = 8 nodes?. Any suggestion will be helpful.<br><br>Thanks.<br><br>Joseph =D1a= =F1ez <br>Systems Engineer<br>National Unversity of Colombia.

------=_Part_6861_28773483.1190817830021--