[Vserver] mpi application problems within vserver

From: Dimitris Zilaskos <dzila_at_tassadar.physics.auth.gr>
Date: Sun 26 Mar 2006 - 15:46:00 BST
Message-ID: <Pine.LNX.4.64.0603261720590.27960@tassadar.physics.auth.gr>

         Hi,

         I have setup a Scientific Linux 4.2 (RHEL 4 compatible) as a host
system for a Scientific Linux 3.0.5 (RHEL3 compatible) guest, following the
example for CentOS. I am attemping to run a fortran mpi application within
the guest but I am getting random application crashes or invalid results. The
same setup works fine on a non vserver Scientific Linux 3.0.5 system with
the same software installed. However, after say 5-20 attemps at one point
the program will execute and complete correctly within the vserver.
         I suspect it has something to do with the way vserver handles lo
interface, since in the non-vserver system I can see all mpi communication
going through the lo interface, and there is just one connection between
the processes. On the other hand, on the vserver system I can see a lot of
connections opening and closing all the time. If i run tcpdump inside the
vserver for the lo interface I also see the mpi traffic(I have enabled
RAW_NET capability to do so).
         I have tried with both mpich1 and mpich2 but it made no
difference. MPICH and the application are compiled with intel c/fortran
compilers. Binding mpich to a spcific ip address did not help,
neither giving a name to the vserver interface or playing with /etc/hosts
and pointing localhost to the vserver's ip address.
         Hardware is a Sun dual core dual cpu opteron for the vserver
system and a dual cpu xeon 3.6 for the non-vserver. Vserver system is
using 2.6.14.3 with vserver 2.0.1 and the other one is running
2.4.21-37.0.1.ELsmp.
         I would like to listen if other people have any experience on this or
any ideas on how to proceed.

         Best regards,

--
============================================================================
Dimitris Zilaskos
Department of Physics @ Aristotle University of Thessaloniki , Greece
PGP key : http://tassadar.physics.auth.gr/~dzila/pgp_public_key.asc
 	  http://egnatia.ee.auth.gr/~dzila/pgp_public_key.asc
MD5sum  : de2bd8f73d545f0e4caf3096894ad83f  pgp_public_key.asc
============================================================================
_______________________________________________
Vserver mailing list
Vserver@list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver
Received on Sun Mar 26 15:46:23 2006
[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Sun 26 Mar 2006 - 15:46:28 BST by hypermail 2.1.8