[mpich-discuss] Is MPICH2 connection failure aware ?

Seifer Lin seiferlin at gmail.com
Wed Apr 1 11:27:13 CDT 2009


Hi:

Unplug the wire of node1 : mpiexec abort

Unplug the wire of node2: the MPI task at node2 finally disappeared, the MPI
task at node1 hangs (from the console window it just hangs)

regards,
Seifer

2009/4/1 Jayesh Krishna <jayesh at mcs.anl.gov>

>  Hi,
>  Did the application crash (or was it an abort ?)?
>
> Regards,
> Jayesh
>
>  ------------------------------
> *From:* mpich-discuss-bounces at mcs.anl.gov [mailto:
> mpich-discuss-bounces at mcs.anl.gov] *On Behalf Of *Seifer Lin
> *Sent:* Wednesday, April 01, 2009 12:31 AM
> *To:* mpich-discuss at mcs.anl.gov
> *Subject:* [mpich-discuss] Is MPICH2 connection failure aware ?
>
>   Hi all:
>
> I have run 2 MPI tasks in 2 different nodes. (2 nodes are with Windows XP
> Professional)
>
> The mpiexec command is run on node1.
>
> During the parallel computing, if I unplug the network wire of node1, the
> parallel computing will crash.
>
> But if I unplug the network wire of node2, the parallel computing just
> hangs! I notice that the hanging results from
> that I unplug the wire before a MPI_Barrier, therefore node1 waits at
> MPI_Barrier, but node2 will never reach that point.
>
> Is there any mechanism to make the parallel computing crash after the
> network wire of node2 is unplugged ?
>
> regards,
>
> Seifer
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20090402/0df08793/attachment-0001.htm>


More information about the mpich-discuss mailing list