Re: [BProc] Two problems with CM3 (NFS and MPI)

[Jack N. <sl...@qu...>]

> > * Second one: MPI.
> > I can't run a mpi program on 2 or more nodes. I get the next messages:
> > 
> >    [root@nereapc examples]# mpirun -P -np 3 -d ./cpi
> >    listen: n-1 33276
> >      0    0 31728 192.168.1.100   32773
> >      1    1 31729 192.168.1.101   32773
> >      2    2 31730 192.168.1.102   32773
> >    Process 0 on n0
> >    rank 2 pid=31730 exited with signal 2
> >    rank 1 pid=31729 exited with signal 2
> >    xm_31728: (0.007068) net_recv failed for fd = 6
> >    xm_31728:  p4_error: net_recv read, errno = : 104
> >    rank 0 pid=31728 exited with signal 13
> > 
> > Also, when I try to run 'cpi' without mpirun, I get:
> > 
> >    [root@nereapc examples]# ./cpi
> >    p0_31794:  p4_error: init_p4_brdcst_info: my master indx bad: -1
> >        p4_error: latest msg from perror: No such file or directory
> 
> Yeah, the p4 hack is horribly broken.  It's somewhere on the list of
> things to fix.
> 
> - Erik
> 

Is there a status update on this?  *Offers to test*

I'm becoming more convinced by the day that taking parts of Scyld and
Clustermatic and brewing my own Beowulf is crack induced, and
Clustermatic is the only thing I know of that's sane.  Got to have a
working p4 device though.

Jack Neely


-- 
Jack Neely <sl...@qu...>
Linux Realm Kit Administration and Development 
PAMS Computer Operations at NC State University
GPG Fingerprint: 1917 5AC1 E828 9337 7AA4  EA6B 213B 765F 3B6A 5B89