[SCore-users] Score limitations?
Amik St-Cyr CFD Lab
amik at cfdlab.mcgill.ca
Wed Sep 18 07:20:35 JST 2002
Hi Gents,
Something strange is happening with our new
(not yet fully functional) Beowulf.
We have a 128 nodes(dual) cluster connected with both a myrinet2k
switch and an ethernet 100MB network.
The only problem (yet) is that when I try a very simple
program the scrun/mpirun hangs forever when the number of
nodes used is greater then 64.
Example:
--------
| amik at stokes 17:49:23 amik> scout -g cn -n 64
SCOUT: Spawning done.
SCOUT: session started.
| amik at stokes 18:09:08 amik> scrun -nodes=64x1 ./myname
SCore-D 5.0.1 connected.
<0:0> SCORE: 64 nodes (64x1) ready.
My name is cn13.clumeq.mcgill.ca
...etc.
WORKS FINE
then:
| amik at stokes 18:09:30 amik> scrun -nodes=64x2 ./myname
SCore-D 5.0.1 connected.
<0:0> SCORE: 128 nodes (64x2) ready.
My name is cn1.clumeq.mcgill.ca
...etc.
ALSO WORKS FINE
Now booting with 65 nodes:
| amik at stokes 18:09:41 amik> exit
SCOUT: Session done.
| amik at stokes 18:11:09 amik> scout -g cn -n 65
SCOUT: Spawning done.
SCOUT: session started.
| amik at stokes 18:11:23 amik> scrun -nodes=65x1 ./myname
Hangs forever...
| amik at stokes 18:11:52 amik> scrun -nodes=65x2 ./myname
Hangs forever...
| amik at stokes 18:12:39 amik> scrun -nodes=64x2 ./myname
Hangs forever...
| amik at stokes 18:13:18 amik> scrun -nodes=64x1 ./myname
Hangs forever...
It seems that when we scout with 65 nodes even the 64 nodes
case does not work.
Do you folks think that:
1)the problem is myrinet2k related ?
2)the driver for the myrinet switch has a problem ?
we are using:
/* PM/Myrinet */
myrinet2k type=myrinet2k \
-firmware:file=/opt/score/share/lanai/lanaiM2k.mcp \
-config:file=/opt/score/etc/pm-myrinet.conf
3) Bad connections for the myrinet
4) faulty configuration of node 65?
5) Score is the culprit?(maybe because of a misconfiguration)
Best regards,
Amik St-Cyr
--
_____________________________________________________
Dr. A. St-Cyr
Research Associate, CFD Lab
Department of Mechanical Engineering
McGill University
688 Sherbrooke Street West, 7th floor
Montreal, Qc, Canada H3A 2S6
Tel: +1 (514) 398-1710, Admin. Fax : 2203
amik at cfdlab.mcgill.ca
_____________________________________________________
More information about the SCore-users
mailing list