[Mauiusers] Test of Maui causes core dumps

[Wightman]

Maui should never core dump even if there is an incompatibility with a
resource manager.  Can you run maui under gdb with the environment
variable MAUIDEBUG=yes?  Send the output of "where" when the program
crashes.

- Douglas

On Tue, 2006-04-04 at 17:13 -0500, Michael Homa wrote:
> Hi:
> 
> I'm running torque 1.20p3 under RH 9.0 on our 64-node cluster. The
> production scheduler is PBS sched. I'm trying to replace that with Maui
> (3.26p14). I created the Maui executables without problem and have the
> basic cfg file. I configured maui to run in test mode for evaluation
> purposes. Each time I start Maui, messages are written to the log and
> then the Maui scheduler core dumps. Edited version of the log:
> 
> 04/04 16:54:17 INFO:     starting Maui version  ##################
> 04/04 16:54:17 MAMSetDefaults()
> 04/04 16:54:17 ServerProcessArgs(1,ArgV,0)
> 04/04 16:54:17 starting  version Maui (PID: 24030) on Tue Apr  4 16:54:17
> 04/04 16:54:17 MPBSInitialize(ARGO-NEW,SC)
> 04/04 16:54:18 WARNING:  cannot bind to port 15004, errno: 98 (Address
> already in use)
> 04/04 16:54:18 WARNING:  cannot connect to PBS scheduler port 15004
> 04/04 16:54:18 __MPBSSystemQuery(ARGO-NEW,RCount,SC)
> 04/04 16:54:18 INFO:     connected to PBS server :0 on sd 1
> 04/04 16:54:18 INFO:     XRMInitialize not supported
> 04/04 16:54:18 MPBSClusterQuery(ARGO-NEW,RCount,SC)
> 04/04 16:54:18 __MPBSGetNodeState(Name,State,PNode)
> 04/04 16:54:18 INFO:     PBS node argo4-4 set to state Idle (free)
> 04/04 16:54:18 INFO:     node slot not set on node 'argo4-4'
> ...
> 04/04 16:54:18 INFO:     resources detected: 64
> 04/04 16:54:18 MPBSWorkloadQuery(ARGO-NEW,JCount,SC)
> 
> The messages stop at the PBSWorkloadQuery. I would expect Maui shouldn't
> be able to talk to PBS via 15004 because that port is already in use as
> part of the production environment.
> 
> The only thing I can think of is my old version of torque is not compatable
> with the more recent Maui?
> 
> Can anyone make any suggestions or point me in the right direction. I've
> persused the logs and nothing seems to jump out. Thanks.
> 
> Michael Homa
> Operating Systems Support and Database Group
> Academic Computing and Communication Center
> University of Illinois at Chicago
> email:  mhoma at uic.edu
> 
> _______________________________________________
> mauiusers mailing list
> mauiusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/mauiusers