[Mauiusers] Node idle but load is HIGH

[Chris Samuel]

On Sat, 29 Sep 2007, Jan Ploski wrote:

> In particular, I am concerned about the 4:4 indications which I
> suppose is "class initializers". If these were inconsistent with
> the actual number of running jobs, then more than the configured
> number of jobs would be able to start in the given class (if I
> understand the concept correctly).

It is very odd that Torque doesn't think that there are jobs running 
there when there are!

Whilst Garrick's mention of max_load will help it would be useful to 
try and track down what's happening with the jobs.

The situations I've seen this happen with are when a PBS script forks 
a process into the background and then exits, or when an SSH/RSH 
based MPI launcher doesn't clean up nicely after itself (Google 
for "mpiexec" for a really nice replacement that uses the PBS TM 
interface instead).

Good luck!
Chris
-- 
Christopher Samuel - (03) 9925 4751 - Systems Manager
 The Victorian Partnership for Advanced Computing
 P.O. Box 201, Carlton South, VIC 3053, Australia
VPAC is a not-for-profit Registered Research Agency