[Mauiusers] insufficient idle procs available ?

Itay M itaym.tau at gmail.com
Tue Jan 22 05:20:10 MST 2008


pbsnodes -a says:
state = busy
np = 4   --- which is correct, this machine has 4 processors, but at
the moment only 2 processrs (=jobs) are running on it. And this is
where I think the problem is - while the node should allow up to 4
procs to be used on it, it only utilizes 2 procs. The other 2 are
doing nothing.
And yes, this is consistent with the diagnose -n shows : each node
that has the (for example)  "WARNING:  node 'node17' has more
processors utilized than dedicated (4 > 2) " problem, also uses less
processors than it should at the moment.

How can I make sure the node allows to use it's maximum number of processors?

(btw: I'm using showres as an administrator.)



On 1/22/08, Jan Ploski <Jan.Ploski at offis.de> wrote:
> "Itay M" <itaym.tau at gmail.com> schrieb am 01/22/2008 12:36:42 PM:
>
> > I don't see any reservation, but perhaps I don't understand what the
> > rejection state means. Here the checkjob -v (shortened) output:
> >
> > /=====================/
> > State: Idle
> > ...
> > Req[0]  TaskCount: 1  Partition: ALL
> > Network: [NONE]  Memory >= 0  Disk >= 0  Swap >= 0
> > Opsys: [NONE]  Arch: [NONE]  Features: [NONE]
> > Exec:  ''  ExecSize: 0  ImageSize: 0
> > Dedicated Resources Per Task: PROCS: 1  MEM: 512M
> > NodeAccess: SHARED
> > NodeCount: 1
> > ...
> > job cannot run in partition DEFAULT (idle procs do not meet
> > requirements : 0 of 1 procs found)
> > idle procs:   4  feasible procs:   0
> > ...
> > Detailed Node Availability Information:
> > node01         rejected : State
> > node02         rejected : State
> > ...
> > node28         rejected : State
> >
> >
> > /=====================/
> >
> > What does the rejected: State means?
>
> According to the documentation, this means that the "node is not Idle or
> Running". What does pbsnodes -a say about the node states and job-node
> assignments? Is the information consistent with what diagnose -n shows?
>
> > showres shows as if all currently running jobs are 'reserved' . That
> > is, the number of  'Active Jobs' (using showq) equals to the number of
> > 'reservation located' (using showres) . The job numbers are also
> > identical in the two outputs. Is this how it is supposed to be?
>
> Yes, it is as it should be.
>
> By the way, showres run as a normal user may show fewer reservations than
> when you run it as administrator. However, had your job been rejected due
> to a standing reservation, the rejection reason would have been
> "Features", I think.
>
> Regards,
> Jan Ploski
>


More information about the mauiusers mailing list