[gridengine users] Long delay starting jobs, even when compute nodes are empty
dowobeha at gmail.com
Sat Mar 12 13:11:17 UTC 2011
On Fri, Mar 11, 2011 at 7:40 PM, Dave Love <d.love at liverpool.ac.uk> wrote:
> Lane Schwartz <dowobeha at gmail.com> writes:
>> I do know that more scheduling info can be printed out. There's
>> currently a job waiting in the queue that asked for mem_free=40960M,
>> and when I run qstat -j on that job, I get a bunch of output saying
>> that the job can't be run on various hosts because they don't offer
>> enough mem_free.
> If you're not getting sensible results with respect to vmem, you might
> be falling foul of the bug reporting values on 64-bit hosts, though I
> cant remember the symptoms. There was discussion about it on the old
> list (see gridengine.markmail.org) and one source of Rayson's patch for
> it that I have to hand is
I'm definitely running 64-bit hosts. Do you know if the bug you're
referring to is fixed in the current open source release?
>> But, for most of my queued jobs, scheduling info
>> only prints out the message about the disabled node.
> I don't know under what other circumstances it may happen, but one
> reason for nodes not being listed is that they're involved in a
I've got a pretty simple and straightforward setup. We only have a
handful of grid users, and we don't ever use reservations.
More information about the users