Re: [Condor-users] In Dynamic slot to match a new job to slot1.2 need condor_negotiator

[Mag Gam <magawake@xxxxxxxxx>]

The users can wait ;-)



On Tue, Jan 12, 2010 at 9:47 AM, Matthew Farrellee <matt@xxxxxxxxxx> wrote:
> On 01/12/2010 08:10 AM, Mag Gam wrote:
>> This is in reference of
>> https://lists.cs.wisc.edu/archive/condor-users/2009-January/msg00034.shtml
>>
>> I have all 16 core boxes with 64gig of memory and running dynamic
>> slots and the problem I have is  when I submit a job like this:
>>
>> request_cpus = 1
>> request_memory = 1024
>> request_disk = 10240
>>
>>
>> I have to constantly run condor_reschedule to have the jobs executed.
>> My default negotiator cycle is 60 seconds therefore I have to have
>> 60*number of jobs to be pushed through.
>>
>> I tried placing this
>> NEGOTIATOR_IGNORE_USER_PRIORITIES = TRUE
>> NEGOTIATOR_MATCHLIST_CACHING = FALSE
>>
>> but still no luck. Any better way to do this with dynamic slots?
>>
>> Version 7.2.4
>
> You're only going to get one partition per negotiation cycle. Depending on the size of your pool you could push the cycle interval down below 60.
>
> Best,
>
>
> matt
>