Hi,

Am 19.12.2012 um 18:13 schrieb Ian Kaufman:

> Well, that is what you asked for - setting it to YES means applying it per 
> slot per the man pages.
> 
> What kind of behavior are you really looking for? If set to JOB, then the 
> resource is debited from the queue resource, which in your case, has no limit 
> (h_vmem is set to INFINITY). So, since there is no queue limit set, SGE 
> ignores the host limits when set to JOB, and you can oversubscribe. 

I disagree here. The tighther limit should be take. Sometimes negative values 
show up for unknown reason.

If it should be consumed from a queue instance, it would be necessary to attach 
the it to complex_values also to the queue (this is per queue instance), but 
the limit further down in the queue definition is per job.

-- Reuti


> Ian
> 
> 
> On Wed, Dec 19, 2012 at 8:57 AM, Brett Taylor <[email protected]> wrote:
> Ok, that seems to be working, but it changes the submission time `-l h_vmem=` 
> to be per thread instead of per job.  Not a big deal, just a little annoying 
> to have to calculate each time.
> 
>  
> 
> Thanks
> 
>  
> 
> Brett Taylor
> 
> Systems Administrator
> 
> Center for Systems and Computational Biology
> 
> 
> The Wistar Institute
> 
> 3601 Spruce St.
> 
> Room 214
> 
> Philadelphia PA 19104
> 
> Tel: 215-495-6914
> 
> Sending me a large file? Use my secure dropbox:
> 
> https://cscb-filetransfer.wistar.upenn.edu/dropbox/[email protected]
> 
>  
> 
> From: Ian Kaufman [mailto:[email protected]] 
> Sent: Wednesday, December 19, 2012 11:37 AM
> To: Brett Taylor
> Cc: [email protected]
> Subject: Re: [gridengine users] vmem allocation
> 
>  
> 
> Hi Brett,
> 
> Try setting consumable to YES instead of JOB.
> 
> Ian
> 
>  
> 
> On Wed, Dec 19, 2012 at 8:05 AM, Brett Taylor <[email protected]> wrote:
> 
> Hello,
> 
>  
> 
> I seem to be having some issues with the h_vmem setting and proper 
> allocation.  I was under the impression that this would function much like 
> the CPU slots, and once all of the h_vmem had been allocated, that host would 
> not accept more jobs.  This however does not seem to be the case.  I have two 
> queues running on the same nodes. 
> 
>  
> 
> My hosts are defined:
> 
>  
> 
> complex_values        slots=48,h_vmem=95G
> 
>  
> 
> complex config:
> 
>  
> 
> #name               shortcut   type        relop requestable consumable 
> default  urgency
> 
> #----------------------------------------------------------------------------------------
> 
> h_vmem              h_vmem     MEMORY      <=    YES         JOB        4G    
>    0
> 
>  
> 
> queue config, in both queues:
> 
>  
> 
> slots                 24
> 
> h_vmem                INFINITY
> 
>  
> 
> So if I submit one job to A.q, using –l h_vmem=95G, then `qstat -f -u "*" -F 
> h_vmem` shows:
> 
>  
> 
> ---------------------------------------------------------------------------------
> 
> [email protected]       BIP   0/24/24        1.02     lx26-amd64
> 
>         hc:h_vmem=0.000
> 
>    3970 0.60500 bt_script. btaylor      r     12/17/2012 11:26:45    24
> 
> ---------------------------------------------------------------------------------
> 
> [email protected]        BP    0/0/24         1.02     lx26-amd64
> 
>         hc:h_vmem=0.000
> 
>  
> 
> but if I then submit to the 2nd queue, the h_vmem will go negative and allow 
> both jobs to run at the same time:
> 
>  
> 
> ---------------------------------------------------------------------------------
> 
> [email protected]       BIP   0/24/24        1.01     lx26-amd64
> 
>         hc:h_vmem=-95.000G
> 
>    3970 0.60500 bt_script. btaylor      r     12/17/2012 11:26:45    24
> 
> ---------------------------------------------------------------------------------
> 
> [email protected]        BP    0/24/24        1.01     lx26-amd64
> 
>         hc:h_vmem=-95.000G
> 
>    4012 0.60500 bt_script. btaylor      r     12/19/2012 11:03:04    24
> 
>  
> 
>  
> 
> Is this not supposed to act like the cpu slots?  Any ideas on how I might be 
> able to treat available vmem the same as the cpu slots?
> 
>  
> 
> Thanks,
> 
> Brett
> 
>  
> 
>  
> 
>  
> 
> Brett Taylor
> 
> Systems Administrator
> 
> Center for Systems and Computational Biology
> 
> 
> The Wistar Institute
> 
> 3601 Spruce St.
> 
> Room 214
> 
> Philadelphia PA 19104
> 
> Tel: 215-495-6914
> 
> Sending me a large file? Use my secure dropbox:
> 
> https://cscb-filetransfer.wistar.upenn.edu/dropbox/[email protected]
> 
>  
> 
> 
> _______________________________________________
> users mailing list
> [email protected]
> https://gridengine.org/mailman/listinfo/users
> 
> 
> 
> 
> -- 
> Ian Kaufman
> Research Systems Administrator
> UC San Diego, Jacobs School of Engineering ikaufman AT ucsd DOT edu
> 
>  
> -- 
> This email was Anti Virus checked by Astaro Security Gateway. 
> http://www.astaro.com
> 
> 
> 
> -- 
> Ian Kaufman
> Research Systems Administrator
> UC San Diego, Jacobs School of Engineering ikaufman AT ucsd DOT edu 
> _______________________________________________
> users mailing list
> [email protected]
> https://gridengine.org/mailman/listinfo/users


_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to