RE: [Condor-users] To set dynamic limit on the jobs on condor

[Prashant Lal <lalp@xxxxxxxxxxx>]

Hi Mike

NO. I don't have any misconceptions.

Let's put it like this 

I have got 2 dual cpu's machines.

opteron1 and opteron2

have 10 jobs both for compile and run (so it becomes 20 execution for the condor)

Now compilation part:

4 then 4 and then 2 jobs are compiled. 

now when it comes to run iwant it to submit the job (for the run part ) only on opteron2
so it will be like 2 then 2 then 2 then 2 then 2 on opteron2 ( I hope I am clear here)

Now when the opteron2 is having the initial 2 (run jobs) jobs, what the reaming 8 are doing.Obviously they don't get executed on the cvainopt1 as i have made the script like that.

So my problem is does those remaining 8 jobs will keep on asking the cpu's of opteron2 for the availability so that the job no. (run job) 3 and 4 can be submitted on it. And if it does then how much time does it waste for the same.

I hope I have made my self clear.

On Wed, 2005-08-31 at 00:10, Michael Yoder wrote:
> > Hi Nick and others
> > 
> > Can you help in the following scenario.
> > 
> > I have got 20 jobs. All the 20 jobs are to be "compiled" and "run"
> there
> > fore total no. of execution are 40.
> > 
> > Noe run part has to be executed after the completion of the
> compilation
> > part i.e. after the 20 compilation, 20 run should start.
> > 
> > Now the cluster is having say 5 cpus
> > 
> > so at the compilation time no problem as it will take 4 submission to
> > compile the 20 jobs.
> >
> > Now here what I want is when it comes to the submission of 20 run
> jobs,
> > out of which say 5 has been submitted to the 5 cpu's and the next 5
> jobs
> > of the remaining 15 will keep on pinging/asking  the cpus to check
> whether
> > it is free or not to submit the next lot of 5 jobs.
>
> I think you have some misconceptions about how Condor works.  You submit
> jobs into the scheduling system (the condor_schedd in the condor lingo)
> and let it schedule the jobs where and how it wants.  You shouldn't be
> concerned about what cpus handle what jobs.  You may want to have a look
> at the "User's Guide" section of the Condor manual: 
>
> http://www.cs.wisc.edu/condor/manual/v6.7/2_Users_Manual.html
>
> What you really want to do is to use DAGMan, condor's job dependency
> manager.  See
>
> http://www.cs.wisc.edu/condor/manual/v6.7/2_12DAGMan_Applications.html
>
> You'll create 40 condor jobs, and have one DAG manage them.  You can set
> it up so that the first 20 jobs (the compile jobs) run to completion
> before the next 20 (the run phase) start off.  Alternately, you could
> have the DAG specify the "parent-child" relationship between each
> compile-run pair.
>
> Hope this helps, 
>
> Mike Yoder
> Principal Member of Technical Staff
> Ask Mike: http://docs.optena.com
> Direct  : +1.408.321.9000
> Fax     : +1.408.321.9030
> Mobile  : +1.408.497.7597
> yoderm@xxxxxxxxxx
>
> Optena Corporation
> 2860 Zanker Road, Suite 201
> San Jose, CA 95134
> http://www.optena.com
>
>
> _______________________________________________
> Condor-users mailing list
> Condor-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users

Thanks and Regards

P r a s h a n t  L a l

Cadence Design Systems

Noida Export Processing Zone,

Noida - 201301,

Phone:+91 120 2562842, extn 4009

Fax:+91 120 2562231

Cell:+91 98101-44168

mailto: lalp@ cadence.com