Mailing List Archives
Authenticated access
|
|
|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] submitting LOTS of jobs
- Date: Thu, 12 Nov 2009 12:11:47 -0600 (CST)
- From: "R. Kent Wenger" <wenger@xxxxxxxxxxx>
- Subject: Re: [Condor-users] submitting LOTS of jobs
On Thu, 12 Nov 2009, Mag Gam wrote:
Do you have an example of a DAGMan job for a situation like this?
Lets say he would only like to have 10 jobs running and 10 jobs idle
while still trying to resolve 1000000000. Is that possible to do ?
Well 1000000000 separate jobs would be too big for a DAG. But you could
do, say, 1000000. You just make a big DAG that has all of the jobs (there
don't have to be any parent/child relationships in a DAG):
JOB Node0000000 job0000000.submit
JOB Node0000001 job0000001.submit
...
(You might be able to use VARS to re-use the same submit file if your
jobs are similar enough.)
Then you just say:
condor_submit_dag -maxjobs 10 -maxidle 10 <dag file>
Or if you don't want to have to remember to specify the maxjobs and
maxidle on the command line, you can put them in a per-DAG configuration
file:
large.config:
DAGMAN_MAX_JOBS_SUBMITTED = 10
DAGMAN_MAX_JOBS_IDLE = 10
large.dag:
CONFIG large.config
JOB ...
...
Then just do:
condor_submit_dag large.dag
If you really need to handle a workflow of 1G jobs, I guess you could
use nested DAGs -- have each sub-DAG handle 1M jobs, and have the
top-level DAG run 1k sub-DAGs, with maxjobs for the top-level DAG set to
1, and appropriate throttles on the sub-DAGs.
Kent Wenger
Condor Team