Hello Geert,
I don't believe any such functionality is available out of the box,
but I am confident clever use of dynamic job runners
(http://lists.bx.psu.edu/pipermail/galaxy-dev/2012-June/010080.html)
could solve this problem.
One approach would be to maybe move all of your job runners out of
galaxy:tool_runners maybe to a new section called
galaxy:tool_runners_local and then create another set of runners for
your HPC resource (maybe galaxy:tool_runners_hpc).
Next set your default_cluster_job_runner to
dynamic:///python/default_runner and create a python function called
default_runner in lib/galaxy/jobs/rules/200_runners.py.
The outline of this file might be something like this:
from ConfigParser import ConfigParser
def default_runner(tool_id):
runner = None
if _local_queue_busy():
runner = _get_runner("galaxy:tool_runners_local", tool_id)
else:
runner = _get_runner("galaxy:tool_runners_hpc", tool_id)
if not runner:
runner = "local://" # Or whatever default behavior you want.
return runner
def _local_queue_busy():
# TODO: check local queue, would need to know more...
def _get_runner(runner_section, tool_id):
universe_config_file = "universe_wsgi.ini"
parser = ConfigParser()
parser.read(universe_config_file)
job_runner = None
if parser.has_option(runner_section, tool_id):
job_runner = parser.get(runner_section, tool_id)
return job_runner
You could tweak the logic here to do stuff like only submit certain
kinds of jobs to the HPC resource or specify different default runners
for each location.
Hopefully this is helpful. If you want more help defining this file I
could fill in the details if I knew more precisely what behavior you
wanted for each queue and what the command line to determine if the
dedicated Galaxy resource is busy (or maybe just what queue manager
you are using if any).
Let me know if you go ahead and get this working, I am eager to hear
success stories.
-John
------------------------------------------------
John Chilton
Senior Software Developer
University of Minnesota Supercomputing Institute
Office: 612-625-0917
Cell: 612-226-9223
Bitbucket: https://bitbucket.org/jmchilton
Github: https://github.com/jmchilton
Web: http://jmchilton.net
On Mon, Sep 24, 2012 at 3:55 AM, Geert Vandeweyer
<geert.vandewey...@ua.ac.be> wrote:
> Hi,
>
> The admin pages state that it is possible to specify multiple clusters in
> the universe file. Currently, we are investigating if we can couple the
> university HPC platform to galaxy, to handle usage peaks. It would be ideal
> if the job manager would check the load of the dedicated cluster (eg queue
> length) and send jobs to the second cluster when load is above a threshold.
>
> Does such an approach exists already, or will it become available in the
> near future? As far as I understand, it is now only possible to specify
> which jobs run on which cluster, without dynamic switching?
>
> Best regards,
>
> Geert
>
> --
>
> Geert Vandeweyer, Ph.D.
> Department of Medical Genetics
> University of Antwerp
> Prins Boudewijnlaan 43
> 2650 Edegem
> Belgium
> Tel: +32 (0)3 275 97 56
> E-mail: geert.vandewe...@ua.ac.be
> http://ua.ac.be/cognitivegenetics
> http://www.linkedin.com/pub/geert-vandeweyer/26/457/726
>
> ___________________________________________________________
> Please keep all replies on the list by using "reply all"
> in your mail client. To manage your subscriptions to this
> and other Galaxy lists, please use the interface at:
>
> http://lists.bx.psu.edu/
___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client. To manage your subscriptions to this
and other Galaxy lists, please use the interface at:
http://lists.bx.psu.edu/
