Behavior of preempted jobs?

[jamesgao]

Hi, I stumbled upon this repository while searching for queuing solutions in Openstack. In the presentation found here, https://agenda.infn.it/getFile.py/access?contribId=17&sessionId=3&resId=0&materialId=slides&confId=7915, it says that when resources aren't available, older instances are terminated to make room for new instances.

Where in the code does this termination occur? Is it possible to suspend the instance and add it back into the queue for later processing?

[zangrand]

Hi James,

you are referring to a very old presentation and in such slides we shown
the instance termination explicitly done by hand. The fair-share
scheduler prototype is evolving and it will be in the next future an
OpenStack service. The new version will be able to terminate the extra
quota VMs, but the current one is not able to do it.
Thanks a lot.
Cheers,
Lisa

[edwins]

Is there a name for that project? I am highly interested in this scheduler (or something like it). I would be happy to contribute to this effort if it has begun work. Or, should we make contributions to this repository?

Thanks,
Ed

[jamesgao]

Thanks for the update! This is a critical feature for us and I can spare programmers to help get it up and running. How can I contribute to this effort?

[zangrand]

Hi Edwin,

thanks a lot for your interest on our work.
At the moment the repository is not public but it will become public as
soon as we produce the first stable release.
About the contributions, we are planing to open a position for
developers and if you or someone of your team would like to work in
Italy and contribute on the evolution of our scheduler, please let me
know in advance.

Thanks a lot.
Cheers,
Lisa

[zangrand]

Hi James,

we are planing to open a position for developers and if you or someone
of your team would like to work in Italy and contribute on the evolution
of our scheduler, please let me know in advance.
Thanks a lot for your interest.
Cheers,
Lisa