Feature/cr queue

[Sanger-ad7]

Changed the queue selection when blcr checkpointing is set so that the default is set to the normal queue

[jrandall]

The intention here is that the queue selection should have nothing to do with the run length of the job when using checkpointing. One of the main features of checkpointing is that you can submit jobs in a normal queue and then have it checkpoint and restart again in the same queue to keep running. When this feature is desired, it makes no sense to calculate the expected runtime of a job and submit it to a longer (e.g. long or basement) queue.

[pd3]

Checkpointing is costly and error prone operation and it should be possible to reschedule failing jobs in a queue with longer time limit. This would not be possible if we accepted this pull request.
Note that one can enforce the behavior you requested even with the existing code by overriding the default queues set_limits(queue_limits=>{single_queue=>time}).

[jrandall]

Ok, if we document the way to override the list of queues that sounds like
it should be fine!

On Mon, May 19, 2014 at 4:56 PM, pd3 notifications@github.com wrote:

Checkpointing is costly and error prone operation and it should be
possible to reschedule failing jobs in a queue with longer time limit. This
would not be possible if we accepted this pull request.
Note that one can enforce the behavior you requested even with the
existing code by overriding the default queues
set_limits(queue_limits=>{single_queue=>time}).

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/49#issuecomment-43522236
.