cubicweb-condor #2591325 Adapt Condor for multiple queues [validation pending]

Current condor code uses condor_q command for inspecting the queue.

Problem is that the requesting server will only see the jobs in his queue.

When having several servers you can face following problems:

  • You will not see your jobs running. This can happen if a worker on another server launched the job.
  • We have in our Pylos cube a loop task that looks at the queue and at the running jobs. If there is no jobs in the queue while there are some jobs supposed to run, the loop task declares that "Something went wrong, probably in Condor".

A patch is provided:

Using "- global" allows to inspect all queues.

Two adaptations are necessary:

  • Reading the output of "condor_q - global" is slightly more complicated
  • When all queues are empty, "condor_q - global" returns an error code that we do not want to log. we used the trick of "ignore_output_errors" to bypass this issue.
priorityimportant
typebug
done in0.4.0
load3.000
load left0.000
closed by#0d131ec82cf8 [commands,views] condor_rm takes the schedd name as parameter (closes #2591325), #a1d43447c820 [commands] Allow to have more than one queue (closes #2591325)