Hello Kyle,

> I have encountered a cluster using the PBS/Torque scheduling system on
> which the qstat and tracejob commands give no information about job IDs 
> that have finished running. Instead the checkjob command must be used. This 
> prevents the PBS backend from determining which jobs have completed and 
> thus the Engine class is unable to transition completed jobs to the 
> TERMINATED state. I think new regex's will need to be written to parse the 
> checkjob output, as I think it is different from the tracejob output.

Thanks for reporting! I have already replied to the GC3Pie bug report --
but I thought I could provide (for future reference and for Google
search results) the set of questions that need to be answered for
implementing/extending batch-queuing system support in GC3Pie.

For implementing or extending batch-queueing system support in GC3Pie,
these info and outputs are needed:

* what command is used to submit a job? does it require a shell script
  or can it submit aribtrary (even binary) commands? what command-line
  option (or other mechanism) is used to specify that a process requires
  several CPUs, all on the same node?

* what command is used to check the queued/running/finished status of a
  job? if the job is finished, does this check command exit with a
  non-zero status? can you provide an example such output for each of
  the three statuses? (queued/running/finished)

* what command is used to check the exit status of a *finished* job? can
  you provide a sample output? how long after the job has finished does
  this information persist (i.e., it can be queried via the
  aforementioned command)?

* what command (if any) is used to check the resource usage of a
  *finished* job? (i.e., how much wall time did it consume, how much CPU
  time, etc.) can you provide a sample output? how long after the job
  has finished does this information persist (i.e., it can be queried
  via the aforementioned command)?

Ciao,
R

-- 
Riccardo Murri, Schwerzenbacherstrasse 2, CH-8606 Nänikon, Switzerland

-- 
You received this message because you are subscribed to the Google Groups 
"gc3pie" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to gc3pie+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to