Command reference for Condor
We define advanced research computing to mean harnessing computing power beyond that of a typical desktop or laptop computer to support research computing across a wide range of disciplines. There are two particular systems that we run to enable research computing and the commands listed below describe those used in Condor.
Condor harnesses the spare computing capacity of Windows workstations in Schools and libraries across campus. Condor is an ideal system to use if your program currently runs on your desktop computer and can reduce the time taken to run your jobs from days to hours.
- condor_status - Displays the status of machines in the Condor pool
- condor_q - Displays the status of jobs on the Condor pool
- condor_submit - Queues one or more jobs to run on the Condor pool
- condor_rm - Removes one or more jobs from the queue
Calling "condor_status -totals" displays the status of the Condor pool.
- Total = The number of processor cores currently in the pool
- Owner = The number of processor cores in use by people running their own applications
- Claimed = The number of processor cores in use by Condor running researchers applications
- Unclaimed = The number of processor cores not in use by people or Condor
Calling "condor_q" displays the status of jobs on the Condor pool submitted by you.
Calling "condor_q -global" displays the status of jobs on the Condor pool submitted by all users.
- ID = The cluster id of the job
- OWNER = The owner id of the job
- SUBMITTED = M/D HH:MM the job was submitted
- RUN_TIME = D+HH:MM:SS the job has run
- ST = Status of the job (H = Held, R = Running, I = Idle)
- PRI = The user specified priority of the job
- SIZE = The size of the job
- CMD = The name of the executable
Calling "condor_submit submit.sub" queues the jobs defined in the submit.sub script.
An example submit script looks similar to this:
universe = vanilla
submitdir = D:\Current\Working\Directory
output = job-out.txt
error = job-error.txt
log = job-log.txt
executable = $(submitdir)\MyProgram.exe
transfer_input_files = $(submitdir)\MyData.dat
use the vanilla universe (this is the only universe available on windows)
define a variable called submitdir that is the full path to the files you are working with
send any output from the executable to a file called job-out.txt
send any error messages from the executable to a file called job-err.txt
send the condor job log information to a file called job-log.txt
set the executable to be MyProgram.exe (which is in the submitdir directory)
send the file MyData.dat (which is in the submitdir directory) that is to be used by MyProgram.exe
you have to list the files to transfer because Condor does not have a shared file system
queue one job
Calling "condor_rm 1" removes job 1 from the queue.
Calling "condor_rm -all" removes all of your jobs from the queue.