Supervising jobs
You can determine job status by using the qstat
command, adding -u $USER
to display only your own jobs. Include -t
to expand any array jobs, and/or -rn
to display only jobs that are running, with a list of the
nodes on which they are running. See
man qstat
for more flags and information.
For best viewing of qstat
output, use a terminal window that is at least 94 characters wide. For a more visual display, try
.pbstop
Scheduling information
Maui's user commands are available to answer questions about job scheduling, e.g.
showstart 1234 |
When is job #1234 expected to start? |
checknode bo32 |
What reservations/schedules exist for node bo32 ? |
showbf -n 2 -f bora |
What is the walltime limit for a two-node job on Bora starting immediately? |
checkjob -v 1234 |
Why isn't job #1234 running? |
For specific node availability information, use the commands
freenodes
,pbstop
, orpbsnodes -l all
while logged in to any of the login servers for the cluster you're interested in. Note that TORQUE considers a node "free" (e.g. for pbsnodes -l free
) when there is at least one available core. Since users would often rather know how many nodes are completely idle, we provide a wrapper around pbsnodes
called freenodes
, which reports counts of nodes that are not running any job whatsoever.
Stopping/canceling a job
You can terminate/delete a job you submitted at any time by using the command:
qdel {job number}
Note that when naming a member of an array job you may need to escape (with a backslash) any brackets so they are not interpreted as part of a glob by the shell, e.g.
qdel 123456\[78\]