When you are submitting a job on Cowboy, walltime will determine your job's time limit.
The "walltime" line in your script will look like: #PBS -l walltime=10:00
The 10:00 means 10 minutes. Thus, if you needed 10 hours, you'd change the line to read 10:00:00, and 100 hours would look like 100:00:00, etc.
It is recommended you try to most accurately predict your walltime with perhaps a few extra hours tagged on because:
Requesting too much (or the max) walltime can prevent your job from starting as it waits for an open node
Requesting too little (or underestimating) can result in your job being "killed" if it doesn't complete in time
If you've most accurately predicted your job's walltime, but it is still taking longer than expected, please email email@example.com ASAP** to request a walltime extension.
**Please be aware of when your jobs are scheduled to end. We do not mind helping you extend your walltime, but we also cannot predict if we will get the request in time if you send in an extension request when, for example, you only have an hour of walltime left.
Walltime & Queues
Walltime can also be affected by which queue you specify in your submission script. Below are descriptions of our four queues and maximum walltime limits for each:
batch: The 'batch' queue is the default queue. The walltime limit is 120 hours (120:00:00). If your job needs to run longer than this and your software does not have checkpoint/restart capabilities, please email firstname.lastname@example.org for assistance as far in advance of your need as possible.
express: The 'express' queue is for short jobs and debugging/testing scripts. The express queue contains 2 compute nodes and has a walltime limit of one hour (1:00:00).
bigmem: The 'bigmem' queue directs jobs to one of the two compute nodes that have 256 GB RAM and a NVIDIA Tesla C2075 GPU card. The walltime limit is 120 hours (120:00:00).
killable: The 'killable' queue is for long running jobs that are unable to use a checkpoint/restart feature. The walltime limit is 504 hours (504:00:00). Jobs in this queue are subject to being killed, at the discretion of HPCC administrators, for hardware and software issues. The killable queue uses the same nodes as the batch queue.
NEED HELP? email@example.com