Can I increase walltime of a running job in HPC?

Question

[md.solayman]

I have submitted a job for performing genome alignment by PBS scripting. Now it is in running status. I found the job will not finish within the walltime I have requested.

Is it possible to increase the walltime for a running job?

Thanks.

[GenoMax]

Your cluster admins can add more time. When you are experimenting with new analysis tools it be best to start with long times and lower memory. Once a job completes you can modulate the time request for next one. If you ask for a ton of RAM generally those jobs tend to pend longer since most clusters give those a lower priority or may assign them to a queue that have nodes with large amount of RAM (which tend to be fewer than normal nodes).

Answer 1

[ATpoint]

Unfortunately not, at least not without admin rights as this would deny the purpose of the scheduling system. You should contact your local admin for support specific to your HPC. I actually recommend always talking to your cluster admin first when it comes to HPC problems because they know the system best and in fact it is their job to provide support.

[md.solayman]

Thanks a lot for your kind reply. It's really hard to predict the proper resource for me. After running several hours, my job will be terminated soon.

What is the best way to calculate the resource?

[ATpoint]

What is the best way to calculate the resource?

No gold standard here. Only experience helps. You can try to run the job on a small subset and then extrapolate, or simply try to book as much walltime as possible if you really have no idea how long it will take.