3

I have a program, called "minvappmpi", which is parallelised and itself calls qsub as a subroutine.

When I use the program on the cluster, it throws the error 

qsub: command not found

If I want to submit another job using "qsub", then everything is fine. It is only when I call "qsub" inside my program that somehow the command is not found.

I think the issue is something to do with paths: somehow the environment I am working in knows where the executable "qsub" is sitting, but the cluster doesn't?

I'd like to know: firstly, how to make sure what the path is for the batch environment, and if it is different to the one for the login environment, and, secondly, if there is a way to change the path for the batch environment without root permissions?

Thanks so much for your help!

KR_Henninger
  • 61
  • 1
  • 5
  • 1
    For a guide to setting up "PATH", see https://unix.stackexchange.com/questions/26047/how-to-correctly-add-a-path-to-path . If that doesn't help, you'll need to expand on "When I submit a job using the program on the cluster" - how do you submit this job? `at`, `cron`, something custom to the machine in question? – JigglyNaga Sep 07 '16 at 12:03
  • Thanks for the reply! Sorry to have been vague...I meant submit to an SGE cluster (hence qsub). I think the cluster is managed via Torque, if that is useful to know... – KR_Henninger Sep 14 '16 at 12:39
  • Oh, is `qsub` the command that submits a job? I can't follow which program is calling (or trying to call) `qsub`, and what program `qsub` itself is submitting. Does "submit a job using the program" mean "use the program to submit a job", or "submit (a job that uses the program)"? – JigglyNaga Sep 14 '16 at 13:27
  • Yes! Sorry...I guess I was a bit vague earlier :-) I have re-edited the question to make it more clear! – KR_Henninger Sep 16 '16 at 07:57
  • If `qsub` is an executable on the system, how does minvappmpi call it "as a subroutine"? Does it use eg. the [system()](http://linux.die.net/man/3/system) function, as in `system("qsub")`? – JigglyNaga Sep 16 '16 at 11:28

1 Answers1

3

Okay! I have made progress!

I tried submitting my job with

qsub -V run.sh

As I learned that this exports all the variables in qsub's command environment to the batch environment. This worked great, and now the batch environment knows where "qsub" is :-).

But I have a new problem.

I need to submit jobs on the cluster with 

qsub -P Projectname

And my program doesn't know that! It just calls "qsub".

Is there any way I can get around this without having to recompile? I don't want to have to recompile because I'm running the program remotely and I don't have all the right libraries on the remote machine (and can't install them without root permissions :-))

Thanks so much for your help!

KR_Henninger
  • 61
  • 1
  • 5
  • It sounds like `qsub -V` is the correct fix for your original problem, so you should end your answer after that. The second part, about running with extra command-line options, is a new question. – JigglyNaga Sep 16 '16 at 11:55