Signal handling

Using SIGINT to checkpoint and stop fluidity

If your Fortran compiler supports it, it is possible to send an operating system signal to your fluidity process which will cause it to checkpoint and then quit.

Do I have signal support?

You can check whether you have signal handing support in your version of fluidity by typing:

 fluidity -V

If you see the line:

 Signal handling support         yes

then you have signal handling support. If you see:

 Signal handling support         no

then your Fortran compiler doesn't support the signal function and signal handling won't work.

WARNING: If you do not have signal handling support then sending SIGINT to fluidity will cause it to crash without checkpointing!

Sending signals to fluidity

To signal a single processor fluidity job to checkpoint and quit use:

 kill -INT <pid>

where pid is the process number of the fluidity job.

For a parallel job under PBS use:

 qsig -s INT <job_identifier>

job_identifier is returned by qsub and may be found in the output of qstat.

If you use LSF (on Bull clusters, for example) then the command is:

 bkill -s INT <job_ID>