How to monitor open file handles in a process

OK, that seems to work. I still need to figure out how to get the number of file handles for a specific process. The settings in that article provide the load average per cpu.

I _did_ say selection would be a problem ;-)

I don't have an HP-UX system with Glance, but if you can select a single process (-s Process_ID ?)  or list all open files for processes (-F ?)  from the command line and post the output, then (we) can help in interpreting the output using a script.

I found a better answer: If I use lsof -p <PID> it shows all the open files/sockets. I pipe into wc and it tells me the number of open file handles.
Thanks for the suggestions on glance. I think I can use that for other tasks.
tfewster: I will still give you the points for the good direction to research.

Also, glance has example scripts in /opt/perf/examples/adviser including a file handle monitor

Note that you can limit the number of file descriptors that a process can open, using the built-in unix command 'ulimit':

   ulimit -n 12       # will limit the number of *available* file descriptors to 12 + 1

If the process attempts to allocate an FD beyond this number, the behaviour is determined by the code itself, or if no provision is made by the code, the process will be issued a HUP or KILL signal, or be coredumped depending on the unix flavour.

The effect being that your process is stopped when it reaches a limit you have set.

The advantage of this method is that you can run a script that sets ulimit, then executes the underlying process in the FOREGROUND, but in a never ending loop.  When the process dies, the loop will cycle, and you can do some quick maintenance, and then restart the process.  No delays are necessary while you wait for a cron job to kick in.

Note that the semantics of ulimit are slightly different by flavour of unix, so you should test this technique thoroughly.

Actually, the reason I need to monitor is that I have a file handle leak that I cannot fix (in 3rd party code) and I need to be prepared to restart the process when the number available is getting low. The OS kernel is configured for 8000 per process. The process is already under a management system that restarts it, but I prefer to do it gracefully rather than waiting for errors.

Glad to hear you resolved the problem yourself; i suggest you post a (0 point) question in http://www.experts-exchange.com/Community_Support/General/ with the link to this question, asking them to accept your solution at
http://www.experts-exchange.com/Q_22551088.html#a19032800
as the answer and refund your points