DTLSPAWN-Step stays alive but should terminate

[Ulrich Schmidt]

Hello,
this is probably a very basic question, but I don't know how to deal with it right now.
We are running a job consisting of two steps. The first one starts some daemons and at last it runs my REXX and then does a exit - at least the protocol I get from "set -x" says so. But the process still stays alive.
Do you have any hints about this?
brgds,
Ulrich Schmidt

[Ulrich Schmidt]

And to say more: If I goback from DTLSPAWN to BPXBATCH as a Launcher for this sequence it works fine.
brgds,
Ulrich Schmidt

[dovetail]

Ulrich,

What version of COZBATCH (formerly DTLSPAWN) are you using?

Can you run with '-LT' and post or email the trace to support@dovetail.com?

Also, if you can provide us with a small test case that reproduces the problem, that would help us.

Thanks

[Ulrich Schmidt]

Have sent the requested data.
Ulrich

[dovetail]

Ulrich,

Thanks for sending the trace data; sorry that it took me a while to look at it carefully.

What I believe is happening is that COZBATCH is starting your shell (/bin/sh) as a child process and then it redirects its stdin, stdout, and stderr to pipes. COZBATCH then has a select() loop that does three things at once:

1) reads any records from DD:STDIN and writes them to the stdin pipe
2) reads any data from the stdout pipe and writes to DD:STDOUT
3) reads any data from the stderr pipe and writes to DD:STDERR

This continues until all three of the pipes to the shell have been closed (EOF).

From the trace that you sent me, I can see that COZBATCH is waiting for more data from stdout and stderr, but it never gets and EOF (pipe closed).

I assume that what is happening is that your shell forks a daemon (child) process (how is this done? with nohup? or just &? ). But when the shell does "exit 0", the daemon process is still running and has duplicate handles to the stdout/stderr pipes, which is why they never close.

Under Posix standards, a child process has a life of its own, and doesn't necessarily terminate when its parent terminates. So I'm not sure what COZBATCH *should* do in this case, since there is still a running process that is writing output to stdout->DD:STDOUT and stderr->DD:STDERR.

What happens if you fork your daemon child process, but redirect both its stdout and stderr to files (or /dev/null)? I would assume that this would allow the pipes in the main shell process to close and the job still will terminate.

[Ulrich Schmidt]

Thank you for your answer. It let's me understand what ist happening there. And your hint, to redirect the stdout and stderr solved the problem.
Thank you again.
brgds,
Ulrich Schmidt