An example of how to write MCollective Agents to run in background.

$ mco rpc bg run_bg -I node -j
{"status": 0, "result": "2434"}

$ mco rpc bg status pid=2434 operation=run_bug -I node -j
{"status": null, "result": "running"}

$ mco rpc bg status pid=2434 operation=run_bug -I node -j
{"status": 0, "result": ""}

There's a somewhat harmless issue with this approach. Since the parent of the forked child process does not wait for the process to end, there is a good chance of the child process turning into a zombie process.

There is a fix for this problem which requires defining SIG_IGN as the default disposition for the SIGCHLD signal. This causes the Kernel to reap the zombie process. This can easily be done by using the following statement in Ruby: Signal.trap("SIGCHLD", "SIG_IGN").