restarting supervisor fails

[JMLX42]

The current handler implementation fails "sometimes" because Ansible will not call:

service supervisor restart

but will call something like this instead:

service supervisor stop && service supervisor start

This will fail "sometimes" because supervisor programs (and supervisor itself) are not completely stopped when service supervisor stop returns. Then, when service supervisor start is called immediately, it will fail.

See this post: https://stackoverflow.com/questions/32738415/supervisor-fails-to-restart-half-of-the-time

It can be fixed by rewriting the restart handler this way:

- name: stop supervisor
  service:
    name: supervisor
    state: stopped
  listen: "restart supervisor"

- name: waiting for supervisor to be stopped
  wait_for:
    path: /var/run/supervisord.pid
    state: absent
    sleep: 5
  listen: "restart supervisor"

- shell: ps faux | grep supervisor

- name: start supervisor
  service:
    name: supervisor
    state: started
  listen: "restart supervisor"

Edit: the proposal above does not work as expected yet. I'll edit it when I can confirm it works.

Another solution might be to call service supervisor restart using the Ansible shell module.

[JMLX42]

FYI this is the implementation of restart:

restart)
    echo -n "Restarting $DESC: "
    start-stop-daemon --stop --quiet --oknodo --pidfile $PIDFILE
	[ -n "$DODTIME" ] && sleep $DODTIME
	start-stop-daemon --start --quiet --pidfile $PIDFILE \
		--startas $DAEMON -- $DAEMON_OPTS
	echo "$NAME."
	;;

As you can see, it calls sleep $DODTIME with $DODTIME set to 5 (seconds) by default.
So supervisor expects a 5 seconds timeout between start and stop.

[JMLX42]

I've updated my original proposal with the actual code to use.

Should I make a PR for this?

[JMLX42]

PR ready to be merged: #48

[JMLX42]

@tersmitten any chance this will be addressed?