service_autorestart should perform a configtest before restarting

[ross-w]

Right now if service_autorestart is true (which is the default), but nginx encounters an issue with the configuration upon restarting, nginx will shut down and stay shut down.

If instead, a configtest could be performed and only notify the nginx service on passing configtest, this would ensure that nginx keeps running on the old config until fixed.

Challenge will be to somehow make it fail /every/ time not just for one puppet run.

[sammcj]

👍

This is an issue for me too, we managed to cause a small production outage because of this.

[sammcj]

Just wondering if any progress has been made with this bug? It's caused a few outages now.

[javierbertoli]

Hi @sammcj and @ross-w. I'm checking on this issue, and both in CentOS 6, Debian 7 issue a configtest in the /etc/init.d/nginx script for both the restart and reload functions. But as the hasrestart in the service in this module defaults to false, pupet is issuing stop/start instead of a real restart. And start, in the init script does not have that check in place.

Will add a restart parameter set to true, I think this will fix this issue.

[sammcj]

Thanks @javierbertoli , that sounds like it would fix it.