I noticed that when the bug hits the pid of nginx is always the same (I monitored it using `for x in {1..10000}; do cat /var/snap/maas/current/http/nginx.pid; sleep 0.5; done`).
On the other side, when it works properly the pid changes.
Looking at http://nginx.org/en/docs/control.html , in our case it's enough to send a HUP signal to reload the configuration.
Still we can't really understand what happened, but it must be a nasty race with NGINX and pebble (like https://trac.nginx.org/nginx/ticket/424).
I noticed that when the bug hits the pid of nginx is always the same (I monitored it using `for x in {1..10000}; do cat /var/snap/ maas/current/ http/nginx. pid; sleep 0.5; done`).
On the other side, when it works properly the pid changes.
Looking at http:// nginx.org/ en/docs/ control. html , in our case it's enough to send a HUP signal to reload the configuration.
Still we can't really understand what happened, but it must be a nasty race with NGINX and pebble (like https:/ /trac.nginx. org/nginx/ ticket/ 424).