[Pkg-nginx-maintainers] Bug#863602: nginx: Restart should check conf files before stopping service

[Christos Trochalakis]

On Fri, Nov 23, 2018 at 10:02:15AM +0100, Olaf van der Spek wrote:
>Op vr 23 nov. 2018 om 09:52 schreef Christos Trochalakis
><ctrochalakis at debian.org>:
>> Yes, restarting Nginx is something that can be avoided, Nginx provides
>> such utilities.
>
>> You can just run something like `nginx -t && nginx -s reload`, that
>> checks the config and the gracefully reloads nginx (no restart). If
>> nginx configuration is not valid the script will return a non-zero
>> exit status.
>>
>> >
>> >I think the conf files should be checked in restart before the old
>> >service is stopped.
>>
>> This is working in sysvinit systems but is not supported in systemd
>> systems: systemd doesn't give an entrypoint for a pre restart command.
>
>What about ExecStartPre?

It's already at `ExecStartPre=` [0], the problem is that in `systemd
restart nginx` ExecStartPre is executed **after** the service is
stopped, not in the start of the sequence.

[0] $ systemctl cat nginx.service|grep ^ExecStartPre=
>
>> ExecStartPre=, ExecStartPost=
>
>> Additional commands that are executed before or after the command in ExecStart=, respectively. Syntax is the same as for ExecStart=, except that multiple command lines are allowed and the commands are executed one after the other, serially.
>
>> If any of those commands (not prefixed with "-") fail, the rest are not executed and the unit is considered failed.
>>
>> >Perhaps even better, but I'm not sure if systemd and nginx supports
>> >this, would be for the new instance to be fully started before the old
>> >one is stopped.
>> >
>>
>> This can be done (in both sysvinit & systemd) systems by running
>> `/etc/init.d/nginx upgrade` but in your case a binary upgrade is not
>> needed, just a config reload (nginx -s reload).
>
>Could "service nginx restart" use this same logic?
>Users and packages probably don't know about nginx upgrade and will be
>using the standard interfaces...

No, service restart just calls systemctl restart on systemd systems.
I think that both nginx restart & upgrade are commands that are rarely
needed in my experience: `service nginx reload` is what you usually need
and luckily that works as expected on both systemd & sysvinit.