On Wed, Oct 08, 2014 at 06:16:16AM -0700, Tyler Riddle wrote:
I think that gracefully stopping nginx is a better default the forcibly
stopping the process.
I certainly don't - both cases have uses. Please do not force sysadmins
to wait for end user behavior *by default*. When I want end users to be
able to influence my reboot time I'll specify a specific command. The
default should not be to extend reboot time defined by the activity of
people external to the system and this is what you have done by
converting to graceful shutdown by default.
I don't think that reboot times are an issue for a web server (at least
by default). And, for me, it seems that not killing active request, and
giving nginx a few seconds to handle things gracefully is a better
default behavior. We can switch to 5s if 10secs seems like a lot of time
(it might be).
Those 5 or 10 seconds are a capped, well defined, period. External
users can't extend your stop time beyond that time, so this is not a
threat to the system.
This strategy has been made configurable in a /etc/default/ setting, so
it can be easily tweaked to match anyone's needs.
You've reduced the functionality of this patch and decreased the
utility of init. Why are you attempting to optimize for reduced command
count?
Why we should introduce a new command when you can simply run
`nginx -s stop` or `nginx -s quit`?
Also, jessie ships with systemd by default, and systemd doesn't support
custom commands. So, it's not a good plan to divert the initscript from
the service file.
The correct thing to do, for both systemd and sysvinit users, is to keep
the initscript plain simple, and introduce something like a nginxctl
script that handles things like upgrading the nginx executable, and
other complex things that an administrator might need.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
