tell me your feature request about nginx

[Alexander Staubo]

On Fri, Jun 13, 2008 at 1:11 AM, Rt Ibmer <rtibmx at yahoo.com> wrote:
> We could REALLY use a way to get the status of upstream servers - in particular a notification that nginx has declared an upstream server (as used in load balancing) offline, and a notification that nginx has detected that a previously unavailable upstream server it took offline is now back on line.

Seconded. A status page such as HAProxy's (see
http://www.igvita.com/posts/05-08/haproxy-large.png) would be awesome.

More real-time metrics would be extremely handy. These could be used
to plug into tools like Munin.

You could look to Varnish (http://varnish.projects.linpro.no/) for
inspiration here. Instead of logging to files, Varnish uses
low-overhead circular buffers in shared memory to log and output
entries and metrics; this way you can catch the output any way you
like -- syslog, files, over the net. (I'm also pretty sure that
Varnish does not output any information unless someone is listening.)

Varnish comes with a few tools that let you track this output on the
terminal. For example, this is the output of "varnishstat", which is a
top-like tool that updates in real time. Suffice to say, it gives you
tons of useful data:

http://www.grabup.com/uploads/6c8b4ead957aa49ba626c51d36d9cd01.png

Another tool is varnishhist, which draws an ASCII diagram of the response times:

http://www.grabup.com/uploads/02cc7d92445c598d31efb86fa8452756.png

Other tools are varnishlog (gives you a real-time dump of the log) and
varnishtop (a top for log lines -- each unique log line is counted and
printed in order of frequency).

Alexander.