Question about http_stub_status

[Maxim Dounin]

Hello!

On Fri, Aug 22, 2008 at 01:51:12PM +0200, Marcus Bianchy wrote:

>we are using the "Stub Status Module" now for a couple of days, and
>there is a something strange effect with the values from this module:
>
>The first few hours the values for  "Active connections" and "Waiting"
>were reasonable and low (some few hundred).
>After a while, both of these values suddenly increased to ~5000, then
>few hours later to ~20000, then to ~30000.
>It doesn't matter if we are rotating our logfiles (with SIGHUP): These
>values are still high around ~30000.

JFYI: you should rotate logfiles with SIGUSR1, not with SIGHUP.  With 
SIGHUP nginx re-configures itself, while with SIGUSR1 it does exactly one 
thing needed to rotate logfiles: it reopens logs.

>For what are these values standing for? What is happening there? BTW:
>netstat et al. says some hundred or few thousand concurrent (i.e.
>ESTABLISHED) connections (and some FIN_WAIT etc. too)
>
>Maybe there is one who can explain this behavior?

Sudden changes in stub_status numbers usually coresponds to died or 
killed nginx worker.  Take a look at logs.

Maxim Dounin