HttpUpstreamModule: Need more detailed Information

Maxim Dounin mdounin at
Fri Aug 10 13:45:36 UTC 2012


On Fri, Aug 10, 2012 at 01:36:46PM +0200, Axel wrote:

> Hello all,
> i'm new to nginx and first of all i have to say it's a great piece
> of software.
> I need some more detailed Information about nginx behaviour
> regarding HttpUpstreamModule and I hope you can give me some hints
> and links where i can learn more about it.
> I set up nginx/1.2.2 as reverse proxy in front of a bunch of apache
> server(*) which are located in different housing locations.
> Now I have some questions and I can't find any docs or wiki pages
> with detailed answers.
> - how does nginx detect if one ore more upstream servers has diappeared?
>   - what kind of mechanism does nginx use? icmp or something else?

It detects based on status of requests to upstream servers.  If 
requests fail - the server is considered down and additional 
requests aren't routed to it for some time.

>   - how often does nginx request the status of upstream servers?

For alive servers - as often as normal requests are routed to the 
servers.  For servers already considered down - once per 
fail_timeout (per worker, see below).

> - how can I monitor the status of upstream server seen by nginx (I
> monitor the status of running apache prcesses on the upstream server
> separately)

Currently, there is no way.  Moreover, each nginx worker process 
has it's own idea about status of upstream servers.

Maxim Dounin

More information about the nginx mailing list