Re: Maintenance mode

From: Alexander Staubo <alex#bengler.no>
Date: Thu, 11 Sep 2008 00:36:37 +0200


On Thu, Sep 11, 2008 at 12:11 AM, Aleksandar Lazic <al-haproxy#none.at> wrote:
> Please can you add a little bit more detail about your setup, due the
> fact that you have started with a nginx example and I'am not sure if you
> still use nginx or not?
>
> client -> ? -> haproxy -> mongrel -> rails

Actually the issue of maintenance mode is completely orthogonal to whatever HAProxy is proxying. We have Mongrel behind HAProxy, but it could be anything. (Sometimes it's Varnish or Nginx, too.)

> You mean all of your mongrel servers was down, sounds a very bad day,
> get my sorry ;-(

It happens with surprising frequency. Whatever you do, don't ever outsource server management to a professional hosting company. :-)

> ###
> errorloc 502 /maintence.http
> errorloc 503 /error.http
> acl error_state status 503
> acl site_dead nbsrv(dynamic) lt 1
> acl site_dead nbsrv(static) lt 1
>
> block if error_state site_dead
> ###

Would this not accidentally trigger a maintenance page if all backends are suddenly down or unresponsive? That's my point -- a maintenance page should not be shown on errors! On the contrary, bringing up a maintenance page is inherently a *manual* operation; whatever the solution, the person bringing down the site has to tell HAProxy explicitly. HAProxy does not possess psychic powers (that I know of).

And a maintenance page should clearly return 503, nor 502, by the way.)

Alexander. Received on 2008/09/11 00:36

This archive was generated by hypermail 2.2.0 : 2008/09/11 00:45 CEST