> -----Ursprüngliche Nachricht----- > Von: Jean-Frederic > Gesendet: Freitag, 8. Juni 2007 13:16 > An: dev@httpd.apache.org > Betreff: Re: ProxyTimeout does not work as documented > > On Tue, 2007-05-22 at 10:07 -0400, Jim Jagielski wrote: > > On May 21, 2007, at 5:30 PM, Ruediger Pluem wrote: > > > > > > > > > > > On 05/21/2007 02:44 PM, Jim Jagielski wrote: > > > > > >> The logic should be: > > >> > > >> 1. If a per-worker value is set, use that. > > >> 2. If not, then if a ProxyTimeout value is set, use that. > > >> 3. Otherwise, use Timeout > > >> > > >> +1 on fixing that :) > > > > > > This sounds sane and I plan to do this, but what about > the original > > > question? > > > Do I get you right that you propose to adjust the > documentation for > > > ProxyTimeout? > > > The current behaviour of ProxyTimeout is to fall back to > Timeout if > > > no ProxyTimeout > > > is set. The documented behaviour is to have a default > value of 300 > > > secs if there > > > is no ProxyTimeout set (regardless of the setting of > Timeout, which > > > also defaults > > > to 300). > > > > > > > I think that the above logic makes the most sense and that > > the code and the docs should be adjusted to match the > > logic :) :) > > The timeout is set to c->base_server->timeout in core_pre_connection() > called by ap_proxy_connection_create via ap_run_pre_connection. > Quick patch is apr_socket_timeout_get() before > ap_run_pre_connection and > apr_socket_timeout_set() after if needed. > > Comments?
This is hackisch, but I guess we have no other choice. Another good reason to hate the approach we are using to do HTTP client operations in the proxy code. Inverting the filter chains we use for request handling in order to do HTTP requests by ourselves has really a lot of nasty sideeffects. Enough rant for now :-). Regards Rüdiger
