a technique i've used for such code is to have a pool of pre-connected sockets to the internal servers. that way when you get a request you don't have to wait to get a new server connection... and you don't have to spend time annoying the user timing out when one of your servers is down.
this sort of thing only works well in a threaded or state-based server. you need to play a bit with the algorithms for populating the pool. start off with one conn to each server. have a periodic task wake up to add conns to servers if there's deemed to be insufficient for the load you're experiencing (or remove conns if the load has dropped). and whenever you finish using a server conn, immediately relaunch one to that server (this maintains the pool size). -dean On Tue, 27 Mar 2001, Ian Holsman wrote: > Hi. > we have a need to implement timeouts in the reverse proxy code ( for > example ... if the remote server hasn't responded in 10 seconds) > I was wondering how I would go about doing this. > > I'm looking in proxy_http.c code, and notice that it is doing a > apr_bucket_read.. it there a way to tell the bucket read function how > long for it > to wait? > > Thanks > > ..Ian > > > Oh.. the cookie thing was due to 1 cookie being host based, and the > other being domain based. > I've also raised a bug with weblogic for the HTTP 1.1 problem > > >
