Hi Florian, On Tue, May 20, 2014 at 12:07:32PM +0200, Florian Munz wrote: > >But since version 1.5-dev25, you have > >unix sockets and abstract namespace sockets which have the benefit of > >showing a lower latency (2.5 times lower than TCP) and not consuming > >ports, so they're better suited for local communications. > > Ah, I see, cool. Are there any system limits I should be aware of if I > replace the multiple bind/server with a single bind/server going over a > unix socket?
Not that I'm aware. Usage of unix sockets to the server is very recent so we'll learn over time if some tunables are needed to improve experience. > >Warning, you have a syntax issue here : > > > > timeout client 2m if host_is_api > > > >It's not correct. That sets your timeout to 2m, because the rest of > >the line after the value is simply ignored. > > ah, damn. So there is no way to have conditional client timeouts based > on the Host header? Unfortunately not, conditional timeouts are in the roadmap but since we added a number of new timeouts (timeout tunnel for example), there has been less incentive in implementing them. You should have a large client timeout to cover all uses, and use your backend to set a different server-side timeout. Since you can now set a client-fin and server-fin timeout to handle half-closed connections, you can easily get rid of dead client connections that way. Hoping this helps, Willy
