On Thu, Aug 13, 2009 at 12:16:06PM +0100, Pedro Mata-Mouros Fonseca wrote: > Greetings, > > Picking this up again, there's a little problem. The directives > timeout client and timeout server should be set to the same value. > Imagine an upload scenario again. Imagine that both timeout directives > are set to 8s (or 30s, or 120s, it doesn't really matter). Client > starts uploading and data is being sent to the server. Client timeout > counter is not reached because data is being sent, as John said > earlier. However, if uploading takes more than 8s (in this example), > the timeout server is swiftly activated...
No, not at all. The timeout is reset each time there is activity. This means that during the upload from the client to the server, both timeouts are constantly refreshed. However, if the server suddenly stops reading the data, its timeout will eventually expire. > Any obvious way to go around this, aside from increasing the timeouts? > I don't want to increase them because I don't really know how much > time an upload can take. I also know that a very obvious solution > would be to have all uploads go to a different domain and not pass by > HAProxy. Anyone dealt with this problem already? It's already OK. Try it. Some people on this list deal with uploads of gigabytes during more than one hour. So I'm pretty sure it's working OK. However, one old version (1.3.15.2 I think, but I may be wrong) had a bug which could cause what you describe. I don't think there are many such versions in the wild given that there have been 8 patches since. Willy
