On Mon, Dec 17, 2012 at 7:29 AM, Jean-Marc Spaggiari < [email protected]> wrote:
> I was doing that the oposite way. I mean, I was starting the request > on a thread, checkinv every 100ms if it was done, and killing it after > a certain period of time. But some time the abort was not working and > it was feeling my memory with pending threads... > If you're worried about filling your app with abort checking threads then you could adapt my solution to use only a single thread that runs for the entire app lifetime. Just (in a multithreaded way) add each request to a queue sorted by the absolute timeout time. Then your abort thread wakes up every second and shifts items off the bottom of the queue that should be aborted. > If I use such a solution, does it mean that the 2 TimeOut that I'm > already setting are useless? Will the abort close both the TCP and the > HTTP connections? > I wouldn't say they're useless, but I would suggest setting the thresholds lower. For example, if the server takes 20 seconds to finish the TCP connection handshake, will it really perform your request in the remaining 10? Unlikely.
