On Sat, Jun 11, 2016 at 2:14 PM, Gert Doering <g...@greenie.muc.de> wrote:
> One thing I noticed when testing which is not *caused* by this patch but > which you/we might want to fix while at it. > > When talking to a "deaf" HTTP proxy - namely, one that accepts the TCP > connect, but then never sends back anything (try 195.30.4.1:23, it's > a "Teergrube" to annoy wannabe hackers) the timeout will fire at the > correct interval - 25s in my test, but then it won't restart but exit: > > Sat Jun 11 20:05:59 2016 Attempting to establish TCP connection with > [AF_INET]195.30.4.1:23 [nonblock] > Sat Jun 11 20:06:00 2016 TCP connection established with [AF_INET] > 195.30.4.1:23 > Sat Jun 11 20:06:00 2016 Send to HTTP proxy: 'CONNECT 10.0.0.1:51196 > HTTP/1.0' > Sat Jun 11 20:06:00 2016 Send to HTTP proxy: 'Host: 10.0.0.1' > Sat Jun 11 20:06:24 2016 recv_line: TCP port read timeout expired: > Operation now in progress (errno=115) > Sat Jun 11 20:06:24 2016 SIGTERM[soft,init_instance] received, process > exiting > Isn't this controlled by the http-proxy-retry option which is off by default? Set that in the config and SIGUSR1 restart should happen, I think... Selva
