On Thu, Jun 11, 2015 at 5:41 PM, Joseph Lynch <[email protected]> wrote: >>> Jun 10 17:27:33 localhost haproxy[23508]: 10.126.160.11:37139 >>> [10/Jun/2015:17:26:03.027] http-in resub-bb-default/njorch0pe16 >>> 30935/0/1/-1/90937 504 194 - - sH-- 16/14/0/0/0 0/0 >>> {569760396|297|RESUB|EMAIL|0|9001|0|0|1.0|NJ|60} "POST /somepath >>> HTTP/1.1" > The interesting bit of this to me is the timing events: > 30935/0/1/-1/90937. My understanding of > http://cbonte.github.io/haproxy-dconv/configuration-1.5.html#8.4 > indicates that this took 30s for the proxy to receive the client > request and over 90 seconds before timing out. What do you have > "timeout server" set to? The docs suggest multiples of 3 usually > indicate packet loss,
Well, retransmit occurs after 3s, 9s, 27s, etc... In his case, I guess the timeout server is 60s, which is not enough, but obviously already high! > so it might be worth running tcpdump on your > outgoing traffic on the proxy and on your incoming traffic on your > service's server and trying to see where these seconds are coming from > (wireshark can be helpful to find these long sessions). If your > application log doesn't show the request then that to me is more > evidence that your requests are having issues getting from your proxy > to your backend servers. Very true, tcpdump is your friend! Have you remarked any common pattern between those 504? Same source IP, same cookie value, same URLs, same server, etc... Baptiste
