Jakov: There could be multiple reasons. My first thought is to run a network sniffer like tcpdump, capture the relevant traffic, then analyze, and go from there.
On Tue, May 27, 2014 at 11:40 AM, Jakov Sosic <[email protected]> wrote: > Hi guys, > > > I can see a lot of errors like this in my haproxy log: > > May 26 20:57:40 localhost haproxy[9762]: <source_ip>:53644 > [26/May/2014:20:57:40.611] main backend/server03 12/0/-1/-1/12 503 212 - - > CCVN 148/105/16/6/0 0/0 {www.example.org|Mozilla/5.0 (Windows NT 6.2; WOW64) > Appl|http://www.example.com/some/link/to/content/} {|} "POST > /cs/compare/check HTTP/1.1" > > > > May 26 21:03:49 localhost haproxy[9762]: <source_ip>:49983 > [26/May/2014:21:03:49.583] static backend/server02 118/0/-1/-1/118 503 212 - > - CCVN 128/45/12/4/0 0/0 {static.example.com|Mozilla/5.0 (Windows NT 6.0) > AppleWebKit|http://www.example.com/lang/find/private/from/2014} {|} "GET > /images/front/image.jpg HTTP/1.1" > > I have a lot of these daily (~2% of total log entries), they are mostly with > flags CCVN and numbers like: > > 118/0/-1/-1/118 > > So, Tq = Tt, which means that connection spent whole time in waiting for the > client to send a full HTTP request... If I'm not mistaken, that means client > aborted before haproxy received full request, and backend servers weren't > contacted at all? > > Does too many of these 'abortions' mean that I may have problem with my ISP, > or my network gear in front of HaProxy somehow breaking incoming > connections? > -- Sasha Pachev Fast Running Blog. http://fastrunningblog.com Run. Blog. Improve. Repeat.
