Hi Pawel,
> Hi. > > We've tried migrating haproxy from 1.4.22 to 1.5.2. As a result we > experienced a serious performance impact. The only thing that was > changed was hapxory version. We tried 1.5.12 afterwards, but that > didn't really help. This is running on AWS Amazon Linux 64bit. 1.5.2 > was AWS stock, 1.5.12 we built ourselves. Check conntrack settings on both proxy and the backends. Different haproxy behavior may trigger conntrack limits by using more connections. I see you set "option httpclose" on the default. Was this set in both 1.4.22 and 1.5.x instances? This settings should theoretically make haproxy behave exactly the same. In its absence, the behavior would indeed have changed, from the old default: option http-tunnel [1] to: option http-keep-alive [2] One thing I can suggest is to replace "option httpclose" in the default section with the following configuration: option http-keep-alive option prefer-last-server timeout http-keep-alive 5s (or whatever keep alive timeout works for you) Regards, Lukas [1] http://cbonte.github.io/haproxy-dconv/configuration-1.5.html#option%20http-tunnel [2] http://cbonte.github.io/haproxy-dconv/configuration-1.5.html#option%20http-keep-alive
