we just enabled HTTP/2 in out frontends (bin … alpn h2,http1.1).
Since we observe, that the number of open connections is 5 times higher than before.
timeout http-keep-alive was 12000 (ms). A reduction to 2000 (ms) did no effect.
Is this effect normal. Did someone else observe this behaviour?
timeout http-keep-alive docs:
When using HTTP/2 “timeout client” is applied instead. This is so we can keep using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2 (where we only have one connection per client and a connection setup).
This is expected behavior (however this will change in haproxy 2.6).