Hi Lukas&Baptiste
Thank you very much.
It means that tunnel mode has been implemented in current haproxy.
But it can not do balance by http transcations since it only processes
first request and response for each tcp connection.
>>Full keepalive support on the server side means something else.
>>For example we could have long-lived backend tcp connection to
>>serve more than one frontend users (theoretically).
If so, haproxy will try to reuse the connection between backend and server.
As a result, there will be less new backend connections than in
http-server-close
mode and the throughput of haproxy will be promoted. That's what I have
understood, am I right?
Best Regards,
Godbach
Hi Godbach!
I am wondering that whether server side keepalive you said is
http keepalive between haproxy and server or not.
Thats it exactly.
After testing lastest snapshot, http keepalive has already
token effect between haproxy and server under such conditions:
What you are seeing is not http keepalive between haproxy
and the server, but is actually between the client and the
server, because the way you configured it, haproxy is in tunnel
mode.
This is from the docs:
By default HAProxy operates in a tunnel-like mode with regards
to persistent connections: for each connection it processes
the first request and forwards everything else (including
additional requests) to selected server.
Full keepalive support on the server side means something else.
For example we could have long-lived backend tcp connection to
serve more than one frontend users (theoretically).
Regards,
Lukas
