Hey,
I'm experiencing some weird thing with Haproxy. I'm new to it, so I hope that you guys will figure this one out in no time :). Here it goes:
I configured Haproxy to do TCP load balancing. That's because I'm balancing HTTP+SSL traffic and my traffic may be too heavy to decipher it on one machine. So the SSL traffic gets forwarded to the backends, they do their job, send the response back, everything's fine. One problem though: It takes too long! It's like 3.5 seconds per one request. My backends are also visible to the Internet and if I talk to the backend directly, I get the response in no time. So something is definitely wrong here. What is really interesting is that if I talk to Haproxy on port 443 and send plain HTTP request (via netcat), I get the response from backend immediately. The response says that I sent plain HTTP request to HTTP+SSL, but never mind. And what's even more interesting, I can't find noone else who would experience such delays. It seems that the issue is only related to my computer and my SSL client library. But I'm worried that my users may also experience such thing. Any idea, what may be wrong? Below is my Haproxy config.
Regards,
Mike
################
maxconn 10000
ulimit-n 22000
nbproc 1
user some_user
group some_group
defults
mode tcp
clitimeout 60000
srvtimeout 30000
contimeout 4000
listen lb lb.mydomain.com:443
balance roundrobin
server backend1.mydomain.com:443 weight 1 maxconn 4096 check #
routing between lb and backend is via local network
# repeat that nine more times
###################Received on 2008/06/15 20:54
This archive was generated by hypermail 2.2.0 : 2008/06/15 21:00 CEST