I have this question on using the HAProxy in a load-balanced manner in some
configurations.
This arrangement is what is described in the following sections in the
architecture document:
2. HTTP load-balancing with cookie prefixing and high availability
2.1 Variations involving external layer 4 load-balancers
The question is how does each proxy keep accurate track of the load on each web
server? If I assign a maxconn value to my servers in the config files, how does
one proxy know if what it the max connections are going to be reached or not?
each proxy knows about the connections itself is establishing and does not know
about the other proxy?
What I can think of would be to halve the maxconn and put that halved value in
each config file. The problem with this approach is if one of the proxies goes
down, the remaining proxy will only load the server(s) up to half of their
capacity (unless I update their config again).
I'm pretty sure this situation has been thought of otherwise the architecture
document would not recommend a network arrangement like that.
I'd love to hear the how!
Thanks
Jeff
