Streamline Highly Availability and Load Balancing

[Johan Ronkainen]

> I am trying to implement the two node ha/load balanced topology. I've been
> successful with the ha part, the servers will fail over, but I've been
> unable to get load balancing to work.


Hi.

We're running two-node cluster doing web-caching (Squid) and virus-scanning
HTTP/FTP traffic (Dansguardian with av-patch and F-prot). No separate
front-end LVS boxes, just these two servers. There's 700-800 workstations
behind these two machines and weekly combined traffic is ~40GB so
utilization is fairly low.

Two identical dual AMD machines (MP2100+, 2GB RAM, 4*80GB IDE, 3*ETH). Both
have one ethernet interface to internal LAN (eth0), one with cross-over
cable to another node (eth1) and one connected to Internet (eth2).

No special kernel patches. Keepalive handling VRRP and squid service
monitoring.

Clients connect to 10.80.80.80:8080 (eth0) and LVS passes request to
10.99.99.1 (node1) or 10.99.99.2 (node2) via eth1. We're using LVS-DR with
persistence.

Only LVS related iptables rule we have is port 8080 redirection.

iptables -t nat -A PREROUTING -p tcp -d 10.80.80.80 --dport 8080 -j REDIRECT -i eth1


-- 
Johan.Ronkainen at iki.fi