[Linux-ha-dev] Packets after failover not routing correctly

[tuliol at sybatech.com]

I have 2 directors:
Director1 (public ip = 10.103.40.2, private ip 10.10.10.1)
Director2 (public i = 10.103.40.1, private ip 10.10.10.6)

Then I have ldirectord configured to do load balancing of http to 3 real
servers:

virtual=10.103.40.2:80
        real=10.10.10.3:80 masq
        real=10.10.10.4:80 masq
        real=10.10.10.7:80 masq
        checktype=negotiate
        service=http
        scheduler=rr
        persistent=600
        protocol=tcp

Everything is working great.  So now my goal is to make the directors
highly available so if one fails then the ip is transfered to the other
and they continue to route and load balance the 3 web servers.

My haresources has the following entry on both directors:
lvsdir1 10.103.40.2 ldirectord::ldirectord.cf

If I shutdown director one the everything goes according to plan.
Director 2 takes over the outside ip of director 1 (10.103.40.2) which the
returns pings when pinged from another host.  The ldirectord services are
started successfully on director 2.

My only problem is that any requests made timeout.   I ran tcpdump and it
shows that the director 2 receives the request and then it passes the
request to one of the real servers but then the real servers cant route
the packets back to the client.  The reason I think is because the real
servers have a default gateway of 10.10.10.1 which is the internal ip of
director 1 which is now down.

How do I get around this problem?

Any help or guidance will be appreciated.