Network Adaptive Fault Tolerance with Switchback Load Balancing

Network Adaptive Fault Tolerance with Switchback Load Balancing

Post by Brian Seppane » Sat, 14 Dec 2002 05:31:32



In my quest for as much redundancy as possible I have a Dell 1650
Poweredge Server running Redhat linux 7.3 with the latest e1000 drivers
from Intel.   I've configured eth0 and eth1 into a single team called
failover, so eth0 and eth1 are configured with the same ip address and
are controlled by a virtual adaptor called vfailover.   The problem
comes about when I try to put this server in a Switchback Server Load
Balancing Environment.   In this environment there is a Foundry
ServerIron that connects to the servers providing a typical service.
The foundry executes layer 3,4 and 7 health checks and allocates
incoming client requests to the fastest responding host.   It provides
for fault tolerance when a server fails for any reason as well as the
obvious load balancing.   Anyways with the failover interface the IP
corresponds to two MAC Addresses for each interface, and the moment I
bind the service port in the config, the foundry fails on the health
checks and refuses to allocate any clients to the box.   The failure is
most likely due to the fact that it is getting two seperate mac
addresses claiming to be the same IP.   Has anyone found a way to work
network interfaces configured for Adaptive Fault Tolerance into a Server
Load Balanced switch.   I might just have to settle for simplicity and
lose the fault tolerance for this host's interfaces, but it seems such a
shame not to use Adaptive Load Balancing or Adaptive Fault Tolerance
when the interfaces support them.

Thanks,
Brian Seppanen