Coyote Point Systems Equalizer GX Series, Administration Manual

Page: 584 / 866

Failover
Network Design for N+1 Failover
The design of the host network is critical to a successful failover configuration.
The essential concept of active-active failover is that resources that are required for a cluster to
serve client requests are organized into "failover groups". For any cluster, the required resources
include:
l
the cluster object and all objects to which it points including server pools, server instances,
servers, responders, certificates, etc.
l
the subnet on which the cluster IP address resides
l
the subnet (or subnets) on which all server IP addresses in the server pool reside
If you instead locate a cluster IP address on one subnet and the servers in the cluster’s server
pool all reside on another subnet, then both those subnets would be considered part of the
cluster’s failover group.
So, in order to allow each cluster to fail over separately to another Equalizer, the cluster IP
address and all server IP addresses need to be located on Equalizer subnets that are distinct from
the subnets on which other cluster and server IP addresses reside.
Once you configure cluster and server IPs and enable active-active failover, the clusters, servers,
subnets, etc., are organized into "failover groups" that can be passed between all the peers at
network connectivity issues occur.
How a Peer is Chosen for Failover in N+1 Configuration
A failover occurs when Equalizer detects that there is an issue with one of the subnets on which a
cluster’s IP address or one of its server IP addresses resides. This typically means that Equalizer
has lost connectivity on a subnet, and can happen for any number of reasons; for example, the
failure of a downstream hub, router, or other networking device.
A failover event can be simulated by either removing a cable from Equalizer’s front panel or
rebooting a peer.
In our example configuration, each VLAN subnet is connected to Equalizer through a separate
port. When you remove a cable from Equalizer, it recognizes that it has lost connectivity on that
subnet and attempts to fail over all the resources on that subnet (the "failover group") to another
peer.
When Equalizer detects a network connectivity failure, it does the following:
1. It determines which failover groups are affected by the failure.
2. It examines the heartbeat information it has received from the other peers in the failover
set, and determines which other peers can provide connectivity on the subnets that have
failed.
3. If there is only one peer that can provide the required connectivity, the failover group is
moved to that peer.
4. If there is more than one peer that can provide the required connectivity, Equalizer checks
the ‘preferred peer’ setting on the cluster (or clusters) in the failover group (or groups), and
if the preferred peer can provide connectivity, the failover groups are moved to that peer.
584 Copyright © 2014 Coyote Point Systems, A Subsidiary of Fortinet, Inc.