Setting Up A Failover Configuration; Failover Partition - Coyote Point Systems E350GX Installation And Administration Manual

Chapter 5: Configuring Equalizer Operation

Setting Up a Failover Configuration

You can set up two Equalizer GX systems in a hot backup, or failover, configuration. In such a configuration, one of
the systems handles incoming requests (the primary system), while the other (the backup system) waits for a failure
to occur and automatically takes over if the Equalizer that is currently handling requests fails. The two Equalizers
are called failover peers or siblings in such a configuration.
To use a second Equalizer as a hot backup or failover peer, you need to install both Equalizers so their network
interfaces have corresponding configurations (see Figure 13 on page 36). The two peers need to be able to
communicate over the configured failover aliases.
Note – Be sure that you do not create a network loop between the interfaces of the two units when connecting
them to the network.
You must designate one of the Equalizers as the preferred primary; the second is the preferred backup. When you
boot both Equalizers at the same time, the preferred primary Equalizer is activated. If the primary Equalizer fails, the
backup takes over. When you bring the failed unit back online, it assumes the backup role until another failure
occurs or you reboot its peer.
A failover configuration requires an additional IP address on each VLAN, called the failover alias. These IP
addresses are initially assumed by the preferred primary system and are used as the network-visible interfaces of the
Equalizer. When a failover occurs, the failover aliases are then assumed by the backup system.
When an Equalizer in failover is brought online, it checks to make sure that the configured network interfaces are
link active by pinging a configured server and the failover peer. If the interfaces are not active, Equalizer sits in a
loop waiting for them to become active (and sends comments to the console).
Note – In order for failover to properly initialize, there must be at least one active server directly connected to the
subnet defined for one of the VLANs in the configuration.
Once the network interfaces are active, the failover peers begin a negotiation in which one system becomes the
primary unit and the other becomes the backup unit. This is accomplished by the backup system performing a
reboot.
When a backup Equalizer loses contact with its failover peer, it tries to determine the cause. If it cannot identify the
cause, it will try to assume the primary role. If no other system has configured the gateway IP address and cluster
addresses, the backup Equalizer assumes those IP addresses and starts handling traffic.

Failover Partition

A partition occurs when both failover Equalizers are unable to communicate with each other over the failover IP
addresses defined in the VLAN Configuration (all VLANs defined on both failover peers must have a failover IP),
but both systems are still up and running. Each Equalizer continually probes the status of the other Equalizer over all
failover IPs. Equalizer determines that a partition has occurred when the following conditions are observed:
• three (3) failover probes have failed on one failover IP
• at least one (1) failover has occurred on all other failover IPs
When the two conditions above are met, Equalizer enters primary mode. Equalizer does this to maintain cluster
availability, and this may result in network errors since both Equalizers assume the IP addresses for all clusters.
Once the issue causing the partition is fixed and both units regain communication, the two systems resolve this
dispute by choosing one system to reboot itself. Generally, this means that the system that is configured as the
default backup will reboot; upon coming back up, it will enter backup mode.
90 Equalizer Installation and Administration Guide