Coyote Point Systems Equalizer, Руководство по установке и администрированию

Страница: 24 / 144

Chapter 1:Overview
14 Equalizer Installation and Administration Guide
Should either Equalizer fail to respond to a status message probe, the survivor begins a
diagnostic cycle. As part of the diagnostic cycle, the survivor attempts to contact its sibling
via the other network interfaces. If these attempts fail, the sibling is considered to be down.
In addition to status probes, the sibling Equalizers exchange current configuration
information. When you update the configuration on either machine, the configuration on
its sibling is automatically updated.
When the backup Equalizer determines that its sibling is down, it initiates a failover
process:
1. The virtual cluster aliases are configured on the external interface. When the backup
Equalizer configures these aliases on the external interface, it sends out “gratuitous
ARP” packets that cause any hosts on the external network to replace ARP table
entries that point to the failed Equalizer with the physical address of the backup unit.
2. A failover gateway alias is configured on whichever interface is local to the servers. In a
non hot-backup configuration, the servers use the IP address of the internal or external
interface as their default gateway. In a hot-backup environment, the gateway address
can migrate between the primary and backup unit so an additional address is
required.
3. The Equalizer kernel moves out of BACKUP mode into PRIMARY mode. When an
Equalizer is in PRIMARY mode, it performs gateway routing of packets between its
internal and external interfaces as well as address translation and load balancing.
When a failed unit is brought back online, it begins to exchange status messages with
its sibling. Once both Equalizers have synchronized, the newly-started unit assumes
the backup role.
Using Reserved IP Addresses
Equalizer supports placing servers on reserved, non-routable networks such as the class A
network 10.0.0.0 and the class C network 192.168.2.0. In environments where the
conservation of IP addresses is important, using reserved IP addresses can minimize the
number of “real” IP addresses needed.
For example, an ISP hosting several hundred unique web sites replicated on three servers
might not want to assign real IP addresses for all of them because each virtual cluster
would consume four addresses: three on the back-end servers and one for the virtual
cluster. In this case, the ISP might use 10.0.0.0 (the now-defunct Arpanet) as the internal
network and assign virtual server addresses out this network for the servers.
Figure 9 shows a reserved network configuration in detail.