Release Notes

Amigopod |Technical Note High Availability Deployment Guide |5
2 About High Availability
Terminology & concepts
A cluster consists of a primary node and a secondary node, configured so that a failure of
either node will not prevent the cluster as a whole from performing its normal functions.
The primary node is the active server in a cluster. The cluster’s network services are
always delivered by the primary node.
The secondary node is the backup server in a cluster. If the primary node fails, the
secondary automatically takes over and continues delivering network service.
Fault tolerance is the ability of a server cluster to continue operating if either the
primary or secondary node experiences a hardware failure.
Fail-over is the process by which the secondary node assumes control of the cluster once
the primary node has failed.
A cluster’s virtual IP address is a unique IP address that will always be assigned to the
primary node of the cluster. In order to take advantage of the cluster’s fault tolerance, all
clients that use the cluster must use the cluster’s virtual IP address, rather than each
node’s IP address.
Replication is the process of ensuring that the secondary node maintains an exact copy of
the primary node’s database contents and configuration. Replication is used to ensure that
if a fail-over is required, the secondary node can continue to deliver an uninterrupted
service to clients of the cluster.
Keep-alive is the process by which cluster failures are detected. The primary and
secondary nodes verify that each is able to communicate with the other node by sending
network requests and answering with a response.
Database replication is the process of ensuring that all changes to the database,
including new guest accounts, changes to existing guest accounts, RADIUS roles, NAS
servers, and RADIUS accounting information, are replicated from the primary node to the
secondary node. This replication process occurs continuously in a normally operating
cluster. Replication is required so that in the event of a primary node failure, the secondary
node is up to date and can continue to deliver the same network services to clients.
Configuration replication is similar to database replication, but occurs at a slower rate
due to the reduced frequency of configuration updates.
The downtime threshold is the time for which the primary node of the cluster must
remain offline before an automatic fail-over will be initiated. This is 30 seconds by default.
Network architecture
Diagram 1 shows the network architecture for a high availability cluster.