High availability (HA) is the industrial term for continuous, uninterrupted services. In this chapter, you will learn about the history, concepts, and implementations of high availability software and the relation between PostgreSQL replication and high availability.
We will cover these topics in detail in this chapter:
- Understanding the purpose of high-availability
- Measuring availability
- History of high-availability software
- OpenAIS and Corosync
- Linux-HA (Heartbeat) and Pacemaker
- Terminology and concepts
- High-availability is all about redundancy
- PostgreSQL and high-availability
- High-availability with quorum
- High-availability with STONITH
To quote Murphy's law:
"Anything that can go wrong will go wrong."
"Anything" really includes everything in life. This is well understood by all service providers that intend to keep their customers. Customers usually aren't satisfied if the service they want is not continuous, or not available. Availability is also called uptime and its opposite is called downtime.
Depending on the service, downtime can be more or less tolerated. For example, if a house is heated using wood or coal, the homeowner can stack up a lot of it before winter to avoid depending upon the availability of shipping during the winter. However, if the house is heated using natural gas, availability is a lot more important. Uninterrupted service (there should be enough pressure in the gas pipe coming into the house) and a certain heating quality of the gas are expected from the provider.
The provider must minimize downtime as much as possible. If possible, downtime should be completely eliminated. The complete lack of downtime is called high availability.
Also, we can talk about perception of high availability when the downtime is hidden.