Jaguar’s Open Directory infrastructure, while viable for smaller organizations, suffered from one nearly debilitating flaw. The Open Directory Master itself was a single point of failure. And clients, upon that failure, had a very negative user experience. Simple things, like a Finder file listing or starting Terminal, could take two or three minutes, and the login process, which worked only for local users, could be excruciating. Clients were very tightly coupled to the directory services infrastructure, and that infrastructure consisted of a single server, subject to all of the hardware failures and network outages that implies. With Panther, however, Apple has introduced a stable and fault-tolerant replication infrastructure designed to meet the needs of a modern directory service. While not without flaws, it is a great step forward, and it deserves much of the credit for Panther’s acceptance in increasingly critical Directory Service roles.
Open Directory Replication also parallels a common trend in highly available systems, relying on application level clustering and numerous parallel hosts rather than the massively redundant single systems of the mainframe age.
This chapter examines that replication architecture, discusses the implications of the design choices that formed it, and describes best practices associated with its deployment.
Creation of an Open Directory Replica is, on a fundamental ...