Depending on the password policy settings, a client account can be locked out of an account when the number of failed bind attempts exceeds the number of allowed bind attempts. In a replicated topology the client is locked out of all instances of Directory Server, not just the instance to which the client was attempting to bind. This feature is called global account lockout.
In versions of Directory Server prior to Directory Server 6, account lockout was based on integer counters. By default, these counters were not replicated.
In this version of the product, bind failures are recorded by using timestamps. By default, the timestamps are replicated, and prioritized replication is used to replicate updates to the lockout data that are caused by failed bind requests.
Global account lockout can be used in the following scenarios:
When replication is used to propagate bind failures
Bind requests must not be directed to read-only consumers. When a client fails to bind to a read-only consumer, the lockout data is not replicated. Therefore, if a bind request fails on a read-only consumer, the lockout data is updated on that instance only and is not replicated across the topology.
Even if all bind attempts are directed at master replicas, the client might be able to perform bind attempts on multiple servers faster than the lockout data can be replicated. In this way, a client can exceed the limit on failed bind attempts for the password policy. Note that this risk is present even though bind failures are replicated by using prioritized replication.
When Directory Proxy Server manages the routing of bind operations
The Directory Proxy Server can achieve global account lockout by using the hash algorithm for load-balancing to route all bind requests for a given account to the same Directory Server. For information about using the hash algorithm for global account lockout, see Operational Affinity Algorithm for Global Account Lockout.