In this context, validation means checking that you are building the right thing, whereas verification means checking that you are building it in the right way.
Before you begin to write specific tests, step back to review the big picture. Do the security requirements make sense for the directory service, and for the users of the service?
For the directory service, aim to prevent problems caused by security threats. The directory service must expose sensitive information only in secure ways. It must require strong authentication credentials, and limit access.
For the users, the service must permit access from any device or application using standard protocols and tools. The directory should protect your sensitive information while making it easy to connect.
In refining the security requirements for the directory service, make sure there is an appropriate balance between security and user needs. An ultimately secure directory service is one that denies all user access. An ultimately flexible directory service is one that relinquishes all control. The appropriate balance is somewhere in the middle.
Be aware that directory service security is only part of an overall strategy, one that aims to help users and application developers make appropriately secure choices. In validating the requirements, understand how the directory service fits into the overall identity and access management strategy.
Long before you roll out the directory service, when you start to prepare server configurations in a lab environment, begin by testing the directory's functional capabilities. As you understand your users' requirements, reproduce what their client applications will do in your tests. In most cases, it is not feasible to exhaustively reproduce everything that every directory client will do. Instead, choose a representative sample of actions. Test your expectations, both for normal and for insecure client application and user behavior.
The goals for your functional testing are to verify compliance, to uncover problems, and to begin automating tests early in the project. Test automation should drive you to use version control software, and continuous integration software as well. Be ready to roll back any change you make if a test fails, and make sure that every change is reviewed and tested before it is pushed further along in the process.
Aim to keep the automated tests both representative and short. As you build out the deployment and complexity grows, automated tests let you build the service with confidence, by repeatedly iterating with small changes to fix problems, and to better match users' expectations.
Before you apply a change to a production environment, verify the impact under conditions as close as possible to those of the production environment.
In the test environment, the directory service should mirror production for configuration, client application configuration and load, and directory data, including access policies. This is the environment where end-to-end testing first takes place.
Although you might not test at a scale that is identical to production, the test environment must remain representative. For example, when using replicas in production, also use replicas in the pre-production test environment. When using secure connections everywhere in production, also use them in the test environment. If you expect many client applications accessing the production directory, and particular client usage patterns, also simulate those in the test environment.
Automate your testing in the pre-production environment as well. Each change for the production environment should first pass the pre-production tests. You will need to iterate through the tests for each change.
In deployments where updates to new servers would not harm production data, consider using canary servers. You deploy a small group of canary servers in production that have the change. You then test and monitor these servers to compare with unchanged servers. If the change looks good after enough testing and monitoring, you roll out more servers with the change. If something goes wrong, you have only exposed a small proportion of production clients to the change. Only use this when you are sure that replication from a canary server would not corrupt any production data.
Directory services integrate with many monitoring solutions. Access monitoring information over HTTP, LDAP, SNMP, and JMX, and by sending Common Audit events to local log files and remote systems for further processing.
Adapt some of your tests to verify operation of the service in production. For example, a few end-to-end tests in production can alert you to problems early before they impact many users.