Sentinel provides a framework for the execution of system monitoring tests. Individual tests are defined in a sentinel.conf file. Tests are run at scheduled times as defined in the configuration file (similar in nature to crontab schedule definitions). Tests can be contained in shared libraries which are loaded dynamically at runtime or in separate executables (compiled or shell scripts).

Upon startup, Sentinel reads its configuration file, within which is defined all the various tests to be run, required test parameters, and test schedules. If the test is contained within a shared library object, the object is loaded at that time.

Each test can contain up to three separate entry points - an init(), run(), and cleanup() routine. Each init() routine is called by Sentinel for each test upon startup. This allows each test to perform any initialization processing that might be required (opening files, allocating data structures, etc). The test's run() routine is called at each scheduled time interval as configured in the sentinel.conf file. It is during the run() stage that the specific system test is performed. Finally, the cleanup() routine is called upon shutdown of the process, allowing each individual test to release any allocated resources (close files, free memory, etc.).

The following figure displays the basic design of this process.

Figure 1.

Source code: Sentinel Server, Utility Routines, Configuration File, Sample Test