Developer messages needed only when debugging code. May include recording of such things as entry and exit from methods or internal branch tracking.
This level should never be required by end users.

Informational message used to trace system inputs and actions. Significant actions in the execution of some activity;
for example, the receipt of a message or an important decision in its disposition.
This level should be sufficient for an administrator or support person to determine what occured in the system when debugging a configuration or interoperability problem in the field.

Normal but significant events. Events that are expected but provide important context, such as service restarts and reloading configuration files.
This is the default logging level, and generally logging should always include at least these messages.

Conditions that imply that some failure is possible, but not certain. Generally, external inputs that are not as expected and possibly invalid. Especially useful in low level routines that are going to return an error that may be recoverable by the caller.

An unexpected condition likely to cause an end-user visible failure.
This level should be used whenever an error response is being sent outside the system to provide a record of the internal data that are important to understanding it.

Endangers service operation beyond the current operation.
MUST be logged prior to any 'assert', or when exiting for any abnormal reason.

Fault to be communicated to operations.
Should be replaced by usage of the new Alarm subsystem.

Should be replaced with CRIT.