Description: This persistent variable indicates the time period in minutes that is used by the feedback/pin trap limiting algorithm as specified in RFC 1224.

The smallest value allowed for 'trapWindowPeriod' is 1 and the largest value allowed is 1440 which specifies a period of 24 hours. The default value is 15.

This variable only takes affect when 'trapControl' is set to 'feedbackPinLimited'. This method operates as described below:

If 'trapMaxPerWindow' traps occur and are sent by the Agent in the time period specified by the 'trapWindowPeriod' and another trap needs to be sent, then the Agent ignores this over the limit trap, and forces itself into a mode where it will send no more traps. The Agent sends the 'trapsDisabled' Trap, prior to entering this mode. The Agent then sets the 'trapFlow' control variable to 'closed' which allows no more traps to be sent until this variable is set to 'open', by an SNMP set operation.

The agent sends the 'trapsDisabled' Trap, so that an SNMP manager will have an indication that it has sent an excessive number of traps. The agent sets the 'trapFlow' variable to 'closed', so that a Manager can poll this variable periodically, and determine that this Agent has sent too many traps. A Network Administrator can then perform more selective MIB inquiries to determine a more complete status of theis unit.

When this limiting condition occurs, the Agent also logs a system event with a severity of 'MAJOR', marking the event as effecting the health of the system. This impact on the health of the NMU is canceled when either 'trapFlow' is set back to 'open', or 'trapControl' is changed to any value other than 'feedbackPinLimited'.

If 'trapWindowPeriod' is modified during an active timing period of this feedback method, then the Agent performs in this manner. 1) If the valid new value is larger than the previous value for 'trapWindowPeriod', then the new value is used for any trap timing in the future, and the previous trap history is expanded to the new period size. 2) If the valid new value is smaller than the previous value for 'trapWindowPeriod', then the current trap history is shorten (by reducing the older entries), and the new value is used for any trap timing in the future.

A simple example: The Agent can be configured to allow one trap per minute by setting 'trapWindowPeriod' to 1, and 'trapMaxPerWindow' to 1. With this configuration, if at any time a trap needs to be sent within 1 minute of a previous trap, the Agent would limit itself.

This implementation of the Feedback Pin technique uses a timing quantum that is one minute as determined by the system's clock. The timing quantum is synchronized to a particular second determined by the first trap of a new window period. All traps within 60 seconds of this starting second accumulate to represent traps sent within that quantum-minute. Traps continue to be counted, and assigned to a new quantum-minute depending upon the starting second, and an aggregate counter represents the total number of traps within the current period. As the full period of history accumulates, traps associated with a quantum-minute at the front of the period reduce the aggregate total as the current clock moves into a new quantum-minute. Thus this window period is a moving period related to the current clock.

A new synchronizing second is not choosen unless there is a full 'trapWindowPeriod' that has no traps sent. With this situation, a new starting second is chosen when the new trap needs to be sent.

The NMU's system clock resolution is not extremely accurate, but provides reasonable accuracy for limiting traps with this technique.