Has there ever been any discussion of hysteresis values?
That is where a service has passed a threshold where it is then in a warning
state, but when it goes below that threshold it doesn't return to an ok state
until it passes a second threshold.
For example, monitoring the size of a mail queue. The queue goes to warning
when there are say 350 messages in the queue.
Because of the nature of a mail queue it might float around this threshold for
some time. It might even end up in a flapping state.
What if it were to not go back to an ok state unless the size of the queue
were to go below 300.
Plugins could have hysteresis values optionally specified on their command
lines. If it was specified then the previous state would also need to be
provided so the plugin could determine if the hysteresis value should be
considered.
ie: previous state is ok, then hysteresis value is irrelevant.
if the previous state is warning then the hysteresis value should be
considered if the new state is to be ok.
It's possible this behaviour could be implemented as utility functions dealing
with absolute values. These functions would determine the returned state
based on the value, previous state and new state.
--
-- Sam
This post was automatically imported from historical nagios-devel mailing list archives
Original poster: [email protected]