I was expecting this to be a post about machine monitoring.
It's not, but the same rule applies.
Much of the past year has been tweaking Nagios and other alert settings. Some have been disabled entirely, some simply set to much more sane values (as in: if this value is hit, things are not sane).
My read:
If you can't take some action within an alert interval to change / make things better, you don't need to know about it.
If alert A means that condition B has already occurred (e.g.: site is down, whatever), then alert A is 1) useless and 2) will be polluting your (likely already overstressed) mental bandwidth / handheld device constraints at a time when it's least convenient.
E.g.: while you're trying to read a really, really crappy vendor manual in Android's PDF viewer, which helpfully resets you back to page 1 of 892 every time you navigate away from the document -- say, because a new mail/SMS alert has just arrived.
The anomalous condition should be logged for forensics, and explored for predictive ability, but if it's not triggered until the shit's already hit the fan, ditch the alerting, keep the logging.
It's not, but the same rule applies.
Much of the past year has been tweaking Nagios and other alert settings. Some have been disabled entirely, some simply set to much more sane values (as in: if this value is hit, things are not sane).
My read:
If you can't take some action within an alert interval to change / make things better, you don't need to know about it.
If alert A means that condition B has already occurred (e.g.: site is down, whatever), then alert A is 1) useless and 2) will be polluting your (likely already overstressed) mental bandwidth / handheld device constraints at a time when it's least convenient.
E.g.: while you're trying to read a really, really crappy vendor manual in Android's PDF viewer, which helpfully resets you back to page 1 of 892 every time you navigate away from the document -- say, because a new mail/SMS alert has just arrived.
The anomalous condition should be logged for forensics, and explored for predictive ability, but if it's not triggered until the shit's already hit the fan, ditch the alerting, keep the logging.