(noun) /ˈɑːmstrɒŋ ˈmɛtrɪk/
A metric which is too noisy to be monitored (does not indicate system health), but which nonetheless is helpful in understanding system behavior.
Named for Jeff Armstrong.
Example:
We can’t set up an alert for transient errors because the number/rate of transient errors is an Armstrong metric - it fluctuates outside of our control and sometimes for desired reasons.
These should be traced and sampled instead.
Josh Beckman
Widgets
Comments & Replies
You can subscribe or follow or reply here: