OllyGarden APIis now available!Access telemetry quality insights programmatically to integrate with your existing tools and workflows.Explore API →
On this page
Verbose Log Ratio
NormalSignal: Logs
Problem
DEBUG and TRACE records dominate a service’s log output. That verbosity is development detail, invaluable on a laptop but mostly unread in production, and at high ratios it means the service ships far more bytes than its operators ever look at.
Insights measures the severity mix per service over a 1-hour sliding window, evaluated only once the window holds enough logs to be statistically meaningful. The insight fires at Normal severity when DEBUG and TRACE exceed 60% of volume, and escalates to Important at 80%.
Impact
Most of the log bill buys development detail in production.
Higher-severity lines get harder to spot in searches and live tails.
Verbose logging often runs hot in the request path, costing CPU too.
Example
Verbose:
DEBUG entering handler path=/checkout
TRACE deserializing payload bytes=418
DEBUG cache lookup key=cart:8812 hit=true
DEBUG exiting handler status=200 elapsed=41ms
INFO order placed order=8812
Balanced:
INFO order placed order=8812
# debug detail available via a runtime toggle when needed
How to fix
Set production log levels to INFO or above; keep DEBUG for development and short-lived investigations.
Make verbosity a runtime toggle with an expiry, so raising it for an incident doesn’t become permanent.
If some debug lines are genuinely needed in production, promote the valuable ones to INFO and delete the rest.
Expected gains
Log volume and cost drop to what operations actually reads, and the lines that matter stand out again.