Tame Noisy Logs and Cut Costs with Adaptive Logs Drop Rules

Logging is essential, but not all logs are valuable. Health checks, DEBUG messages, and verbose INFO logs can quickly balloon your bill and drown out useful signals. With Adaptive Logs drop rules, now in public preview, you can define custom logic to drop low-value logs before they ever reach Grafana Cloud Logs. This guide answers common questions about how to use drop rules to eliminate noise, save money, and simplify log management.

What are Adaptive Logs drop rules and how do they work?

Adaptive Logs drop rules, now in public preview, let you define custom logic to drop low-value log lines before they are written to Grafana Cloud Logs. You can combine log labels, detected log levels, or line content to create rules. When a log line arrives, it is evaluated against your rules in priority order. The first rule that matches applies its drop rate (0% to 100%). This allows you to eliminate known noise such as health check logs or verbose DEBUG messages without requiring changes in individual services.

How do drop rules help reduce log noise and costs?

By dropping logs before ingestion, you reduce storage and processing costs immediately. Noisy logs that inflate your bill—like forgotten DEBUG logs from little-used services—can be filtered out with a single rule. Centralized teams can enforce standards across all services without complex infrastructure changes. This saves money and reduces clutter in your log analysis tools.

What are examples of using drop rules?

You can drop logs by level (e.g., all DEBUG logs), sample chatty repetitive logs by specifying a drop percentage, or target specific noisy producers with a label selector combined with log level or text string. For instance, a health check service that logs every second can be dropped 100%, while a batch job that logs progress can be sampled at 90% to keep representative data.

How are drop rules evaluated alongside exemptions and patterns?

When a log line arrives, it first checks exemptions—protected logs pass through untouched. Next, drop rules are evaluated in priority order; the first match applies its drop rate. Finally, patterns (optimization recommendations) are applied to remaining logs that weren't exempted or filtered. This three-step system ensures you can protect critical logs, drop known noise, and sample the rest.

Can drop rules be used to sample logs instead of dropping entirely?

Yes. Drop rules allow you to specify a drop percentage between 0% and 100%. A 100% drop rate discards all matching logs. A lower percentage, like 90%, keeps 10% of the logs, effectively sampling them. This is useful for logs you don't want to lose entirely but can tolerate reduced volume—like repetitive batch job logs.

Who benefits most from using drop rules?

Platform and observability teams gain the most. They can centrally manage log costs without requiring every development team to change logging configuration. Drop rules give them a simple, toilsome-infrastructure-free way to eliminate noise. Individual service owners also benefit because their logs won't be cluttered with noise, and budgets stretch further.

How do drop rules compare to Adaptive Logs recommendations?

Drop rules are your custom inputs, while recommendations are intelligent optimization suggestions from the system. You can use both together: drop rules for known noise you want 100% gone, and recommendations for patterns the system identifies as repetitive. Exemptions protect high-value logs. Together they form a complete log cost management system.