Mastering Log Cost Control: Custom Drop Rules for Adaptive Logs

By ⚡ min read

Introduction

Most platform and observability teams grapple with logs that are nothing but noise. Health-check pings, forgotten DEBUG statements, or verbose INFO messages from rarely used services can inflate your logging bill without providing real value. The challenge has always been how to get rid of them quickly and efficiently, without involving cumbersome infrastructure changes. With the new drop rules feature in Adaptive Logs (now in public preview), Grafana Cloud offers a straightforward way to define rules that drop low-value logs before they are ever written to storage. This reduces noise and saves money immediately.

Mastering Log Cost Control: Custom Drop Rules for Adaptive Logs

How Drop Rules Work

Drop rules allow you to create custom logic using any combination of log labels, detected log levels, or line content. When a log line arrives in Grafana Cloud, it passes through a sequence of checks. The first step is exemptions—protected logs that must never be dropped. Next, drop rules are evaluated in priority order; the first matching rule applies its drop rate. Finally, remaining logs that weren't exempted or dropped can be further optimized with pattern-based recommendations (as already available in Adaptive Metrics and Adaptive Traces).

Real-World Examples

Here are a few practical scenarios where drop rules shine:

Drop by log level: Eliminate all DEBUG logs from a particular environment to instantly cut your logging budget.
Sample chatty, repetitive logs: Instead of dropping 100%, specify a drop percentage (e.g., 90%) to keep a representative sample without drowning in duplicates.
Target a specific noisy producer: A misconfigured service suddenly emits flood of INFO logs. Combine a label selector with log level or text string to surgically reduce that stream.

These examples barely scratch the surface. You can chain multiple criteria to precisely control what stays and what goes.

The Complete Log Management System

Drop rules are just one piece of a comprehensive log cost management system within Adaptive Logs. Together with exemptions and pattern recommendations, they give you full control:

Exemptions protect critical logs from any sampling or dropping.
Drop rules eliminate known noise—for example, a platform team can enforce a 100% drop on health-check logs across all services without requiring individual teams to change their logging configuration.
Pattern recommendations apply intelligent optimization to the remaining logs, reducing volume without losing important signals.

This layered approach ensures you never accidentally drop valuable data while systematically removing waste.

Key Benefits

Using drop rules, teams can achieve:

Immediate cost savings by stopping ingestion of low-value logs.
Reduced noise in dashboards and alerts, making it easier to spot real issues.
No infrastructure changes—rules are defined centrally in Grafana Cloud.
Fine-grained control with label, level, and content filters.

Getting Started

To begin using drop rules, visit the How Drop Rules Work section above to understand the evaluation flow, then explore the Adaptive Logs documentation for detailed instructions. Start with a single rule targeting your noisiest service—set a modest drop rate, monitor the impact, and adjust as needed. With drop rules, you reclaim control over your log pipeline and your budget.