What Are Drop Rules?
Every observability team knows the pain of log clutter. Whether it's automated health checks, forgotten DEBUG statements, or verbose INFO messages from rarely used services, these logs inflate storage costs without providing meaningful insight. Until recently, eliminating this noise required cumbersome infrastructure changes and coordination across teams. Now, with the Adaptive Logs drop rules feature in Grafana Cloud (currently in public preview), you can define custom rules to drop low-value logs before they are ever written to storage.
Drop rules allow you to create precise logic using any combination of log labels, detected log levels, or line content. Once a rule matches, the log is discarded—instant cost savings and cleaner dashboards. This capability parallels what was already available in Adaptive Metrics and Adaptive Traces.
How Drop Rules Fit into Adaptive Logs
Adaptive Logs employs a three-step evaluation pipeline when a log line arrives. Understanding this order helps you design effective rules:
- Exemptions — Logs that match exemption rules pass through untouched. No sampling or dropping occurs. Use exemptions to preserve critical audit logs or compliance data.
- Drop Rules — If a log line isn't exempted, drop rules are evaluated in priority order. The first matching rule applies its configured drop rate (0–100%).
- Patterns — Remaining logs (those not exempted or dropped) are candidates for intelligent optimization recommendations. Adaptive Logs suggests patterns to reduce volume without losing signal.
This layered approach ensures that only truly noisy logs are removed, while valuable data remains intact.
Practical Applications of Drop Rules
Drop rules are flexible—here are three common use cases:
1. Drop Logs by Level
Many teams have services emitting excessive DEBUG logs that consume the logging budget. A single rule can target level=debug and set a 100% drop rate, eliminating them entirely without asking developers to change code.
2. Sample Chatty, Repetitive Logs
Some logs are useful but too frequent—like periodic status messages. Instead of dropping them completely, use a drop percentage (e.g., 90%) to keep a representative sample. This maintains observability while slashing volume.
3. Target a Specific Noisy Producer
A particular service may suddenly start emitting high-volume, low-value logs. Combine a label selector (e.g., service=my-service) with additional criteria such as log level or a text string to surgically drop only the offending lines.
A Complete Log Cost Management System
Drop rules are one piece of a comprehensive cost management suite in Adaptive Logs. Each component serves a distinct purpose:
- Exemptions — Protect essential logs from any alteration.
- Drop Rules — Eliminate known noise (e.g., health checks) or apply sampling to specific workloads.
- Recommendations — Adaptive suggestions for optimizing remaining log volume.
For example, a platform team can create a drop rule that discards all health check logs (100% drop rate). This standard applies across every service automatically—no need for individual teams to reconfigure their logging libraries. Meanwhile, a batch processing job that generates repetitive output can be targeted with a stream selector and a 90% drop rate, preserving a representative stream without overwhelming storage.
By combining drop rules with exemptions and pattern-based recommendations, you gain fine-grained control over log ingestion—reducing costs and noise while keeping the data you actually need.