Sampling decides which traces are forwarded. It reduces trace volume before traces reach OpsRamp, while retention controls how long accepted spans remain available.

Choose the decision point

An SDK sampler can make an early decision when a span or trace begins. The OpsRamp trace proxy can buffer spans, assemble more of a trace, and apply trace-aware rules using fields observed across that trace. The proxy approach can retain selected error or high-latency traces while reducing routine traffic, depending on the configured rules.

Trace-proxy sampler options

Trace-proxy sampler options include deterministic, dynamic, exponential-moving-average dynamic, rule-based, and throughput-based sampling. Confirm which options are supported by your deployed version before configuring them. Throughput-based sampling targets spans per second per proxy node, not traces per second.
For field-based rules, prefer stable, bounded fields that distinguish useful traffic. High-cardinality fields can make most traces appear unique and defeat reduction. Provide fields needed for the decision when spans are created, because later attributes may arrive after a sampling decision.

Roll out safely

Review sampling rules with representative normal and error traffic. If the deployed configuration supports a dry-run mode, use it to inspect would-be decisions before allowing drops. Compare trace volume and retained diagnostic coverage after rollout; dropped traces cannot be recovered by changing retention later.