Add the tailsampling processor to the Collector configuration and connect it after the receiver in the traces pipeline
Define policies such as latency thresholds, status-code error sampling, and probabilistic sampling as named policy blocks
Combine policies with the and or composite policy type to create nuanced sampling decisions
Set the decision_wait duration long enough to receive all spans for a trace before making the sampling decision
Deploy the Collector as a stateful gateway tier so all spans for a given trace arrive at the same Collector instance
Known gotchas
Tail sampling requires all spans for a trace to arrive at the same Collector instance; use a load balancer with trace-ID-based routing (the loadbalancing exporter) upstream
A decision_wait that is too short causes incomplete traces to be evaluated, leading to incorrect sampling decisions for slow or async operations
Memory consumption grows with decision_wait duration and trace volume; monitor the Collector's memory usage and set appropriate resource limits
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp