Sampling Policy
Purpose
Trace (and log) sampling policy — what we keep, what we drop, and why.
Trace sampling
- Default: tail-based sampling at the collector.
- Always keep: errors, high-latency outliers, sampled baseline.
- Don’t hard-code SDK head sampling without review.
Rationale
Balances trace fidelity against Tempo ingest cost.
[stub: sampling-rates]— fill this in. Greppable doc-debt marker.
Related
Local graph
Linked from 4 notes
Traces Instrumentation Guide
How to instrument a service for **distributed traces** on the ShipSolid observability platform (OTel
Data Plane Architecture
How telemetry flows from source to storage — the data plane.
Collector Config Templates
Reusable Alloy / OTel collector configuration templates by workload class.
05 — Platform Configuration
The knobs: collector configs, alert/dashboard catalogs, sampling and retention.
Related notes
Alert Rules Catalog
Catalog of alert rules managed by the platform (Terraform/config-driven).
Collector Config Templates
Reusable Alloy / OTel collector configuration templates by workload class.
Dashboard Catalog
Catalog of shared dashboards and the golden-signal starter pack.
Feature Flags & Config Management
How platform feature flags and configuration are managed and rolled out.