Sampling

Trace sampling decides what you keep. The Sampling tab makes that decision explicit and reversible.

What it shows

• The active sampling policy.
• Suggested rules, derived from the traffic actually observed — the high-volume, low-value paths that are worth dropping first.
• The current rule set, editable.
• The ingest budget the rules have to fit inside, and how close you are to it.

Tail, not head

Sampling here is tail-based: the decision is made once a trace is complete, so errors and slow traces can be kept while the healthy bulk of the same endpoint is dropped. Head sampling cannot do that — it decides before it knows how the request went.

Rules match the root span

A drop rule matches on the root span. This is the part people get wrong: a rule written against an attribute that only appears on a child span will not match, and the trace stays. Check the suggestion’s own preview count before you apply it.

Changes are undoable from the toast, and the budget panel updates with the projection.