Telemetry
Browse docs
GuidesUpdated July 27, 2026Reviewed by the Telemetry product team2 min read

Use this doc with your coding agent

Copy a Telemetry prompt and ask Claude Code, Codex, or Cursor to instrument the workflow covered here.

PromptsGet API key
On this page
  1. Verify the query manually
  2. Verify the condition
  3. Inspect evaluation history
  4. Check recipients and email delivery
  5. Reduce noise safely

Alert Delivery and Troubleshooting

An alert has four independent parts: the query, the evaluated points, the threshold state, and notification delivery. Check them in that order so an email problem is not confused with a query that returned no eligible data.

Verify the query manually

Open the saved Explore or SQL query and run the same time range used by the alert. Confirm it returns a numeric value and a chronological timestamp column when the alert evaluates a series.

Use complete buckets. The newest point may still be changing, so enable Ignore last data point when a partially filled bucket could cross the threshold.

Verify the condition

Check:

  • aggregation such as average, maximum, sum, or p95
  • number of recent points
  • comparison operator
  • threshold and unit
  • evaluation interval
  • minimum traffic or denominator safeguards in the query

An error-rate alert should return both failures and requests before calculating the percentage. A latency alert should state whether the value is milliseconds or seconds.

Inspect evaluation history

Open the alert detail page and review recent evaluation status and events. Distinguish:

  • no data returned
  • query execution failed
  • points evaluated but did not breach
  • breach detected
  • notification delivery attempted

If the query fails, run it in the Query workspace and inspect table names, field types, SQL compatibility, and the configured timestamp column.

Check recipients and email delivery

Confirm recipient addresses are current and do not contain spaces or copy-and-paste artifacts. Check spam filtering and allow messages from the configured Telemetry sender. Use a controlled test threshold during setup, then restore the production threshold after delivery is confirmed.

Do not repeatedly lower a production alert threshold merely to test it. Clone or temporarily isolate the test when paging or revenue workflows could be affected.

Reduce noise safely

Require multiple complete points, minimum volume, or an absolute impact threshold before notifying. Send informational lifecycle changes to scheduled review; reserve immediate alerts for conditions with a clear owner and response.

Read Alerts, then use the alert guidance included with every SQL recipe.