Skip to content
downpipes docs

Push downpipes backup-health metrics to Datadog

Datadog is the one supported monitoring tool whose OTLP intake accepts the JSON body this engine sends, so downpipes can push its backup-health metrics straight to Datadog with no agent and no collector in between. Datadog’s OTLP metrics intake takes HTTP JSON and auto-facets the attributes, so downpipe_id and downpipe_name land as tags with nothing to map.

If you would rather pull than push, Datadog’s Agent can also scrape the /metrics endpoint through its OpenMetrics check.

What you need

  • A Datadog API key that can submit metrics, and your Datadog site’s OTLP metrics intake URL.
  • Owner access to the downpipes console to set the OTLP metrics push, which is Owner-gated and dual-control once a second Owner exists.

Set it up

  1. In Datadog, create an API key that can submit metrics. Note your site’s OTLP metrics host: on US1 it is https://otlp.datadoghq.com/v1/metrics, and other Datadog sites use a different host, so check the one for your account.
  2. In the console, open Integrations, choose the Datadog tile, and set up the OTLP metrics push there.
  3. Set the Endpoint URL to your Datadog OTLP metrics URL. For US1 that is https://otlp.datadoghq.com/v1/metrics.
  4. Set the auth header name to dd-api-key and the auth header value to your API key. The value is write-only: the engine seals it and never reads it back.
  5. Enable the destination. There is no test send for the OTLP push, so the first evidence it works is the delivery trail after the next scheduler tick, roughly fifteen minutes later. Check the trail shortly after.
  6. In Datadog, confirm the downpipe_* metrics arrive, then build a staleness monitor on downpipe_backup_last_success_timestamp_seconds against each downpipe’s cadence. Widen any “no data” window past fifteen minutes so the push cadence does not trip it.

Good to know

  • Datadog’s OTLP metrics intake accepts HTTP JSON, which is exactly what this engine sends, so the direct push works with no agent. Datadog is the only supported vendor for which that is true today; the others use the /metrics scrape.
  • The push rides the engine’s fifteen-minute reconciliation tick and carries a fresh snapshot each time, with no cursor and no backfill, so a missed tick is a gap in your series rather than lost history. The pull alternative through Datadog’s OpenMetrics check needs a minted metrics token and a few config fields. Both directions are covered on OTLP metrics push and the /metrics endpoint.

Last updated .