Datadog log integration via OpenTelemetry Collector

What's new with OTLP

The OTLP pipeline delivers structured records instead. A few highlights specific to how Datadog handles them:

• SOURCE is set to "cloudamqp" automatically. Datadog uses source to route logs through its processing pipelines and apply automatic parsing. Your broker logs will be recognised as a CloudAMQP source from the moment they arrive, with no manual pipeline configuration required.
• Severity arrives as a number via otel.severity_number (9 for INFO, 13 for WARN, 17 for ERROR) rather than embedded as text in the message. Datadog log monitors and faceted search work correctly against it without regex extraction.
• Message content is clean. No timestamp prefix, no PID, no log level string. Just the log.
• One stream per node. On multi-node clusters each node ships independently. In Datadog you see individual host streams rather than a single aggregated one, making it straightforward to isolate behaviour on a specific node.

Custom tags

The integration accepts optional tags in key=value format (comma-separated, e.g. env=prod,team=platform ). These are attached as attributes to every log record and appear as facets in Datadog Log Management. Setting a service= tag overrides the service name, useful if you want broker logs grouped with a specific application service rather than the cluster name.

Setting up the integration

You will need your Datadog API key from Organization Settings → API Keys. In the CloudAMQP console, open the Integrations tab for your instance, select Datadog, enter your API key, choose your region (US1, EU, US3, US5, UK1 or AP2), and optionally add tags. Save, and logs will start arriving within a minute.

Switching from the previous integration

If you are on RabbitMQ and have the previous Datadog log integration active, the two can run in parallel. Before switching, review any monitors, dashboards, or log processors that depend on the old format. The key differences to account for:

• Severity is now otel.severity_number rather than a text string in the message body
• Message content no longer includes a timestamp or level prefix
• Multi-node clusters produce one log stream per host instead of a single aggregated stream

Full details on the log format and setup are in the CloudAMQP Datadog log integration documentation.

If you have questions, reach out to support@cloudamqp.com