Skip to main content

​
Getting Started

azureeventhubs

​
Configuring Destination Fields

✅ Traces ✅ Metrics ✅ Logs ❌ Profiles
  • AZURE_EVENT_HUBS_NAMESPACE string : Namespace. Event Hubs namespace name, without the .servicebus.windows.net suffix. Requires Standard, Premium or Dedicated tier; Basic does not support Kafka.
    • This field is required
    • Example: my-namespace
  • AZURE_EVENT_HUBS_CONNECTION_STRING string : Connection String. Namespace-level SAS connection string with Send permission. Use a dedicated policy and omit EntityPath when exporting multiple signals.
    • This field is required
  • AZURE_EVENT_HUBS_TRACES_TOPIC string : Traces Event Hub. Pre-create this event hub in Azure. Each signal requires a separate event hub for downstream OTLP decoding.
    • This field is optional and defaults to otlp_spans
  • AZURE_EVENT_HUBS_METRICS_TOPIC string : Metrics Event Hub. Pre-create this event hub in Azure.
    • This field is optional and defaults to otlp_metrics
  • AZURE_EVENT_HUBS_LOGS_TOPIC string : Logs Event Hub. Pre-create this event hub in Azure.
    • This field is optional and defaults to otlp_logs
  • AZURE_EVENT_HUBS_ENCODING string : OTLP Encoding. Use the same encoding in the downstream Kafka receiver. OTLP JSON allows payload inspection in Azure Data Explorer.
    • This field is optional and defaults to otlp_proto
  • AZURE_EVENT_HUBS_COMPRESSION string : Compression. Use none on Standard tier. Kafka gzip compression requires Premium or Dedicated tier.
    • This field is optional and defaults to none
Event Hubs transports OTLP telemetry; it does not provide an APM query interface. Profiles are unsupported.Create a namespace on Standard, Premium or Dedicated tier (Basic has no Kafka endpoint). Pre-create one event hub per enabled signal, using the configured names (defaults: otlp_spans, otlp_metrics, otlp_logs). Kafka topic auto-creation is disabled. Allow outbound TCP port 9093.In Azure Portal, open the namespace’s Shared access policies, create a dedicated policy with Send permission, and copy its namespace-level connection string to the secret field. Do not use a generated SAS token. A namespace-level connection string without EntityPath can publish to all three hubs. This integration supports SASL/PLAIN over verified TLS 1.2 or newer; OAuth and managed identity are not exposed.Use a downstream OpenTelemetry Collector with the Kafka receiver, a separate pipeline per signal, the same topics and encoding, TLS enabled and SASL/PLAIN authentication. Give the consumer its own Listen policy. Set initial_offset: earliest for initial verification and a unique group_id; consume all partitions before retention expires. Forward decoded telemetry to a queryable backend such as Jaeger for traces, Prometheus for metrics and an OTLP-capable logs backend. Each record contains the signal-specific OTLP export request, preserving resource attributes, including multi-resource batches; arbitrary JSON and raw log encodings are not used.The producer limits each uncompressed message to 1,000,000 bytes and buffers one record at a time to keep requests below Azure’s 1,046,528-byte limit. This favors compatibility over throughput. Use none on Standard tier. Kafka gzip compression requires Premium or Dedicated tier; Standard rejects compressed records. Other codecs are unsupported. Oversized records fail permanently; reduce upstream batches or oversized individual records. An oversized OTLP batch can drop all its items, including small records. The OTLP receiver can acknowledge a queued request before this export failure occurs. Retriable failures use queued retries for up to 120 seconds. Delivery is at least once, so consumers must tolerate duplicates. Kafka acknowledgments confirm transport, not downstream processing, and do not contain OTLP per-item partial-success counts.For UI payload inspection, select otlp_json, open the namespace’s Data Explorer, select the event hub and consumer group, select Oldest position, and choose View events. Select a received event to inspect its OTLP body. Alternatively, configure an Azure Data Explorer Event Hubs data connection with JSON format, mapping the signal’s resourceSpans, resourceMetrics or resourceLogs into a dynamic column. Portal message-count charts alone do not prove each signal arrived. Bound downstream fetch sizes and allow sufficient request timeout for the namespace’s throughput capacity.References: Kafka connectivity and authentication, client constraints, quotas, and Kafka receiver configuration.

​
Adding Destination to Odigos

There are two primary methods for configuring destinations in Odigos:
Using the UI
1
Use the Odigos CLI to access the UI
2
Click on Add Destination, select Azure Event Hubs and follow the on-screen instructions
Using Kubernetes manifests
1
Save the YAML below to a file (e.g. azureeventhubs.yaml)
2
Apply the YAML using kubectl