Azure Event Grid

Azure Event Grid is a fully managed event routing service that uses a publish-subscribe model. Publishers (Azure services or custom applications) send events to topics, and Event Grid pushes those events to subscribers (Azure Functions, Logic Apps, webhooks, etc.) based on subscriptions and filters. It provides at-least-once delivery with automatic retry, supports millions of events per second, and charges only for operations used. Event Grid is ideal for building reactive, event-driven applications without managing messaging infrastructure.

Azure Event Grid pricing is straightforward: $0.60 per million operations after the first 100,000 free operations per month. Operations include event publishing, delivery attempts, and advanced filtering matches. For most applications, this results in extremely low costs. For example, 10 million events per month costs approximately $6. Event Grid Namespaces (with MQTT support) cost $0.48 per million operations plus throughput unit charges. There are no upfront costs, minimum commitments, or infrastructure management fees.

Use Event Grid for event-driven reactive programming where you need to trigger actions in response to state changes (blob uploaded, resource created, custom domain events). Use Service Bus when you need enterprise messaging with FIFO ordering, transactions, sessions, and dead-letter queues for reliable command/request processing. Use Event Hubs for high-volume telemetry and streaming scenarios (millions of events/second) with long-term retention and batch processing. Many architectures use all three: Event Grid for event routing, Service Bus for reliable messaging, and Event Hubs for analytics pipelines.

No, Azure Event Grid does not guarantee message ordering. Events may arrive at subscribers in a different order than they were published. If your application requires strict ordering, consider: (1) including sequence numbers in event data and reordering in the handler, (2) using Service Bus with sessions for FIFO guarantees, or (3) using Event Hubs with partition keys for per-partition ordering. Event Grid is optimized for high throughput and low latency rather than strict ordering.

Secure Event Grid webhooks using multiple methods: (1) Validation handshake - Event Grid sends a validation request with a code that your endpoint must echo back to prove ownership. (2) Azure AD authentication - Configure Event Grid to include Azure AD tokens in requests, and validate them in your handler. (3) Secret headers - Include a shared secret in the webhook URL or headers. (4) Private endpoints - Use Azure Private Link to keep traffic within your VNet. (5) IP filtering - Restrict webhook access to Event Grid IP ranges. For production, combine Azure AD authentication with private endpoints for defense in depth.

Event Grid implements exponential backoff retry with up to 24 hours of attempts for HTTP webhooks. Retry intervals start at 10 seconds and increase to 1 hour maximum. If all retries fail, the event is dead-lettered to an Azure Storage container (if configured) for later analysis or reprocessing. Dead-letter events include original event data plus error information. For Azure destinations (Functions, Logic Apps, Event Hubs), Event Grid retries continuously for 24 hours. Configure dead-letter destinations and Azure Monitor alerts to track delivery failures.

Yes, Azure Event Grid can integrate with on-premises systems through several patterns: (1) Webhooks - Expose on-premises endpoints via Azure API Management, Application Gateway, or a reverse proxy with proper security. (2) Hybrid Connections - Use Azure Relay Hybrid Connections to receive events without opening inbound firewall ports. (3) Event Grid Namespaces - Use MQTT protocol for IoT/OT scenarios with devices behind firewalls. (4) Azure Functions proxy - Deploy a Function that receives events and forwards to on-premises via VPN/ExpressRoute. Always secure these connections with authentication, encryption, and network isolation.