Introducing messaging in your design should not hurt so much
Download as PPTX, PDF1 like649 views
The document introduces NServiceBus, a framework for building reliable and scalable distributed systems through messaging and workflow management. It highlights key concepts such as commands, events, endpoints, and error handling, emphasizing the importance of decoupling and asynchronous communication. Additionally, it covers the management of dead letter messages and alert mechanisms to monitor message flow and system health.
Introducing messaging in your design should not hurt so much
- 1. Sean Feldman – NServiceBus Saravana Kumar – ServiceBus360 Introducing messaging in your design should not hurt so much
- 2. + Understanding message flow in a large system
- 3. + Easy when it’s small and simple
- 6. Message Flow with Particular Platform
- 8. What's NServiceBus? Messaging and workflow code-based framework to build distributed systems that are • Simple, reliable, and scalable • Easy to work with and extend • Focus on domain problems and not the middleware
- 9. Why do I need NServiceBus? • Decoupling (RPC/WCF) • Asynchronous communication • Pub/Sub • Ability to breakdown monoliths into smaller components (services) RPC/WCF
- 10. Message Types Command • Known destination (unicast) • Instructs recipient to perform an action • Indicates sender knowledge of recipient Event • Unknown destinations (multicast) • Immutable representation of something that occurred in the past Command and Events are messages with Semantic Meaning. NServiceBus enforces the semantic of command and events. Message
- 11. NServiceBus Concepts • Endpoint is a logical entity that communicates with other Endpoints • Endpoint instances physical deployment of endpoints and scale-out • Host environment where endpoints are running • Transport an abstraction of the underlying technology to provide messaging capabilities • Handler unit of code to operate on a message. Stateless. • Saga handles long running processes. Stateful. Usually represent workflows and coordinates work for handlers.
- 12. But What’s an Endpoint? • A logical entity that communicates with other endpoints using messages • Usually equates to a process • Has a unique name • Contains message handlers • Deployed as a single or multiple instances
- 13. Request/Reply Pattern Endpoint A Endpoint B
- 14. Demo: Pub/Sub Pattern Subscriber A Publisher Subscriber B
- 15. See Message Flow
- 18. What are the errors and how to resolve those 1. What are the errors 2. How to resolve those
- 19. Messaging Over Raw ASB Sender Receiver Azure Service Bus namespace Entities management Maintaining connectivity Factories and clients Error handling Transient or not Retries and backoffs Serialization Receiving loop Error handling …
- 20. NserviceBus Error Handling • Error queue • Recoverability • Immediate and Delayed • Connected to DeliveryCount • Unrecoverable exceptions • Custom Recoverability Policy
- 21. ServicePulse
- 22. Service Pulse: Retrying failed messages
- 24. Investigate Dead letter Messages and take appropriate actions like Message Resubmission or Purging
- 25. How messages are Dead Lettered? Why? Messages get dead lettered due to various reasons HeaderSizeExceeded TTLExpiredException MaxTransferHopCountExceeded Specified by application Session id is null Investigate Dead letter Messages
- 26. Investigate Dead letter Messages Can I see what is in the message content? Take a look at Message Properties Why this message is dead lettered?
- 27. Resubmit or Delete Dead Letter Messages Decide to resubmit or delete the message from the DLQ I want to resubmit to a Queue / Topic
- 28. Resubmit or Delete Dead Letter Messages Can I automate this process? I want to resubmit all messages dead lettered due to a particular reason to the same queue
- 29. Understanding the urgency, importance, and impact of these failures
- 30. Monitor Entities Is my Queue functioning properly? Are the messages read properly? Are there messages retained in the Queue for long time? Can I receive an alert if something is not as expected? If so, something is wrong with my downstream application..
- 31. Receive Alerts to understand the state of your entities What are the possible channels, I can receive alerts? Channelize alerts to specific teams in your organisation
- 32. Receive Alerts to understand the state of your entities What do I see in the alerts? Can I receive categorized alerts? Can I receive alerts for specific parameters? Periodic status reports will help better to understand the health of entities
- 35. +
Editor's Notes
- #7: Message Flow with NServiceBus & Particular Platform
- #16: Screenshot is from Labs, not Insight
- #17: https://github.com/ParticularLabs/RoutingVisualization
- #20: ??? Might not belong here at all
- #21: Error queue vs dead-letter queue
- #23: Exception Type and Stack Trace - groups messages both by exception type and stack trace. It is the default way of categorizing failed messages. Message Type - groups messages by message type. Endpoint Address - groups messages by endpoint address where the failure occurred.
- #35: Zoom into PaymentProcessor to explain how errors affect everything