This sample application consists of a "Contacts" API and an event processor that reads a Cosmos DB Change Feed and publishes domain events to an Azure Service Bus. The sample demonstrates how to use the Transactional Outbox Pattern in combination with Azure Cosmos DB to implement reliable/guaranteed messaging in your microservice-oriented application.
The source code is structured as follows:
- Contacts.Domain - Contact domain logic and interfaces
- Contacts.Infrastructure - Implementation of domain interfaces like repositories
- Contacts.Application - Implementation of CQRS pattern using MediatR library, validation and mapping logic
- Contacts.API - RESTful API consuming application logic components
- Contacts.EventsProcessor - Console application using the Change Feed Processor library to read the Cosmos DB change feed and publish domain events to Azure Service Bus
To run the sample application on your local machine, you need to have the following components installed:
You also need to create some Azure resources:
- Azure Cosmos DB
- Azure Service Bus
- Azure Function (to test event distribution via Service Bus - acts as a subscriber)
You can find a bicep file that will take care of creating these resources (and their dependencies) under Deployment/deploy.bicep.
To run the deployment script, open an Azure Cloud Shell, clone this repository to your environment, switch to the Deployment
folder and run:
export RESOURCE_GROUP=<resource-group-name> # name for your resource group
export LOCATION=<azure-region> # Azure region, for example westeurope, westus etc.
az group create -n $RESOURCE_GROUP -l $LOCATION
az deployment group create -f deploy.bicep -g $RESOURCE_GROUP -o none
az deployment group show -g $RESOURCE_GROUP -n deploy --query properties.outputs
The last command will output most of the relevant parameters that you need to replace in the following files:
Copy the ServiceBus connection string and Cosmos DB account key directly from the Portal! You won't find them in the output of the deployment.
Your Azure resource group should look like this:
WARNING: We running this in production, you should adjust the code to use Managed Identities to access Cosmos DB and Service Bus!
After adjusting the appsettings parameters, you can run the Contacts.API
and Contacts.EventProcessor
applications on your machine.
If you are using VS Code, there are predefined debug configurations to build and execute both projects. Just go to the "Debug" view and run both configurations.
When the API and the events processor console app are running, you can open a browser at https://localhost:5001/swagger/index.html (serving the Swagger UI).
To test the flow, you can use the Swagger UI to create a new contact. Use the POST /api/contacts operation, fill out the request body and execute the operation.
Alternatively, you can also issue a request via curl
:
curl -k -X POST "https://localhost:5001/api/contacts" -H "accept: */*" -H "Content-Type: application/json" -d "{\"firstName\":\"John\",\"lastName\":\"Doe\",\"email\":\"[email protected]\",\"description\":\"Sample description\",\"companyName\":\"Example Inc.\"}"
As soon as you add (or later update) contacts, you'll see events being published to the Azure ServiceBus topic and picked up by the Azure Function.
When you look at the Cosmos DB container, you'll find two documents for the create operation (a Contact
object and the corresponding domain event):
Tip: The Time-To-Live attribute for event documents is currently set to
36000
seconds. That means those documents will be deleted automatically after 10 hours. If you want to adjust the parameter to see how the documents will be cleaned up, go to Contacts.API/appsettings.json file and set a lower value forEvents.Ttl
(e.g.120
seconds).
To clean-up the Azure resources, use the following command:
az group delete -n $RESOURCE_GROUP