Testing using the Azure Cosmos DB Emulator in Azure Pipelines
When working on an application that interfaces with an Azure Cosmos DB backend you could use an actual Cosmos DB instance to test against but that’s a bit of a hassle as it involves setting up and configuring an actual instance. It turns out there is an emulator that runs in a docker container that you can test against instead.
Setting up the Task
First off you need to go to the Azure Pipelines Marketplace and install the emulator in your organization.
Now add the task to your
- task: azure-cosmosdb.emulator-public-preview.run-cosmosdbemulatorcontainer.CosmosDbEmulator@2 displayName: 'Run Azure Cosmos DB Emulator'
Note that the task mentions “public preview”, so this will likely need to change when the emulator becomes generally available.
Making the Connection
Now where you run the test you can pass through the
CosmosDbEmulator.Endpoint variable provided by the task, add in the auth key that is hardcoded in the emulator and disable TLS to prevent self-signed certificate errors (example is with Node.JS).
- script: yarn test displayName: 'Run tests using Cosmos DB emulator' env: ENDPOINT: $(CosmosDbEmulator.Endpoint) KEY: C2y6yDjf5/R+ob0N8A7Cgv30VRDJIWEHLM+4QDU5DE2nQ9nDuVTqobD4b8mGGyPMbIZnqyMsEcaGQy67XIw/Jw== NODE_TLS_REJECT_UNAUTHORIZED: '0'
Adding a Condition for Matrix Builds
The emulator is currently only supported on Windows, so you can either set this up in a separate job for Windows or use a condition for that step if you’re using a matrix build like this:
strategy: matrix: linux: imageName: 'ubuntu-latest' mac: imageName: 'macos-latest' windows: imageName: 'vs2017-win2016' ... - task: azure-cosmosdb.emulator-public-preview.run-cosmosdbemulatorcontainer.CosmosDbEmulator@2 displayName: 'Run Azure Cosmos DB Emulator' # Using this condition will ensure this step only runs on the Windows VM condition: eq(variables['imageName'], 'vs2017-win2016')
Outsite Azure Pipelines
It should be possible to use the emulator outside of Azure Pipelines with Docker for Windows but it would be a little more difficult to setup without the pre-packaged task. Consult the official documentation and the GitHub repo for how to go about doing this.