Manual QA Infrastructure¶
The E2E framework can provision real cloud infrastructure for manual QA without running automated tests. Environments stay alive until you explicitly destroy them (or our cleaner destroys it, every week), giving you direct access (SSH, kubectl, RDP) to inspect the agent in a realistic environment.
Prerequisites¶
Complete the one-time setup from the E2E testing guide before creating any environment.
If you do not want to rely on fake intake, and send data to the real Datadog backend, you should make sure you set a valid API in ~/.test_infra_config.yaml
Stack lifecycle¶
Each environment is a named Pulumi stack. Stacks are automatically prefixed with your OS username:
<username>-<stack-name>
# e.g. alice-aws-vm (default for the aws/vm scenario)
# alice-my-qa (with --stack-name my-qa)
Stacks persist until you destroy them. Run the matching destroy-* task when you are done to avoid leaving cloud resources running.
Agents are automatically tagged stackid:<stack-name> so you can filter metrics and logs in the Datadog UI to a specific environment.
Scenarios¶
| Scenario | Command | What it creates |
|---|---|---|
| AWS VM | dda inv aws.create-vm | EC2 instance + Agent |
| AWS Docker VM | dda inv aws.create-docker | EC2 + Docker + containerized Agent |
| AWS EKS | dda inv aws.create-eks | EKS cluster + Agent DaemonSet |
| AWS ECS | dda inv aws.create-ecs | ECS cluster + Agent |
| AWS KinD | dda inv aws.create-kind | EC2 + KinD cluster + Agent |
| Azure VM | dda inv az.create-vm | Azure VM + Agent |
| GCP VM | dda inv gcp.create-vm | GCP VM + Agent |
Pass --no-interactive to any scenario to disable the clipboard prompt and desktop notification, which is useful when running from an AI agent or CI context.
See Also¶
- Running E2E tests — automated test execution against the same infrastructure