This page lists what is on main. Release notes are in release-notes.md. There are no dates and no version numbers. Anything not built is marked planned.

Unreleased

Infrastructure

• A state backend in Azure: one resource group, one storage account and the tfstate-dev container, created with infrastructure/bootstrap/bootstrap.sh.
• Two Terraform roots for dev, one per layer, each with its own state file.
• The persistent root is applied. It defines the resource group, the Azure Monitor and Log Analytics workspaces, Application Insights, the action group, the four burn-rate alerts, the Key Vault, the identity for pull request plans and the identity of the sample service with its Key Vault roles.
• The session root defines a virtual network, a node subnet and an AKS cluster with Azure network policy, the managed ingress, managed Prometheus, workload identity and Entra ID sign-in with Azure RBAC. It was applied in Session 1 and Session 2 and destroyed after each.
• infrastructure/session.sh creates, destroys and checks the session layer, and runs the bootstrap step.
• session.sh down --yes repeats the destroy once after a wait. The repeat has not been run against a failed destroy.
• The budget with alerts for dev is set in Azure and is not defined in Terraform. Defining it in Terraform is planned.
• The environments staging and prod are planned.

Kubernetes platform

• A sample service in Go with health endpoints, a settable latency and error rate, and its Dockerfile.
• A Helm chart for the sample service.
• Namespaces, quotas and network policies.
• A script that checks the chart and the cluster rules in a local kind cluster.
• The sample service ran on AKS and was served through the managed ingress.
• The sample service mounts a Key Vault secret through workload identity.

GitOps and CI/CD

• The Argo CD bootstrap step of session.sh, and the Applications root, cluster-rules, sample-service and opentelemetry-collector.
• Six GitHub Actions workflows: Terraform checks, Terraform plan, Service, chart and manifest checks, Security scans, Publish image, and Open pull request for the image tag. See workflows.md.
• The image is published to ghcr.io, and a pull request sets its tag in gitops/dev/values/sample-service.yaml.
• A bad image tag was deployed and reverted through a pull request in Session 2.

Observability

• A metrics endpoint on the admin port of the sample service, scraped by managed Prometheus through a PodMonitor.
• Traces from the sample service through the OpenTelemetry Collector to Application Insights.
• Query files under monitoring/: six PromQL files and three KQL files.
• SLOs and four burn-rate alerts. See slos.md. The availability alerts fired and sent mail in Session 2. The latency alerts have not been tested.
• Dashboards are planned. Saved searches and an alert for missing data are planned.

Security

• Roles and secret rules: security.md.
• Trivy scans in CI.
• Pull request plans sign in to Azure through OIDC, with a federated credential that includes the owner and repository IDs.
• Sign-in to the cluster with Entra ID and Azure RBAC.

Incident management

• Ten runbooks. Three were exercised in Session 2: availability-burn, bad-image-tag and node-drain. See runbooks/README.md.
• An incident process with a severity table: incident-process.md.
• A postmortem template and one postmortem: postmortems/README.md.

Load and failure tests

• A k6 script with the scenarios smoke, baseline and ramp.
• The baseline and ramp scenarios were run against the cluster in Session 2.
• Experiments run in Session 2: a pod delete, a node drain, error injection and a bad image tag. The drain does not complete on the two node cluster (#98).

Documentation

• Decision records ADR 0001 to ADR 0030.
• The architecture document, the plan, the project history and the workflows page.
• Run sheets and results for Session 1 and Session 2.
• Files for a documentation site built from the docs/ folder (#51).