cloud-architecture-azurelisted

# Azure Architecture Knowledge Service-level detail for an Azure-backed project. Pairs with Melvin's cloud-agnostic diagnostic checklist (traffic pattern, state location, SLAs, blast radius, cost explosion, coordination, limits, observability) — this pack supplies the Azure-specific answers. For concrete topology, cost tiers, and stack context, read the architecture docs named in `CLAUDE.md`. For deep service docs use the `microsoft-docs` skill (Microsoft Learn MCP). ## Compute ### Azure Container Apps (ACA) - Serverless containers on managed Kubernetes (KEDA-based autoscaling). **Scale to zero** is the cost win — and the latency trap: a scaled-to-zero app pays a **cold start** (image pull + container start, seconds) on the next request. For latency- sensitive services set **`minReplicas >= 1`** (a warm replica) — the ACA analogue of provisioned concurrency. - **Revisions:** each config/image change creates a revision; traffic-split between them for blue/green and canary. The phantom "revision nobody recognizes" is usually a stale active revision still taking traffic — check the revision list and traffic split. - **Scaling:** KEDA scale rules on HTTP concurrency, CPU/memory, or queue length (Service Bus, etc.). Per-app and per-environment replica ceilings apply — know them before you bet a spike on autoscale. - **Good for:** microservices, HTTP APIs, queue/event workers. Prefer it over AKS unless you genuinely need Kubernetes primitives. ### AKS - Full