Implementation Status
This page is the single source of truth for what is built, what is unit-tested, and what has been proven on real hardware. The README, the roadmap, the tool overview and every tool page point here. If another page disagrees with this one, this page wins and the other page is a bug.
IMPORTANT
Nothing in this repository has been proven on a live Azure Local cluster yet. "Implemented" means the scripts exist and the command-construction, parsing and refusal logic pass Pester tests with mocks and fixtures. It does not mean the end-to-end run has been executed on a cluster. Treat every tool as a pre-release until the hardware column says otherwise.
Definitions
| Column | Meaning |
|---|---|
| Implemented | The scripts for install, run and collect exist and are wired together. |
| Unit-tested | Pester tests cover the script structure, parameter handling, command construction and result parsing. No cluster, no network, no Azure calls. |
| Proven on hardware | Someone ran it end to end on a real Azure Local cluster and kept the evidence. No for every tool today. |
Tools
| Tool | Implemented | Unit-tested | Proven on hardware | Notes |
|---|---|---|---|---|
| VMFleet (DiskSpd) | Yes: pipeline (pre-check, install, deploy, test, monitor, collect, report, cleanup), monitoring collectors, base-image preparation | Yes: 143 tests (scripts, profiles, base image, failure-injection opt-in guard) | No | The deploy phase passes VM count, base VHD and CSV to the fleet cmdlet; the per-VM memory, vCPU and data-disk values are resolved and logged only. Verify the sizing options of your VMFleet version on the cluster. |
| fio | Yes: install, start, collect, Ansible role and playbook | Yes: 62 tests | No | Linux VMs only. |
| iPerf3 | Yes: start, collect, Ansible role and playbook (no install script: the Ansible role installs it) | Yes: 39 tests | No | Plain TCP/UDP only. It does not exercise RDMA; see Azure Local network guidance. |
| HammerDB | Yes: install, start, collect | Yes: 43 tests (1 skipped) | No | |
| stress-ng | Yes: start, collect, Ansible role and playbook (no install script: the Ansible role installs it) | Yes: 36 tests | No | Linux VMs only. |
Framework
| Component | Implemented | Unit-tested | Proven on hardware |
|---|---|---|---|
| ConfigManager, Logger, StateManager, CredentialManager, MonitoringManager | Yes | Yes (Config, Logger, State; Credential and Monitoring are not covered yet) | No |
| ReportGenerator (PDF via asciidoctor-pdf, DOCX via pandoc, XLSX via ImportExcel) | Yes | Partly: 3 tests, the external tools are not exercised | No |
| ResultAnalysis: offline HTML + CSV report, baseline comparison, regression exit code, trend | Yes | Yes: 27 tests on fixtures | Not applicable (offline), but the inputs come from runs that are unproven |
Day-2 smoke gate (Invoke-Day2Smoke.ps1, JSON schema) | Yes | Yes: 25 tests (schema, exit codes, fixtures) | No |
| FaultInjection (node pause, drive fault, guarded, opt-in) | Yes | Yes: 43 mocked tests | No (see failure injection) |
| CI (GitHub Actions and Azure DevOps): analyzer, Pester per suite, link check | Yes | Not applicable | Not applicable |
Known documentation gaps
- No script has a
-CredentialSourceparameter, and the tool pages no longer show one. The scripts take a-Credential(PSCredential) or a secret parameter such as-WorkspaceKey, then fall back to the environment variables and the Key Vault mapping (see Credential Management). For any other command example, the scriptparamblocks are the reference. - The
monitoring/andreports/folders are per tool (tools/<tool>/monitoring,tools/<tool>/reports/templates), not at the repository root.
What would change a "No" to a "Yes"
An owner runs the tool on a non-production cluster, keeps the output, and records the date, Azure Local release and result here. The suggested order is in the roadmap.