Storage Migration Dry-Run Tool

A non-destructive, read-only dry-run utility for planning on-prem storage and datastore migrations.

This tool is designed to answer one question before anything moves:

“If we try this migration, what will break — and how long will it take?”

It simulates storage migration scenarios using inventory data, applies configurable safety and risk rules, and produces a clear GO / NO-GO decision with supporting evidence.

---

Why this exists

Storage migrations fail for boring reasons:

• insufficient destination capacity
• snapshot sprawl
• unrealistic throughput assumptions
• hidden blast radius
• unclear outage windows

This tool focuses on preventing those failures by running a rehearsal that:

• does not modify infrastructure
• does not require downtime
• produces artifacts suitable for change review, operations, and management

It is intentionally conservative and opinionated in favor of safety.

---

What it does

During a dry-run, the tool:

• Identifies candidate VMs on source datastores
• Calculates effective data to move, accounting for:
  • copy overhead
  • snapshots
  • thin provisioning penalties
• Evaluates destination datastore capacity with safety headroom
• Estimates migration duration using configurable throughput assumptions
• Flags risk conditions (snapshots, size, tooling, policy violations)
• Produces:
  • a structured GO / NO-GO decision
  • detailed findings
  • a placement plan
  • JSON and HTML evidence reports

No changes are made to infrastructure.

---

What it does not do (by design)

• No live migrations
• No storage writes
• No destructive actions
• No assumptions about cloud platforms

This is a planning and validation tool, not an execution engine.

---

Architecture overview

           +-------------------+
           |  Inventory Source |
           |  (mock / future)  |
           +---------+---------+
                     |
                     v
           +-------------------+
           |   Dry-Run Analyzer|
           |-------------------|
           | • capacity checks |
           | • risk rules      |
           | • time estimates  |
           +---------+---------+
                     |
                     v
           +-------------------+
           |   Evidence Output |
           |-------------------|
           | • JSON report     |
           | • HTML report     |
           | • GO / NO-GO      |
           +-------------------+

The provider layer is intentionally abstracted so inventory sources can be swapped without changing analysis logic.

---

Quick start

Requirements

• Python 3.9+
• No external services required

Setup

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Run a dry-run

python -m dryrun.cli

Or, if installed in editable mode:

pip install -e .
storage-dryrun

Reports are written to a timestamped directory under ./out/.

---

Configuration

config/plan.yaml

Defines what you want to evaluate.

migration:
  source_datastores:
    - DS_SITEA_01
  destination_datastores:
    - DS_SITEB_01
  auto_place: true

run:
  provider: mock
  inventory_path: ./mock_data/inventory.json

config/assumptions.yaml

Defines how conservative the dry-run should be.

Key areas:

• throughput assumptions
• overhead multipliers
• snapshot penalties
• free-space safety margins
• policy rules for GO / NO-GO

These values are deliberately configurable to reflect different environments and risk tolerance.

---

Output

Each run produces:

• dryrun_report.json
  Machine-readable results for automation, review, or archiving.

• dryrun_report.html
  Human-readable evidence pack suitable for:

  • change requests
  • migration reviews
  • post-planning documentation

Example highlights:

• overall decision (GO / NO-GO)
• estimated total migration time
• per-VM placement plan
• categorized findings (GREEN / YELLOW / RED)

---

Mock provider (current implementation)

The current provider loads inventory from JSON to enable:

• safe local testing
• deterministic behavior
• easy review of edge cases

This is intentional.

The provider interface is designed so a read-only vCenter provider can be added without changing the analyzer or reporting layers.

---

Known limitations

See docs/LIMITATIONS.md for a full list.

In short:

• inventory is mock-based
• throughput is assumed, not measured
• guest-level checks are limited

These constraints are deliberate to keep planning safe and repeatable.

---

Intended use cases

• Datastore evacuations
• Data center storage moves
• Migration window estimation
• Change impact analysis
• Pre-migration risk reviews

---

Philosophy

This tool is biased toward:

• clarity over cleverness
• safety over speed
• boring correctness over optimism

It is designed to fail loudly before production does.

---

License

MIT