PATTERN Cited by 1 source

Self-service infrastructure vending machine

A platform pattern where engineers request environments by intent (not by raw resource specification), receive isolated configured infrastructure in minutes, and resources are automatically reclaimed on expiry. "Like any good vending machine: you say what you need, you get it, and when you're done, it goes away."

Structure

1. Template catalog — pre-vetted, hardened environment types (stable serverless on AWS, multi-cloud setup, Lakebase-autoscaling-preconfigured, etc.)
2. Intent-to-infrastructure mapping — user describes purpose → system selects and configures template
3. Provisioning engine — Terraform (or equivalent IaC) executes template against target cloud
4. State database — tracks every resource with owner, purpose, cloud/region, expiration
5. Lifecycle manager — TTL enforcement, notification chain, automatic teardown
6. Agent interface — MCP or API enabling programmatic access by AI agents and automation

Forces

• Engineers need isolated environments quickly for demos, reproductions, hackathons
• Platform teams need governance, cost attribution, and security
• Shared environments create interference, limit contention, and attribution opacity at scale
• Manual provisioning doesn't scale past ~1,500 engineers

Consequences

• Eliminates coordination overhead of shared environments
• Enables burst capacity (1,200 requests/day during internal events)
• Platform limits (catalogs, instances) stay within bounds via lifecycle management
• Security maintained via template vetting rather than per-request review
• Cost attribution becomes automatic via per-resource ownership metadata

Known Uses

System	Scale	Stack
systems/databricks-fevm	5,000+ users, 2,600+ deployments, 3 clouds	Databricks Apps + Terraform + Lakebase + MCP

Seen In

• sources/2026-07-23-databricks-self-serve-infrastructure-vending-machine — first wiki canonicalisation (2026-07-25)