← All deploys
ResilienceMilos17 Aug 2026

redu.cloud vs chaos monkey. High availability

An app, a managed Postgres and a managed ClickHouse are converted from single machines to high availability with one prompt on redu.cloud. Then an open-source chaos monkey destroys members ten times over while the service keeps serving. Not a simulation and not a sandbox: the machines are real, they are actually destroyed, and the billing is real.

  • redu-chaos-monkeyThe chaos tool doing the killing. MIT, dry run by default, kills only machines named in an explicit allowlist and re-checks the name at the moment of the kill.
Chapters
  1. 0:00Intro
  2. 0:21What is already running
  3. 0:47One prompt converts all three to high availability
  4. 3:15Arming the chaos monkey
  5. 6:39First strike, in detail
  6. 8:59Second strike
  7. 9:35Ten strikes, and what the load balancer did
  8. 11:05What is coming in the final video
Deploy it yourself

1. Connect the redu MCP:

Claude Codeclaude mcp add --transport http redu https://mcp.redu.cloud/mcp --client-id redu-mcp
Codexcodex mcp add redu --url https://mcp.redu.cloud/mcp --oauth-client-id redu-mcp

Deploy your own.

Add the MCP, point your agent at any repo, and it ships to a real EU cloud. EU data residency by default.

Get started at redu.cloud

More deploys

Deploy Ghost with managed MySQL and shared media
Agent Deploys

Deploy Ghost with managed MySQL and shared media

An AI agent deploys Ghost with a managed MySQL database and shared media space, then publishes a test post on redu.cloud.

redu.cloud vs chaos monkey. Resilience intro
Resilience

redu.cloud vs chaos monkey. Resilience intro

A chaos monkey deletes a real VM out from under a running app. Not a simulation and not a sandbox: the machine is gone. An AI agent works out what broke, rebuilds it, and reconnects it to the databases that survived.

Deploy WooCommerce as an auto-scaling cluster
Agent Deploys

Deploy WooCommerce as an auto-scaling cluster

An AI agent deploys a WooCommerce store on redu.cloud, upgrades it into an auto-scaling cluster, then proves members are added and removed under load.