“We should move off AWS” is easy to say and easy to postpone. What makes it manageable is breaking it into an order where each step produces a result. Here’s how we do it.
1. Inventory and sort
List your workloads and split them into three piles: standard (EC2, EBS, VPC equivalents), portable with some work (containers, self-run databases) and deeply built into proprietary services. The first two piles move easily; the third needs the most thought.
2. Move the portable first
- Stand up equivalent VMs and networks on Kepler.
- Move stateless services and containers — they feel at home on open standards.
- Verify in a parallel environment before switching traffic.
3. Data and cutover
Sync data, run the two environments in parallel for a while, then shift DNS gradually. Moving traffic in stages keeps risk low and always leaves a way back until you’re confident.
Migrate in the order that gives you a win after each step, not a cliffhanger at the end.
— from Kepler’s migration notes
4. Verify and shut down
Once traffic is stable in the EU, export the whole setup as code so it’s documented and portable, and shut down the old one. Get in touch and we’ll help with the plan.