Heroku to EKS in Six Weeks

Client: Early-stage product company (anonymized)
Package: Launch
Timeline: 6 weeks
Outcome: Production traffic on EKS with GitOps, TLS, baselines for metrics/logs, and a live handoff

Problem

Heroku costs and dyno limits were becoming the roadmap constraint. The team had Docker locally but no production Kubernetes experience. They needed a boring, operable landing on EKS — not a service-mesh science project.

Approach

1. Scoped Launch: one production cluster, staging parity, two primary services first
2. Cluster + IAM baseline on EKS; ingress and certs; secrets handling agreed with least privilege
3. CI → build → GitOps (Argo CD) path the team could extend
4. Metrics/logs/alerts that cover deploy health and saturation — not fifty unused dashboards
5. Runbooks + pair sessions for the engineers who would own day-2 operations

Results

• Production cutover in week 6 within the agreed Launch scope
• Staging and prod promoted through the same GitOps path
• Team independent for routine deploys; optional follow-on for Managed Support left open

Takeaway

Launch works when scope is tight: apps, environments, and “must have day-2” items listed up front. Everything else stays a change order — that is how SMB price floors stay honest.