SaaS
A two-hour deploy of a single Node.js application replaced by independently deployed services on Kubernetes, with the monolith shrunk step by step rather than rewritten.
Every change, however small, meant building and deploying the whole application. Deploys took two hours, failed about one time in five, and the team had settled into releasing once a fortnight to limit the pain.
Traffic from large customers arrived in bursts at the start of each month, and the only way to cope was to over-provision the servers all month long.
How things ran when we started, and once the work shipped.
Approach
4 phases over 4 months, with a working demo at the end of every week.
Weeks 1–3
Analysed call graphs and database access to find boundaries, and agreed with the team to extract billing, notifications and reporting first.
Weeks 3–7
Stood up EKS with Terraform, a shared Helm chart and a GitHub Actions pipeline that builds, tests and deploys each service on merge.
Weeks 6–14
Extracted services behind an API gateway one at a time, moving their data with dual writes, and removed the old code paths once traffic had moved.
Weeks 14–16
Tuned horizontal pod autoscaling against the month-start peak and ran four workshops so every squad could build and ship a service on its own.
Services own their data and deploy on their own schedule, while the remaining monolith keeps running behind the same gateway until it is retired.
“They did not ask us to stop and rewrite. We kept shipping the whole time, and by the end each squad owned its own services and its own release schedule.”
Tell us where things stand today. On a 30-minute call an engineer will sketch how we would approach it, the timeline and a price range.