Isaac Sneddon-Jenkins

Infrastructure

Self-hosted infrastructure migration

Moved Umi's client sites onto infrastructure we control and cut builds from 15 minutes to 2–3.

Role
Lead developer
Year
2025
Stack
  • Docker
  • Node
  • PostgreSQL
  • GitHub Actions
  • Next.js

Results

What changed

  • 15 min to 2–3 min

    Build time per deploy after moving to cached container builds

    Umi Digital

Why move off managed hosting#

Every client site deployed through one shared pipeline. Builds queued behind each other and took around 15 minutes, and when one failed it was hard to see why. The goal was to own the infrastructure and make deploys boring.

What changed#

  • Each site builds from a shared, cached base image
  • CI runs per site, with dependency caches kept between runs
  • Runtime and postgres live on servers we manage
Dockerfiledocker
# Shared base: dependencies cached once, reused by every site
FROM node:22-alpine AS deps
WORKDIR /app
COPY package.json pnpm-lock.yaml ./
RUN corepack enable && pnpm install --frozen-lockfile


FROM deps AS build
COPY . .
RUN pnpm build

Builds now take two to three minutes, and the setup is documented so the rest of the team can run it without me.

Process

How it was rolled out

  1. Audit

    [What was measured first: build times, costs, failure points.]

  2. Pilot one site

    [Which site went first, and what it proved.]

  3. Migrate in batches

    [How the rest moved without downtime.]

  4. Hand over

    [Runbooks, alerts and who owns what now.]

Hiring, or have a project in mind?

Open to lead and senior engineering roles, and to freelance work.

Email me

More work

Related projects

Shared npm packages

2025

Versioned UI and config packages, so a fix lands everywhere with one bump.

  • TypeScript
  • npm