DevOps on the Factory Floor
Lessons from running CI/CD pipelines, Kubernetes clusters, and monitoring stacks that also have to cooperate with PLCs, HMIs, and a very loud plant manager.
Why plants need pipelines
When I joined the industrial side of software, I expected the hardest problem to be uptime. It turned out to be change management. Factory operators will trust a control system that has run unchanged for three years long before they trust the shiny microservice you shipped on Friday. DevOps is the bridge.
Pipelines for PLC-adjacent services
Most of what I ship isn't a PLC program — it's the glue around it. The pipeline I lean on:
- GitHub Actions for build, test, image publish
- Tekton on the shop-floor cluster for in-plant rollouts
- OpenShift as the runtime because ops already knows it
- Prometheus + Grafana so the same dashboard can show line OEE and pod restarts
What I learned the hard way
Keep deploy windows aligned with the production schedule. A canary that works beautifully at 2 a.m. will still break if it lands during a shift changeover. DevOps in manufacturing is half automation, half calendar.
Treat the plant's change-control board as a first-class reviewer. They will catch things your linter never will.
