DevOps ·
Lab MCI, DevOps Engineer.
From a Kerja Praktik building CI/CD DevSecOps from scratch to running the lab's deployment platform, taking projects from manual, non-Docker deploys to automated Docker-based production.
Overview.
Since 9 February 2026 I have been the DevOps engineer at Lab MCI (Laboratorium Manajemen Cerdas Informasi), Informatics Department, ITS. It started as my 3-to-4-month Kerja Praktik (internship) and continued afterwards, alongside my final project. The lab runs most of its applications on a single on-premise server, with many developers working on several projects in parallel.
My job is to take their projects from nothing (often no Docker at all) to production with fully automated, Docker-based CI/CD.
Phase 1: CI/CD DevSecOps (Kerja Praktik)
Designed and implemented an on-premise CI/CD infrastructure that the lab’s developers can reuse:
- GitLab Runners on two servers and a reusable
.gitlab-ci.ymltemplate with build, test, scan, and deploy stages. - Security scanning shifted left into every pipeline: secret scanning (TruffleHog), SAST (Semgrep, SonarQube), and filesystem and container scanning (Trivy).
- Developer tooling for the lab: it-tools, Docmost (knowledge base), and Hoppscotch.
- Evaluated on a real lab project against ISO/IEC 25010 (security and reliability): the pipeline caught 5 third-party dependency vulnerabilities and completed in about 5 minutes end to end.
Phase 2: Deployment platform.
After the pilot, I moved on to a shared deployment platform for the whole lab:
- A central deploy engine and a shared CI component, so each repository only declares a small manifest instead of carrying its own deploy script.
- An isolated Docker network per project and environment (staging and production), with ingress through Cloudflare Tunnel.
- Images stored in the GitLab Container Registry, deployed blue/green with health checks.
- Migrated projects onto it, including a full production cutover of the multi-repository CoBEST system with its database restored and verified.
- Published developer documentation for the platform.
DataDock came out of this work too: the self-service database platform that removed the most repetitive requests from my queue.
Tech Stack.
GitLab CI/CD, GitLab Runner, Docker, Docker Compose, Cloudflare Tunnel, nginx, SonarQube, Trivy, Semgrep, TruffleHog, Linux
Status: ongoing.
Next project