My personal playground — where I deploy, break, and rebuild things on my own terms.
Infrastructure is a big part of my life. I'm obsessed with automation and Kubernetes — and when you combine Kubernetes with ArgoCD, you get a level of automation where I can write code, deploy it, assign a domain on the way, and make it available to internal or external traffic in minutes. There are too few things for my perfectionist brain that could give me joy like typing a new domain name and seeing my code already serving there, working smoothly. I love trying new frameworks, open-source projects, and just playing with servers. I love the feeling of owning real compute — knowing my code is not limited by some cloud tier, watching containers pop up when load hits and scale back down when it's quiet. Optimization, efficiency, and performance are things I keep in mind constantly, both when writing code and when running it. I rarely make compromises there and always try to maximize what I can get out of every core and every byte. This cluster is where all of that happens.
12 cores · 24 GB RAM · 768 GB NVMe each
12 cores · 32 GB RAM · 2 TB NVMe each
20 cores · 64 GB RAM · 3 TB storage each
Everything starts with Ansible — it owns the metal. OS preparation, containerd, kubeadm, Calico CNI with eBPF, Longhorn for distributed storage, and ArgoCD bootstrap. Once ArgoCD is up, it takes over. Every component above the base layer is declared as an ArgoCD Application pointing at a private Git repo. Automated pruning, self-healing, sync-waves for dependency ordering. Ansible touches the machines, ArgoCD owns everything else. No overlap.
Two separate Traefik instances — a public gateway for internet-facing services and a private one behind Twingate for internal tools. Gateway API v1 with wildcard TLS on *.aliboyev.com. External DNS automation through Cloudflare, cert-manager with Let's Encrypt DNS-01 challenges. Private services like ArgoCD, Longhorn, and Grafana are never exposed to the internet.
Victoria Metrics for Prometheus-compatible metrics collection, paired with Grafana for dashboards. The full stack is deployed via ArgoCD with server-side apply for CRD-heavy resources.
A mix of self-hosted services and personal projects. Harbor as a private container registry, n8n for workflow automation, Temporal for workflow orchestration, Vaultwarden for password management, Label Studio for data labeling, and BuildKit for container image builds. Each app follows a consistent pattern — ArgoCD Application, Helm values, optional extra manifests, and a secret bootstrapping script. Some apps run on custom Helm charts I wrote from scratch.