Linux · Platform · Infrastructure

Systems behind the products

A self-hosted engineering environment built around explicit trust boundaries, isolated CI, recoverable application/data infrastructure, independent telemetry and evidence-driven operations.

LinuxKVM / ProxmoxOPNsenseGitHub ActionsPrometheusGrafanaDebian / APT

Operating principle

Source, runtime, data and telemetry should not collapse into one machine.

The infrastructure evolved from a conventional homelab into a more deliberate engineering platform: application workloads, databases, CI, observability, storage and privileged control have separate roles and change boundaries.

GitHub carries reviewed intent and recovery context, but a merged file is not treated as proof that a live router, VM or service actually changed. Consequential infrastructure work requires a target, before-state evidence, rollback path and post-change verification.

Current KVM baseline

Build, runtime and data roles are separated before workloads move.

CI / build

Dedicated CI VM

Dedicated Debian KVM guest · self-hosted GitHub Actions · Docker/Buildx/Compose · internal target isolation

Application runtime

Application runtime VM

Dedicated Debian KVM guest · boot/SSH/QEMU agent verified · clean runtime target

Database

Database VM

Dedicated Debian KVM guest · boot/SSH/QEMU agent verified · clean PostgreSQL target

Current boundary

The v2 VM platform is provisioned and verified. Core and DB are clean targets: normal application workloads and PostgreSQL restoration are not yet claimed as completed. The CI guest remains explicitly separate from application/database work.

Four engineering planes

Each system answers a different operational question.

01

Operate safely

Homelab Ops

KVM/Proxmox, OPNsense, trust zones, DNS, SSH, storage, recovery and explicit live-change contracts.

02

What is happening?

Observability

Prometheus, Alertmanager, Blackbox Exporter and Grafana on an independent local telemetry plane.

03

What is true?

Control Station

Immutable observations, provenance, source authority, freshness, conflicts and bounded read-only collectors.

04

Deliver consistently

Node Platform

Versioned operator tooling packaged as signed Debian artifacts with APT canary/stable rollout and rollback.

CI isolation

The build runner is useful because it is not another general-purpose server.

GITHUBReviewed source / workflow
→
ISOLATED CIDedicated CI VMbuild · test · artifact · no normal app/DB workload
→
RELEASE TARGETseparately authorized runtime

Network controls were reverified after restoring SSH administration: protected private targets remain blocked from the CI guest. Convenience does not silently widen the runner's trust boundary.

Control Station vs Observability

“What is true?” and “what is happening?” are different questions.

Control Station stores evidence about stable infrastructure facts: identity, topology, source authority, freshness, supersession and conflict. A failed probe does not magically prove that something disappeared.

Observability remains a separate plane for metrics, probes, dashboards and alerts. This avoids letting a dashboard, a runbook or one stale Markdown file quietly become the authoritative infrastructure inventory.

Node Platform

Operator tooling is packaged, versioned and roll-backable.

The Rikaha Node Platform turns shared Linux operator/profile tooling into a signed Debian package delivered through APT, with immutable GitHub Release artifacts, canary/stable channels, explicit profile apply and rollback.

It also includes read-only SSH/network diagnostics and local receipts while deliberately avoiding a privileged background daemon or automatic remote mutation.

Honest boundary

Hands-on infrastructure, not hyperscale theatre.

This work supports Linux/platform/DevOps positioning through real virtualization, routing, DNS, CI, telemetry, packaging, deployment and recovery experience.

It does not claim multi-region cloud infrastructure, production Kubernetes at scale or a large corporate on-call environment. The useful evidence is ownership of a small but deliberately engineered platform.

Selected work

Product engineering and platform engineering are part of the same portfolio.

Back to all work