01 / SOFTWARE + CLOUD

SYSTEMSTHAT OPERATESYSTEMS.

00 HARDWARE Rack, chassis, power and thermals. ProLiant nodes, disks, cooling.

[01.1]SYSTEM MAP

I LIKE SYSTEMS WITH LAYERS.

SM-01ACTIVE

INFRASTRUCTURE

Linux
Docker
Kubernetes
Helm
Cilium
Longhorn

SM-02BUILDING

CLOUD

AWS
EC2
VPC
Route 53
Lambda
API Gateway

SM-03EXPLORING

AUTOMATION

Terraform
Ansible
Python
Rust
CI/CD

SM-04ACTIVE

OBSERVABILITY

Prometheus
Grafana
Loki
Node Exporter

SM-05BUILDING

NETWORKING

DNS
TLS
Routing
Load balancing
Service networking

[01.2]SELECTED SYSTEMS

CASE 01

KUBERNETES LAB

A FOUR-NODE ENVIRONMENT FOR LEARNING THE WHOLE CLUSTER PATH

The cluster is a working laboratory, not a finished demo: one control plane and three workers used to trace how scheduling, networking, storage and services behave together.

01 / BUILD

Provision the nodes and assemble the platform from Linux upward.

02 / OBSERVE

Use Cilium and Hubble to inspect traffic instead of treating the network as a black box.

03 / REBUILD

Change one layer at a time, break assumptions, and carry the lesson into the next iteration.

KubernetesCiliumHubbleHelmArgoCDLonghorn

Terminal showing the current four-node Kubernetes cluster and healthy Cilium components — LIVE CAPTURE / CURRENT CLUSTER4 NODES · CILIUM 1.20.1 · HUBBLE RELAY

Grafana dashboard showing node, workload, pod, memory, CPU and network metrics for the Kubernetes cluster — OBSERVABILITY LAYER / GRAFANANODE · WORKLOAD · CPU · MEMORY · NETWORK

CONNECTED VIEW / 01.B

SEE WHAT THE CLUSTER IS DOING.

Grafana turns the same environment into an operational picture. Node pressure, workload count, CPU, memory and network activity can be compared before a small symptom becomes a blind troubleshooting session.

Node exporters
Prometheus
Grafana
Decision

THE FIRST IMPLEMENTATION IS RARELY FINAL.

CASE 02

NORDFIR

A RUST CONTROL PLANE FOR THE HOMELAB

Nordfir began as a terminal view for seeing which machines were alive and acting on them without moving between separate tools.

LONG-TERM DIRECTION

A unified homelab dashboard for machine control, live metrics and power awareness — including watt usage, Wake-on-LAN, and policies that can start or stop nodes when capacity changes.

NOW: Rust console
NEXT: Control + telemetry

RustSSHWake-on-LANPrometheusPower telemetry

CASE 03

TANNHAUSER

LINUX FROM SCRATCH / THE OPERATING SYSTEM BENEATH THE PLATFORM

A custom Linux system built from source with Linux From Scratch and systemd. Tannhauser trades convenience for visibility: toolchain, libraries, boot process and kernel configuration all become parts of the system that must be understood.

01 / TOOLCHAIN

Build the environment that will build the final system.

02 / USERSPACE

Compile and integrate the essential packages rather than inheriting a distribution.

03 / BOOT

Configure the kernel and reach a working login on real and virtual hardware.

Linux From ScratchsystemdKernel 6.18.5GlibcDocker

VIEW REPOSITORY

Tannhauser Linux From Scratch booted to a local TTY, showing kernel 6.18.5 and a successful Vincent login — REAL BUILD / LOCAL TTYLFS · SYSTEMD · KERNEL 6.18.5

[01.3]CURRENT FOCUS

ACTIVE

Kubernetes
Helm
Infrastructure troubleshooting

BUILDING

Nordfir
Homelab automation

EXPLORING

ArgoCD
Terraform
Ansible
AWS infrastructure

EKS
GitOps
Cloud architecture

A FAILED DEPLOYMENT IS STILL DATA.

[01.4]HOW I ENGINEER

P.01

BUILD → BREAK → UNDERSTAND → REBUILD

I often rebuild infrastructure once I understand the weaknesses of the previous design.

P.02

TRACE THE WHOLE PATH

When something fails, I prefer following the problem across application, container, network, operating system and hardware layers.

P.03

AUTOMATE AFTER UNDERSTANDING

I prefer understanding a process manually before automating it.

AUTOMATION SHOULD REMOVE REPETITION,
NOT UNDERSTANDING.