Fabrice Semti

DevOps Engineer, IT-thinkerer, dad, friend, gamer

[Fabrice Semti @FabriceSemti]

Oh boy...his attitude is shocking...and I bet he wonders, why people become antisemitic...well, I doubt it is jealousy...😔 https://t.co/qNfRvGNyws

[Fabrice Semti @FabriceSemti]

If you run Proxmox + Ceph at home, this role pays for itself in consistency alone.

🔗 Repo: https://t.co/q4An3aW924

#Proxmox #Ansible #Homelab #Ceph #InfrastructureAsCode #SelfHosted #DevOps

[Fabrice Semti @FabriceSemti]

Creating VMs on Proxmox through the web UI works for one-off. For a home DC with 3 nodes and Ceph storage, it's inconsistent and slow.

I built an Ansible role to automate it. Now every VM creation is ...

[Fabrice Semti @FabriceSemti]

The best part: this closes the loop for full-stack homelab automation.

Proxmox layer becomes declarative — same as K8s. VMs become cattle, not pets.

Spin up identical test VMs: `ansible-playbook create_...[truncated]

[Fabrice Semti @FabriceSemti]

Lesson: this closes the loop. Your infra is declarative, your apps are declarative — why not your monitoring?

🔗 PR: https://t.co/0gtGjGoyQr

#kubernetes #homelab #gitops #uptimekuma #autokuma #selfhosted

6/6

[Fabrice Semti @FabriceSemti]

Adding an Uptime Kuma monitor for every service manually? Your homelab deserves better.

I set up an automated pipeline that generates Uptime Kuma monitors directly from Kubernetes HTTPRoutes — reviewed in the PR, synced by ArgoCD, reconciled by AutoKuma.

🧵 1/6

[Fabrice Semti @FabriceSemti]

The generator uses yq (jq-for-YAML) to query HTTPRoutes, supports display-name overrides, and falls back to https://t.co/CyfZ27xJuf annotations.

50+ services covered: media stack (Sonarr, Radarr, Jellyfin), infra (ArgoCD, Grafana, Prometheus), and utilities.

5/6

[Fabrice Semti @FabriceSemti]

If you run Proxmox + Ceph at home, this kind of containerised, workflow-driven orchestration is a game changer.

🔗 Repo: https://t.co/G7j50IthTe

#homelab #kestra #proxmox #ceph #selfhosted

6/6

[Fabrice Semti @FabriceSemti]

Powering down a 3-node Proxmox cluster with Ceph storage is a recursive dependency nightmare.

VMs depend on hosts. Ceph depends on all hosts. K8s depends on VMs.

I built ordered shutdown/restart workflows in Kestra to handle it safely.

🧵 1/6

[Fabrice Semti @FabriceSemti]

The clever part: SSH keys injected via Kestra secrets, not baked into scripts. Zero agent install on Proxmox hosts.

Flows sync from GitHub — GitOps for your shutdown sequence. Update the scripts in the repo, next run picks them up.

5/6

[Fabrice Semti @FabriceSemti]

This replaces WireGuard configs + port forwards + DDNS with one declarative setup.

If you run K8s at home and use Tailscale, infra routers are the missing link.

🔗 PR: https://t.co/eiHVah8NaT

#homelab #kubernetes #tailscale

6/6

[Fabrice Semti @FabriceSemti]

Access your entire home network from anywhere with Tailscale subnet routers on Kubernetes.

I run a homelab on 10.0.0.0/16. The problem isn't reaching my cluster — it's reaching my NAS, Proxmox, and media servers from a hotel or coffee shop.

🧵 1/6

[Fabrice Semti @FabriceSemti]

From any random network:
• ssh 10.0.20.5 → Proxmox
• 10.0.10.50:8096 → Jellyfin
• grafana.home.internal → cluster dashboards

All like I'm plugged into my home switch. Managed via ArgoCD — GitOps rollbacks included.

5/6