ComputeProxmox VE 9.2
Intel i7-8700 · 24 GB DDR4 · Debian VM and LXC workloads
03 / Homelab
A platform for learning operations.
My homelab is a phased infrastructure project where I practice architecture, reliability, security, documentation, monitoring, and recovery.
Current platform
Small hardware, deliberate architecture.
The environment separates storage and application workloads while keeping access, backups, and monitoring part of the base design.
StorageMirrored and separated
2 x 2 TB ZFS mirror · Sanoid snapshots · SSD-backed Docker data
ServicesPrivate cloud stack
Samba · Nextcloud · PostgreSQL · Uptime Kuma · Prometheus · Grafana
AccessPrivate by default
Tailscale policies · remote Windows and macOS access tested
Full roadmap
Built in phases, guided by a documented roadmap.
Completed phases show the platform already in use. Remaining phases capture the next infrastructure and automation experiments.
Foundation
00
Hardware & hypervisorValidation · Proxmox VE · storage planning
complete01
Storage platformZFS mirror · datasets · SSD-backed workloads
complete02
File servicesDebian LXC · Samba · user permissions
complete03
Private remote accessTailscale · ACLs · cross-platform testing
completeServices & operations
04
Application hostdocker01 · Docker Compose · SSD data disk
complete05
Operations baselineUptime Kuma · Git-managed configuration
complete06
Backup & recoveryVM backups · app archives · restore testing
complete07
Private cloudNextcloud · PostgreSQL · Samba mounts
complete08
ObservabilityPrometheus · Grafana · three exporter paths
completeCurrent & next
08.5
Edge migrationPi services · network role · migration planning
queued09
Secure ingressNginx · reverse proxy · Cloudflare Tunnel
queued10
Portfolio V1Design · content · multi-page deployment
complete11
Automation & CI/CDRepeatable configuration · testing · pipelines
planned12
OrchestrationKubernetes · service scaling experiments
futureCurrent architecture
Two views of the running system.
The platform map shows where services live. The observability map follows the signals used to understand their health.
Workloads separated from storage.
One physical host, two focused guests, and a ZFS-backed data layer.
Physical hostProxmox VE 9.2Intel i7-8700 · 24 GB DDR4
CT 110fileserverDebian 13 · Samba · Tailscale
ZFS tank mirror2 × 2 TB · Sanoid snapshots
VM 100docker01Debian 13 · 4 vCPU · 6 GB RAM
Nextcloud 34PostgreSQL 17Uptime KumaPrometheusGrafana
Metrics become useful signals.
Infrastructure and container telemetry share one monitoring path, with endpoint checks alongside it.
Proxmoxpve-exporter
Debian hostsnode-exporter
ContainerscAdvisor
metrics
Collect & retainPrometheus 3.14
Query & visualizeGrafana 13.2
Independent health checksUptime Kuma monitors service endpoints
Operational proof
Tested beyond installation.
Each service is useful only when access, recovery, and visibility continue working outside the initial setup session.
Access
Remote file service proven
Samba access through Tailscale was tested from Windows and macOS clients in different cities, including account and share permissions.
Recovery
Backups restored, not assumed
The lab combines Proxmox backups, application archives, Git-managed configuration, and ZFS snapshots. An Uptime Kuma archive was restored successfully.
Visibility
Three telemetry paths unified
Proxmox, Debian hosts, and Docker containers report through dedicated exporters into Prometheus and Grafana, alongside endpoint health checks.
Edge & network lab
Experiments beyond the core host.
Camera, embedded, and network-service work extends the lab into physical devices and local infrastructure.
Validated
ONVIF camera network
Discovered three PoE cameras, validated ONVIF SOAP capabilities, and tested H.264 main and substream RTSP profiles.
ONVIFRTSPH.264PoE
Operational prototype
ESP32-S3 camera API
Built OV5640 health, status, and JPEG capture endpoints and completed a stable 100-request capture test for direct server ingestion.
ESP32-S3OV5640HTTP APIEmbedded
Operating principles
The installation is only the beginning.
I use each phase to practice the work that keeps a service understandable and dependable after it starts.
Make it observable
Health checks, host metrics, dashboards, and service-level visibility before adding more complexity.
Plan for recovery
ZFS snapshots, application backups, restore testing, and clear separation between data and workloads.
Limit exposure
Tailscale for private access today, with secure ingress and reverse-proxy work planned as its own phase.
Document the path
Commands, architecture decisions, changes, and troubleshooting notes are part of the deliverable.