Inside the workshop / Software

The software.
Making it all useful.

The platforms and services running on the lab, what they do for me, and how I work to keep them dependable.

01 / The foundation

Different roles. A connected system.

Virtual machines separate the major jobs, containers package applications, and the website shares the public side of the workshop.

Virtualization

Proxmox VE

Virtualization for application hosting, game services, the web proxy, and monitoring.

Application containers

Docker

Containerized applications with persistent data and repeatable service definitions.

Web traffic & HTTPS

Nginx Proxy Manager

Routes approved web traffic to the appropriate service and manages HTTPS certificates.

The public website

Astro

The static website you're viewing, with source changes tracked in a private Git repository.

02 / Self-hosted services

The services I use.

A selection of the platforms and services I use for files, photos, gaming, and hands-on learning.

Files & collaboration

Nextcloud

Personal files and collaboration.

Photos & memories

Immich

A self-hosted home for photos and videos.

Metrics & monitoring

Grafana & Prometheus

Dashboards, metrics, and service checks.

Logs & troubleshooting

Centralized logging

Loki brings logs together for troubleshooting.

03 / AdGuard Home

DNS filtering with a second path.

DNS translates a name a device asks for into an address it can use. AdGuard Home checks those requests against filtering rules, helping block advertising and tracking domains before the device connects to them. It can also apply domain-based security and content policies.

See the two dedicated Raspberry Pi devices supporting DNS.

Both can answer DNS requests. The network supplies both resolver addresses to clients; the second is not a sleeping standby. Device behavior determines which resolver is used and how quickly it retries another. Neither DNS service needs the main Proxmox server to be running.

Try a DNS example

Illustration only. These examples do not test your device or contact my network.

Allowed request

A normal website can resolve.

  1. Your device asks for a website’s address.
  2. AdGuard checks the name against its rules.
  3. If allowed, it returns a cached answer or asks an upstream DNS provider.
A matching rule stops the lookup

A known tracker is filtered.

  1. A page asks your device to contact a tracking domain.
  2. The domain matches an enabled blocking rule.
  3. AdGuard returns a blocking response rather than the tracker’s usable address.
A second resolver remains available

One node can be maintained at a time.

  1. One AdGuard service is unavailable or undergoing maintenance.
  2. The other independently continues resolving and filtering.
  3. Clients configured with both can use the surviving resolver; retry timing varies by device.

How I reduce the chance of a DNS outage.

Redundancy, maintenance safeguards, monitoring, and recovery work together.

Independent hardware

Two wired Raspberry Pis keep DNS separate from the main virtualization host. A server reboot does not inherently require taking DNS down.

Consistent filtering

Configuration synchronization keeps shared rules aligned. If synchronization stops, each resolver keeps serving its existing configuration.

Backups without a double outage

Scheduled system backups check that the other resolver can answer first. A shared lock prevents the two backup jobs from pausing DNS at the same time, and recovery hooks restart and check the service afterward.

Monitoring & recovery copies

Checks watch DNS answers, filtering, service health, and backup results. Off-device configuration and system backups provide a recovery path beyond simply replacing a failed Pi.

Tested with one service stopped.

In controlled tests, each surviving resolver answered normal requests and blocked a test domain while its peer was down. That demonstrates independent resolver operation; it does not guarantee instant failover on every client.

What this setup does—and does not—protect against

UPS-backed power helps bridge short power interruptions, but shared power, the gateway, and network connectivity remain common dependencies. Two DNS servers do not protect against every outage.

DNS filtering applies to devices using these resolvers. It does not remove every advertisement, inspect downloaded files, or replace endpoint protection. Ads served from the same domains as the content, including many video ads, can remain. Devices using their own DNS or a VPN may bypass the local filtering.

Whole-system shutdown and recovery coordination remains part of the ongoing build.

04 / Protection & recovery

A recovery plan needs evidence.

Snapshots, backup copies, and application database backups cover different parts of recovery. Monitoring helps identify a failed job; restore tests check whether the saved data can actually be used.

Recovery means testing it.

I’ve restored the application server from backup into a separate test VM and verified the recovery. RAID, snapshots, and backups each do a different job; the restore test checks that the pieces can work together.

Acceptance still open

Power & recovery coordination

UPS monitoring is in place. Whole-system automatic shutdown and recovery still need integration and testing against the finished setup.

Follow the build