Home lab serverPart 11 of 13

Build note

Pi-hole became the DNS backbone

The Pi-hole is a Raspberry Pi 4 on the home network. The workstation sends DNS queries to it, and Pi-hole filters domains before returning an answer. When name lookups stopped working, the first clue was that Wi-Fi was still connected. The failure was in DNS, not the radio link.

Pi-hole lives at 10.0.0.42. The router at 10.0.0.1 is a separate resolver; it does not forward through Pi-hole. That difference matters: the router can keep DNS working if Pi-hole is unavailable, but it does not apply the same block list.

The workstation has a resolver in front of Pi-hole

Applications do not send every lookup directly to the Pi. /etc/resolv.conf points to the local systemd-resolved stub at 127.0.0.53, which forwards queries to the DNS servers configured on the network connection. In this case, the intended path is:

Application → 127.0.0.53 (systemd-resolved) → 10.0.0.42 (Pi-hole)

That extra hop makes the visible symptom less obvious. A browser reports a name-resolution error. The Wi-Fi connection still says connected. resolvectl and the Pi-hole query log are the tools that show whether the request reached the filter.

One client can generate a lot of DNS

The workstation produced a burst of 41,872 DNS queries in about twenty minutes. The Pi-hole’s per-client rate limit was 1,000 queries per minute, and the burst included repeated lookups from applications such as Spotify and ChatGPT. When the client crosses the limit, Pi-hole can refuse requests, which looks to applications like DNS simply stopped answering.

The rate limit was raised to 10,000 queries per minute after checking the actual peak from this client. That is a local response to this host’s observed query pattern, not a number I would copy blindly to another network. The right limit depends on how many clients share the resolver and what happens when one of them loops.

Keep the fallback honest

The connection also had 10.0.0.1 as a fallback. That provides availability, with a trade-off: if the router is answering instead of Pi-hole, DNS works but filtering is bypassed. Checking the query log is more useful than assuming a configured first server is always the one handling requests.

The rule that would have prevented the next week

Test the whole resolver chain, not just whether the Pi-hole web page loads. Generate a fresh lookup, confirm it appears in the Pi-hole query log, and check which upstream resolver answered. A working DNS fallback is not the same thing as a working filter.

How this was made

  • Written by: Rupesh, from his Pi-hole query log, NetworkManager profile and systemd-resolved checks, with AI drafting and editing.
  • Query counts and rate limits: from the incident records and Pi-hole config.
  • AI-assisted sections: structure and copy editing.