tomspark wrote:
Really good FYI to flag — thanks for posting it. You're 100% right, and this is a known footgun whenever your router uses an in-stack Pi-hole as its DNS. The upgrade has to talk to Docker Hub / GHCR to pull image layers, and during the brief window where Pi-hole's own container is being recreated, your router has no working resolver, so the pulls bomb out. The temporary 1.1.1.1 dance is the right workaround until we automate it. Logging this as a real product gap — we should either (a) detect when the box's own Pi-hole is the network's DNS and walk the user through a router fallback before the upgrade, or (b) make sparkbox update use an external resolver for itself during the pull window so it doesn't depend on the Pi-hole it's about to restart. Adding to the queue; I'll post here when it ships. Thanks again for the clear write-up — exactly the kind of FYI that helps every other Pi-hole user.