From 42aad8a117de0827dac49bc6baa105019391b514 Mon Sep 17 00:00:00 2001
From: "l.kirchner" <l.kirchner@noreply.gitea.luki-net.org>
Date: Fri, 22 May 2026 00:11:49 +0200
Subject: [PATCH] fix(build): add DNS server prompt for static IPs + better
 network diagnostics

For DHCP setups DNS is delivered with the lease, but with a static IP the
container inherits /etc/resolv.conf from the PVE host - which is often
unreachable from the container's VLAN.

Changes:
- New NAMESERVER prompt (only shown for static IPs, defaults to gateway)
- pct create now passes --nameserver when set
- Network wait loop tests L3 and DNS separately so failures point at the
  actual cause (no route to gateway vs. bad DNS server)
- Refactored pct create args into an array for cleaner conditional flags
---
 lib/build.func | 73 +++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 54 insertions(+), 19 deletions(-)

diff --git a/lib/build.func b/lib/build.func
index df87c30..37f9e70 100644
--- a/lib/build.func
+++ b/lib/build.func
@@ -34,6 +34,7 @@ DEFAULT_BRIDGE="${DEFAULT_BRIDGE:-vmbr0}"
 DEFAULT_TEMPLATE_STORAGE="${DEFAULT_TEMPLATE_STORAGE:-local}"
 DEFAULT_ROOTFS_STORAGE="${DEFAULT_ROOTFS_STORAGE:-local-lvm}"
 DEFAULT_TEMPLATE_PATTERN="${DEFAULT_TEMPLATE_PATTERN:-debian-12-standard}"
+DEFAULT_NAMESERVER="${DEFAULT_NAMESERVER:-}"
 
 # ── error trap ───────────────────────────────────────────────────────────────
 _on_error() {
@@ -64,8 +65,8 @@ show_header() {
 
 # ── prompts ──────────────────────────────────────────────────────────────────
 # Each prompt is skipped if the corresponding variable is already set in env.
-# VLAN_TAG uses ${VLAN_TAG+x} so that explicitly setting it to "" via env
-# skips the prompt (= "no VLAN, don't ask").
+# VLAN_TAG and NAMESERVER use ${VAR+x} so that explicitly setting them to ""
+# via env skips the prompt (= "no VLAN" / "inherit DNS from host").
 prompt_lxc_config() {
   if [[ -z "${CTID:-}" ]]; then
     read -rp "Container ID [auto]: " CTID
@@ -112,6 +113,13 @@ prompt_lxc_config() {
   if [[ "$IPCFG" != "dhcp" && -z "$GATEWAY" ]]; then
     read -rp "Gateway: " GATEWAY
   fi
+  # DNS: only meaningful for static IPs (DHCP gets DNS from the lease).
+  # Default suggestion is the gateway, which doubles as DNS in most homelabs.
+  if [[ "$IPCFG" != "dhcp" && -z "${NAMESERVER+x}" ]]; then
+    local default_ns="${DEFAULT_NAMESERVER:-$GATEWAY}"
+    read -rp "DNS server [$default_ns] (empty = inherit from PVE host): " NAMESERVER
+    NAMESERVER="${NAMESERVER:-$default_ns}"
+  fi
 }
 
 # ── template ─────────────────────────────────────────────────────────────────
@@ -142,36 +150,63 @@ create_lxc() {
     net_opts+=",ip=$IPCFG,gw=$GATEWAY"
   fi
 
+  # Assemble pct args as an array so conditional flags stay clean.
+  local pct_args=(
+    "$CTID" "$TEMPLATE_STORAGE:vztmpl/$TEMPLATE"
+    --hostname "$CT_HOSTNAME"
+    --cores "$CORES"
+    --memory "$RAM"
+    --swap 512
+    --rootfs "$ROOTFS_STORAGE:$DISK_SIZE"
+    --net0 "$net_opts"
+    --features nesting=1
+    --unprivileged 1
+    --onboot 1
+    --ostype debian
+    --password "$ROOT_PASSWORD"
+  )
+  [[ -n "${NAMESERVER:-}" ]] && pct_args+=(--nameserver "$NAMESERVER")
+
   msg_info "Creating unprivileged LXC $CTID ($CT_HOSTNAME)..."
   msg_info "  net0: $net_opts"
-  pct create "$CTID" "$TEMPLATE_STORAGE:vztmpl/$TEMPLATE" \
-    --hostname "$CT_HOSTNAME" \
-    --cores "$CORES" \
-    --memory "$RAM" \
-    --swap 512 \
-    --rootfs "$ROOTFS_STORAGE:$DISK_SIZE" \
-    --net0 "$net_opts" \
-    --features nesting=1 \
-    --unprivileged 1 \
-    --onboot 1 \
-    --ostype debian \
-    --password "$ROOT_PASSWORD"
+  [[ -n "${NAMESERVER:-}" ]] && msg_info "  dns:  $NAMESERVER"
+  pct create "${pct_args[@]}"
   msg_ok "LXC $CTID created"
 
   msg_info "Starting LXC..."
   pct start "$CTID"
   sleep 5
 
-  msg_info "Waiting for network..."
-  local i
+  msg_info "Waiting for network + DNS..."
+  local i ping_ok=0 dns_ok=0
+  # We test L3 (gateway/internet) and DNS separately so we can give a useful
+  # error message instead of a generic "network never came up".
   for i in {1..30}; do
+    if [[ $ping_ok -eq 0 ]] && pct exec "$CTID" -- bash -c "ping -c 1 -W 2 1.1.1.1 >/dev/null 2>&1 || ping -c 1 -W 2 ${GATEWAY:-1.1.1.1} >/dev/null 2>&1"; then
+      ping_ok=1
+    fi
     if pct exec "$CTID" -- getent hosts deb.debian.org >/dev/null 2>&1; then
-      msg_ok "Network up"
-      return 0
+      dns_ok=1
+      break
     fi
     sleep 2
   done
-  msg_err "Network never came up. Check bridge/VLAN/gateway settings."
+
+  if [[ $dns_ok -eq 1 ]]; then
+    msg_ok "Network + DNS up"
+    return 0
+  fi
+
+  if [[ $ping_ok -eq 1 ]]; then
+    msg_err "L3 connectivity OK but DNS resolution failed."
+    msg_err "Container can reach the internet but can't resolve names."
+    msg_err "Re-run with NAMESERVER=<dns-ip> set, or fix /etc/resolv.conf inside the LXC."
+  else
+    msg_err "Network never came up. Check bridge/VLAN/gateway settings:"
+    msg_err "  pct exec $CTID -- ip a"
+    msg_err "  pct exec $CTID -- ip r"
+    msg_err "  pct exec $CTID -- ping -c2 ${GATEWAY:-<gateway>}"
+  fi
   exit 1
 }