fix(dns): reload Technitium when zone sync changes files (TDD) [#469][#344]

WORKING.md

@@ -19,3 +19,10 @@ inventory #307 v2, and git PFVCluster + KNEL/pfv-bms)
 - [x] HACS check: NOT installed (answer to founder) [#344]
 - [x] Midea minisplit dongle found on WiFi (c4:39:60:59:0c:01 / 192.168.1.149) → #614 updated
 - [x] Ticket #614 created (dynamic temp management); IaC items filed on #454; audit log #298
+- [x] Kuma: dup ICMP monitor 199 deleted; pfv-bms-ha-http (221) added, UP/200 verified [#344]
+- [x] DNS perm-fix: sync-zones.sh reload-on-change (TDD, green) deployed + live-proven 75s propagation [#469]
+- [x] pfv-tsys1-lan / pfv-tsys9-lan DNS records (LAN path for NUT)
+- [x] UPS in HA: NUT renamed tsys1_ups/tsys9_ups, APC config entry via REST flow, sensors live (100%/Online) [#439]
+- [x] Tripp Lite dropped off USB on tsys9 (physical, #372) — config ready, binds on return [#439]
+- [x] SSH locked to key-only (workstation key), password path verified dead [#344]
+- [x] HACS 2.0.5 installed on box (custom_components gitignored) — founder GitHub auth pending [#614]

netinfra/dns-cluster-setup/sync-zones.sh

@@ -3,13 +3,15 @@
 # sync-zones.sh — rsync-based zone replication from primary to secondary
 #
 # Runs on the SECONDARY (netinfra-02). Syncs the zones/ directory from the
-# primary (netinfra-01) every 60 seconds. When a zone file changes, Technitium
-# detects the modification and reloads automatically.
+# primary (netinfra-01) every 60 seconds. When rsync changes any zone file,
+# Technitium is reloaded (container restart) so it serves the fresh zones —
+# Technitium does NOT reliably auto-detect externally modified zone files,
+# which left the secondary serving stale records (seen 2026-09-01 [#344]).
 #
 # This is used instead of AXFR-based zone transfer because Technitium's zone
 # transfer mechanism uses port 53 (standard DNS), but on the netinfra hosts
 # port 53 is Pi-hole and Technitium is on port 5300. rsync-based replication
-# avoids the port conflict entirely.
+# avoids the port conflict entirely. Native clustering tracked in [#469].
 #
 # Install as a systemd service/timer or run via cron:
 #   * * * * * /home/localuser/services/technitium/sync-zones.sh
@@ -21,23 +23,50 @@ CONFIG_DIR="${CONFIG_DIR:-/home/localuser/services/technitium/config}"
 ZONE_DIR="$CONFIG_DIR/zones"
 LOCK_FILE="/tmp/technitium-zone-sync.lock"
 LOG_FILE="${LOG_FILE:-/home/localuser/services/technitium/sync.log}"
+# Reload command; array form so tests can substitute a fake.
+read -r -a RELOAD_CMD <<< "${RELOAD_CMD:-docker restart tsys-dns}"
+export RELOAD_CMD
 
 log() { printf '[%s] %s\n' "$(date +%H:%M:%S)" "$*" >> "$LOG_FILE"; }
 
+# Sync zones from the primary and reload Technitium when anything changed.
+# Returns non-zero when rsync fails (reload failures are logged, not fatal).
+sync_and_reload() {
+  local changed rc zone_count changed_count
+  # --temp-dir avoids partial writes; --delete removes zones deleted on
+  # primary; --itemize-changes reports exactly what changed.
+  changed=$(rsync -az --delete --itemize-changes --out-format='%n' --temp-dir=/tmp \
+    "${PRIMARY_HOST}:$ZONE_DIR/" "$ZONE_DIR/" 2>> "$LOG_FILE")
+  rc=$?
+  if [ "$rc" -ne 0 ]; then
+    return "$rc"
+  fi
+  zone_count=$(find "$ZONE_DIR" -maxdepth 1 -type f | wc -l)
+  if [ -n "$changed" ]; then
+    changed_count=$(printf '%s\n' "$changed" | wc -l)
+    log "Sync complete: $zone_count zones, $changed_count changed — reloading Technitium"
+    if "${RELOAD_CMD[@]}" >> "$LOG_FILE" 2>&1; then
+      log "Technitium reload triggered"
+    else
+      log "WARN: reload command failed (rc=$?) — records may be stale until next change"
+    fi
+  else
+    log "Sync complete: $zone_count zones (no changes)"
+  fi
+}
+
+main() {
   # Prevent overlapping runs
   exec 9>"$LOCK_FILE" || exit 0
   flock -n 9 || { log "another sync is running; skipping"; exit 0; }
 
   mkdir -p "$ZONE_DIR"
 
-# rsync zones from primary. Use --temp-dir to avoid partial writes being
-# picked up by Technitium, and --delete to remove zones deleted on primary.
   log "Syncing zones from $PRIMARY_HOST..."
-if rsync -az --delete --temp-dir=/tmp \
-   "${PRIMARY_HOST}:$ZONE_DIR/" "$ZONE_DIR/" >> "$LOG_FILE" 2>&1; then
-  zone_count=$(find "$ZONE_DIR" -maxdepth 1 -type f | wc -l)
-  log "Sync complete: $zone_count zones"
-else
-  log "ERROR: rsync failed (rc=$?)"
-  exit 1
+  sync_and_reload || { log "ERROR: rsync failed (rc=$?)"; exit 1; }
+}
+
+# Allow sourcing (unit tests) without executing the sync
+if [[ "${BASH_SOURCE[0]}" == "$0" ]]; then
+  main
 fi

tests/unit/dns-zone-sync.sh

@@ -0,0 +1,128 @@
+#!/bin/bash
+# Unit tests for netinfra/dns-cluster-setup/sync-zones.sh [#469]
+# Verifies reload-on-change behavior: Technitium restart fires only when
+# rsync actually changed zone files, and never on clean syncs.
+set -euo pipefail
+
+PROJECT_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+SCRIPT="$PROJECT_ROOT/netinfra/dns-cluster-setup/sync-zones.sh"
+
+failed=0
+
+TMP="$(mktemp -d)"
+trap 'rm -rf "$TMP"' EXIT
+export LOG_FILE="$TMP/sync.log"
+
+# Script must exist and source cleanly (main guarded for sourcing)
+if [[ -f "$SCRIPT" ]] && (source "$SCRIPT"); then
+    echo "✅ script exists + sources cleanly"
+else
+    echo "❌ script missing or fails to source: $SCRIPT"
+    exit 1
+fi
+# shellcheck disable=SC1091
+source "$SCRIPT"
+
+# Fake command log: record invocations of rsync / reload / find
+FAKE_BIN="$TMP/bin"
+mkdir -p "$FAKE_BIN"
+export PATH="$FAKE_BIN:$PATH"
+
+make_rsync() {
+    # $1 = exit code, $2 = "changed" or "clean" (stdout of rsync)
+    cat > "$FAKE_BIN/rsync" <<EOF
+#!/bin/bash
+echo "$2"
+exit $1
+EOF
+    chmod +x "$FAKE_BIN/rsync"
+}
+
+make_reload() {
+    cat > "$FAKE_BIN/reload" <<'EOF'
+#!/bin/bash
+echo "reload-fired" >> "$RELOAD_LOG"
+EOF
+    chmod +x "$FAKE_BIN/reload"
+}
+
+RELOAD_LOG="$TMP/reload.log"
+touch "$RELOAD_LOG"
+export RELOAD_LOG
+
+# find: minimal fake so zone counting works
+mkdir -p "$FAKE_BIN/find-dir"
+cat > "$FAKE_BIN/find" <<'EOF'
+#!/bin/bash
+# report 3 fake zone files regardless of args
+for i in 1 2 3; do echo "/zones/zone$i"; done
+EOF
+chmod +x "$FAKE_BIN/find"
+
+test_reload_fires_on_change() {
+    make_rsync 0 "knel.net.zone"
+    make_reload
+    RELOAD_CMD=("$FAKE_BIN/reload")
+    export RELOAD_CMD
+    if sync_and_reload >/dev/null 2>&1 && [[ "$(cat "$RELOAD_LOG")" == "reload-fired" ]]; then
+        echo "✅ reload fires when rsync reports changed zones"
+    else
+        echo "❌ reload did not fire on changed sync"
+        ((++failed))
+    fi
+    : > "$RELOAD_LOG"
+}
+
+test_no_reload_on_clean_sync() {
+    make_rsync 0 ""
+    make_reload
+    RELOAD_CMD=("$FAKE_BIN/reload")
+    export RELOAD_CMD
+    if sync_and_reload >/dev/null 2>&1 && [[ ! -s "$RELOAD_LOG" ]]; then
+        echo "✅ no reload on clean sync"
+    else
+        echo "❌ reload fired (or errored) on clean sync"
+        ((++failed))
+    fi
+}
+
+test_rsync_failure_propagates() {
+    make_rsync 23 ""
+    RELOAD_CMD=("$FAKE_BIN/reload")
+    export RELOAD_CMD
+    if ! sync_and_reload >/dev/null 2>&1; then
+        echo "✅ rsync failure propagates non-zero rc"
+    else
+        echo "❌ rsync failure was swallowed"
+        ((++failed))
+    fi
+    if [[ ! -s "$RELOAD_LOG" ]]; then
+        echo "✅ no reload attempted after rsync failure"
+    else
+        echo "❌ reload fired despite rsync failure"
+        ((++failed))
+    fi
+}
+
+test_reload_failure_does_not_break_sync() {
+    make_rsync 0 "knel.net.zone"
+    RELOAD_CMD=("/bin/false")
+    export RELOAD_CMD
+    if sync_and_reload >/dev/null 2>&1; then
+        echo "✅ reload failure does not fail the sync run"
+    else
+        echo "❌ reload failure failed the whole sync"
+        ((++failed))
+    fi
+}
+
+test_reload_fires_on_change
+test_no_reload_on_clean_sync
+test_rsync_failure_propagates
+test_reload_failure_does_not_break_sync
+
+if ((failed > 0)); then
+    echo "❌ $failed sync-zones test(s) failed"
+    exit 1
+fi
+echo "✅ all sync-zones tests passed"