From 2420915d3079dc44da927e9db17223c23fab7a66 Mon Sep 17 00:00:00 2001
From: Bernd Worsch <bernd.worsch@gmail.com>
Date: Thu, 26 Mar 2026 21:56:19 +0000
Subject: [PATCH] fix(backup): SQLite hot backup instead of etcd snapshot

k3s runs in SQLite mode (no --cluster-init). Replace etcd-snapshot
with sqlite3 .backup for a WAL-aware hot copy of state.db.
Update restore guide to match. Cron installed under root crontab.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 tools/cmd/railiance-backup-s2  | 30 +++++++++++++-----------------
 tools/cmd/railiance-restore-s2 | 22 ++++++++++++----------
 2 files changed, 25 insertions(+), 27 deletions(-)

diff --git a/tools/cmd/railiance-backup-s2 b/tools/cmd/railiance-backup-s2
index f05a80c..79fe69e 100755
--- a/tools/cmd/railiance-backup-s2
+++ b/tools/cmd/railiance-backup-s2
@@ -13,7 +13,7 @@ source "${ROOT}/lib/railiance-print.sh"
 AGE_PUBLIC_KEY="age1aq8twfd78wvpra0had8cezcnj96tj4q0068edrz5jez8d6xwmflqdepsh4"
 BACKUP_DIR="/opt/backup/railiance/cluster"
 KUBECONFIG_PATH="/etc/rancher/k3s/k3s.yaml"
-ETCD_SNAP_DIR="/var/lib/rancher/k3s/server/db/snapshots"
+K3S_STATE_DB="/var/lib/rancher/k3s/server/db/state.db"
 KEEP=7
 TS="$(date -u +%Y%m%dT%H%M%SZ)"
 
@@ -27,22 +27,18 @@ fi
 
 mkdir -p "${BACKUP_DIR}"
 
-# ── 1. k3s etcd snapshot ───────────────────────────────────────────────────────
-if k3s etcd-snapshot ls &>/dev/null; then
-  ok "etcd" "taking snapshot…"
-  SNAP_NAME="railiance-${TS}"
-  k3s etcd-snapshot save --name "${SNAP_NAME}" &>/dev/null
-  SNAP_FILE="${ETCD_SNAP_DIR}/${SNAP_NAME}"
-  if [[ ! -f "${SNAP_FILE}" ]]; then
-    # k3s may append a suffix — find the most recent matching file
-    SNAP_FILE="$(find "${ETCD_SNAP_DIR}" -name "${SNAP_NAME}*" | sort -r | head -1)"
-  fi
-  age -r "${AGE_PUBLIC_KEY}" -o "${BACKUP_DIR}/etcd-${TS}.snap.age" "${SNAP_FILE}"
-  ok "etcd" "encrypted → etcd-${TS}.snap.age"
-  # Prune old snapshots from k3s store (keep last KEEP)
-  k3s etcd-snapshot prune --snapshot-retention "${KEEP}" &>/dev/null || true
+# ── 1. k3s state (SQLite hot backup) ──────────────────────────────────────────
+# This cluster runs k3s in SQLite mode (no --cluster-init).
+# sqlite3 .backup performs a WAL-aware hot copy — no k3s stop required.
+if [[ -f "${K3S_STATE_DB}" ]]; then
+  ok "state-db" "taking hot backup…"
+  TMP_STATE="$(mktemp /tmp/k3s-state-XXXXXX.db)"
+  sqlite3 "${K3S_STATE_DB}" ".backup ${TMP_STATE}"
+  age -r "${AGE_PUBLIC_KEY}" -o "${BACKUP_DIR}/k3s-state-${TS}.db.age" "${TMP_STATE}"
+  rm -f "${TMP_STATE}"
+  ok "state-db" "encrypted → k3s-state-${TS}.db.age"
 else
-  warn "etcd" "k3s etcd not available (SQLite mode?) — skipping snapshot"
+  warn "state-db" "${K3S_STATE_DB} not found — skipping"
 fi
 
 # ── 2. Helm release values ─────────────────────────────────────────────────────
@@ -73,7 +69,7 @@ else
 fi
 
 # ── 4. Prune local cache ───────────────────────────────────────────────────────
-for pattern in "etcd-*.snap.age" "helm-values-*.tar.gz.age" "kubeconfig-*.yaml.age"; do
+for pattern in "k3s-state-*.db.age" "helm-values-*.tar.gz.age" "kubeconfig-*.yaml.age"; do
   find "${BACKUP_DIR}" -name "${pattern}" | sort -r | tail -n +$((KEEP + 1)) | xargs -r rm -f
 done
 ok "prune" "kept last ${KEEP} of each type"
diff --git a/tools/cmd/railiance-restore-s2 b/tools/cmd/railiance-restore-s2
index d613054..0aa3eb4 100755
--- a/tools/cmd/railiance-restore-s2
+++ b/tools/cmd/railiance-restore-s2
@@ -38,9 +38,9 @@ list_type() {
   echo ""
 }
 
-list_type "etcd snapshots" "etcd-*.snap.age"
-list_type "Helm values"    "helm-values-*.tar.gz.age"
-list_type "kubeconfig"     "kubeconfig-*.yaml.age"
+list_type "k3s state (SQLite)" "k3s-state-*.db.age"
+list_type "Helm values"        "helm-values-*.tar.gz.age"
+list_type "kubeconfig"         "kubeconfig-*.yaml.age"
 
 echo "============================================"
 echo ""
@@ -50,14 +50,16 @@ echo ""
 echo "Restore kubeconfig:"
 echo "  age -d -i ${AGE_KEY} ${BACKUP_DIR}/kubeconfig-<ts>.yaml.age > ~/.kube/config-hosteurope"
 echo ""
-echo "Restore etcd snapshot (WARNING: destroys current cluster state):"
-echo "  # 1. Decrypt the snapshot"
-echo "  age -d -i ${AGE_KEY} ${BACKUP_DIR}/etcd-<ts>.snap.age > /tmp/etcd-restore.snap"
-echo "  # 2. Copy to k3s snapshot directory"
-echo "  sudo cp /tmp/etcd-restore.snap /var/lib/rancher/k3s/server/db/snapshots/"
-echo "  # 3. Stop k3s and restore"
+echo "Restore k3s state (SQLite) — WARNING: destroys current cluster state:"
+echo "  # 1. Decrypt the state db"
+echo "  age -d -i ${AGE_KEY} ${BACKUP_DIR}/k3s-state-<ts>.db.age > /tmp/k3s-restore.db"
+echo "  # 2. Stop k3s"
 echo "  sudo systemctl stop k3s"
-echo "  sudo k3s server --cluster-reset --cluster-reset-restore-path=/var/lib/rancher/k3s/server/db/snapshots/etcd-restore.snap"
+echo "  # 3. Replace the state db"
+echo "  sudo cp /var/lib/rancher/k3s/server/db/state.db /var/lib/rancher/k3s/server/db/state.db.bak"
+echo "  sudo cp /tmp/k3s-restore.db /var/lib/rancher/k3s/server/db/state.db"
+echo "  sudo rm -f /var/lib/rancher/k3s/server/db/state.db-shm /var/lib/rancher/k3s/server/db/state.db-wal"
+echo "  # 4. Start k3s"
 echo "  sudo systemctl start k3s"
 echo ""
 echo "Restore Helm values (for re-running helm upgrade after cluster restore):"