Update entrypoint.sh

entrypoint.sh

@@ -82,7 +82,7 @@ if [ -f /app/security-check.sh ]; then
 fi
 
 # ----------------------------------------------------------------
-# State DB helpers
+# Helpers
 # ----------------------------------------------------------------
 STATE_DB="$OPENCLAW_HOME/.openclaw/state/openclaw.sqlite"
 
@@ -106,7 +106,7 @@ dump_owner() {
 const { DatabaseSync } = require('node:sqlite');
 try {
   const db = new DatabaseSync(process.env.DBPATH, { readOnly: true });
-  const tables = db.prepare("SELECT name FROM sqlite_master WHERE type='table' AND (name LIKE '%owner%' OR name LIKE '%lock%' OR name LIKE '%lease%' OR name LIKE '%heartbeat%' OR name LIKE '%machine_state%')").all();
+  const tables = db.prepare("SELECT name FROM sqlite_master WHERE type='table' AND (name LIKE '%owner%' OR name LIKE '%lock%' OR name LIKE '%lease%' OR name LIKE '%heartbeat%')").all();
   for (const t of tables) {
     console.log('[owner-dump] table ' + t.name);
     for (const r of db.prepare('SELECT * FROM "' + t.name + '" LIMIT 5').all()) {
@@ -118,6 +118,26 @@ try {
 JSEOF
 }
 
+# Stale lock files left on the persistent volume by a crashed gateway
+clear_stale_locks() {
+  for d in "$OPENCLAW_HOME"/.openclaw/tmp/openclaw-*; do
+    [ -d "$d" ] || continue
+    for l in "$d"/*.lock; do
+      [ -e "$l" ] || continue
+      INFO=$(head -c 160 "$l" 2>/dev/null | tr '\n' ' ')
+      rm -rf "$l" && log "Removed stale lock: $l [$INFO]"
+    done
+  done
+}
+
+run_doctor() {
+  timeout 300 openclaw doctor --fix --non-interactive </dev/null >/tmp/doctor.out 2>&1
+  cat /tmp/doctor.out
+}
+doctor_locked() {
+  grep -q "could not enter maintenance" /tmp/doctor.out
+}
+
 # Move retired Telegram JSON state aside (blocks doctor migration; kept as backup)
 LEGACY_BAK="$OPENCLAW_HOME/legacy-telegram-backup"
 mkdir -p "$LEGACY_BAK"
@@ -129,6 +149,19 @@ for f in "$OPENCLAW_HOME"/.openclaw/telegram/update-offset-*.json \
   fi
 done
 
+# ----------------------------------------------------------------
+# Restore a state DB that an earlier boot moved aside
+# ----------------------------------------------------------------
+LOCKED=$(ls -t "$STATE_DB".locked-* 2>/dev/null | head -n1)
+if [ -n "$LOCKED" ]; then
+  warn "Found a state DB moved aside by an earlier boot - restoring: $LOCKED"
+  if [ -e "$STATE_DB" ]; then
+    mv "$STATE_DB" "$STATE_DB.fresh-$(date +%s)"
+  fi
+  rm -f "$STATE_DB-wal" "$STATE_DB-shm"
+  mv "$LOCKED" "$STATE_DB" && log "Restored state DB from $LOCKED"
+fi
+
 # If the state DB is corrupt, restore from doctor's own pre-migration backup
 if [ -f "$STATE_DB" ] && ! check_db "$STATE_DB" 2>/dev/null; then
   warn "State DB failed integrity check - looking for an intact backup..."
@@ -146,38 +179,23 @@ if [ -f "$STATE_DB" ] && ! check_db "$STATE_DB" 2>/dev/null; then
 fi
 
 # ----------------------------------------------------------------
-# Doctor (state migration) with lock diagnostics and one-time fallback
+# Doctor (state migration)
 # ----------------------------------------------------------------
-run_doctor() {
-  timeout 300 openclaw doctor --fix --non-interactive </dev/null >/tmp/doctor.out 2>&1
-  cat /tmp/doctor.out
-}
-doctor_locked() {
-  grep -q "could not enter maintenance" /tmp/doctor.out
-}
-
 log "Running openclaw doctor --fix (state migration)..."
+clear_stale_locks
 run_doctor
 if doctor_locked; then
   warn "doctor blocked by state ownership lock - diagnostics follow"
   log "now (UTC): $(date -u +%Y-%m-%dT%H:%M:%S)"
   dump_owner "$STATE_DB" 2>/dev/null
-  find "$OPENCLAW_HOME/.openclaw" -maxdepth 3 -iname '*lock*' 2>/dev/null | head -n 20
+  find "$OPENCLAW_HOME/.openclaw/tmp" -maxdepth 3 2>/dev/null | head -n 30
   warn "retrying doctor in 30s..."
   sleep 30
+  clear_stale_locks
   run_doctor
 fi
 if doctor_locked; then
-  if ls "$STATE_DB".locked-* >/dev/null 2>&1; then
-    warn "still locked, and the one-time fallback was already used - not touching the DB again"
-  else
-    warn "still locked - moving state DB aside (one time) and retrying on a fresh DB"
-    TSX=$(date +%s)
-    for f in "$STATE_DB" "$STATE_DB-wal" "$STATE_DB-shm"; do
-      [ -e "$f" ] && mv "$f" "$f.locked-$TSX"
-    done
-    run_doctor
-  fi
+  warn "doctor still blocked by the state lock - not touching any data"
 fi
 log "Doctor step done."
 
@@ -232,6 +250,7 @@ JSEOF
 log "Starting gateway loop..."
 
 while true; do
+  clear_stale_locks
   log "Gateway starting..."
 
   openclaw gateway \