Update entrypoint.sh

entrypoint.sh

@@ -1,9 +1,8 @@
 #!/bin/sh
-umask 022
 # OpenClaw HF Spaces - Production Entrypoint
-# Storage: HF Storage Bucket at /data (100 GB persistent)
+umask 022
 
-# ── Fix: Force IPv4 for Telegram (IPv6 route broken on HF nodes) ──
+# ── Force IPv4 for Telegram (IPv6 route broken on HF nodes) ──
 sysctl -w net.ipv6.conf.all.disable_ipv6=1 2>/dev/null || true
 sysctl -w net.ipv6.conf.default.disable_ipv6=1 2>/dev/null || true
 export NODE_OPTIONS="--dns-result-order=ipv4first ${NODE_OPTIONS:-}"
@@ -83,21 +82,10 @@ if [ -f /app/security-check.sh ]; then
 fi
 
 # ----------------------------------------------------------------
-# Migrate legacy state from older OpenClaw versions (required by new releases)
+# State DB helpers
 # ----------------------------------------------------------------
-# Move retired Telegram JSON state aside (it blocks doctor migration;
-# only holds update offsets / sent-message cache). Kept as backup.
-LEGACY_BAK="$OPENCLAW_HOME/legacy-telegram-backup"
-mkdir -p "$LEGACY_BAK"
-for f in "$OPENCLAW_HOME"/.openclaw/telegram/update-offset-*.json \
-         "$OPENCLAW_HOME"/.openclaw/agents/*/sessions/sessions.json.telegram-sent-messages.json; do
-  if [ -f "$f" ]; then
-    mv "$f" "$LEGACY_BAK/$(echo "$f" | tr '/' '_')" \
-      && log "Moved legacy telegram state: $f"
-  fi
-done
+STATE_DB="$OPENCLAW_HOME/.openclaw/state/openclaw.sqlite"
 
-# ---- Check shared state DB; restore from doctor's own backup if corrupt ----
 check_db() {
   DBPATH="$1" node - <<'JSEOF'
 const { DatabaseSync } = require('node:sqlite');
@@ -113,7 +101,35 @@ try {
 JSEOF
 }
 
-STATE_DB="$OPENCLAW_HOME/.openclaw/state/openclaw.sqlite"
+dump_owner() {
+  DBPATH="$1" node - <<'JSEOF'
+const { DatabaseSync } = require('node:sqlite');
+try {
+  const db = new DatabaseSync(process.env.DBPATH, { readOnly: true });
+  const tables = db.prepare("SELECT name FROM sqlite_master WHERE type='table' AND (name LIKE '%owner%' OR name LIKE '%lock%' OR name LIKE '%lease%' OR name LIKE '%heartbeat%' OR name LIKE '%machine_state%')").all();
+  for (const t of tables) {
+    console.log('[owner-dump] table ' + t.name);
+    for (const r of db.prepare('SELECT * FROM "' + t.name + '" LIMIT 5').all()) {
+      console.log('[owner-dump] ' + JSON.stringify(r).slice(0, 400));
+    }
+  }
+  db.close();
+} catch (e) { console.log('[owner-dump] ' + e.message); }
+JSEOF
+}
+
+# Move retired Telegram JSON state aside (blocks doctor migration; kept as backup)
+LEGACY_BAK="$OPENCLAW_HOME/legacy-telegram-backup"
+mkdir -p "$LEGACY_BAK"
+for f in "$OPENCLAW_HOME"/.openclaw/telegram/update-offset-*.json \
+         "$OPENCLAW_HOME"/.openclaw/agents/*/sessions/sessions.json.telegram-sent-messages.json; do
+  if [ -f "$f" ]; then
+    mv "$f" "$LEGACY_BAK/$(echo "$f" | tr '/' '_')" \
+      && log "Moved legacy telegram state: $f"
+  fi
+done
+
+# If the state DB is corrupt, restore from doctor's own pre-migration backup
 if [ -f "$STATE_DB" ] && ! check_db "$STATE_DB" 2>/dev/null; then
   warn "State DB failed integrity check - looking for an intact backup..."
   GOOD=""
@@ -129,9 +145,40 @@ if [ -f "$STATE_DB" ] && ! check_db "$STATE_DB" 2>/dev/null; then
   fi
 fi
 
+# ----------------------------------------------------------------
+# Doctor (state migration) with lock diagnostics and one-time fallback
+# ----------------------------------------------------------------
+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)..."
-timeout 300 openclaw doctor --fix --non-interactive </dev/null \
-  || warn "openclaw doctor reported problems (see output above)"
+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 "$OPENCLAW_HOME/.openclaw" -maxdepth 3 -iname '*lock*' 2>/dev/null | head -n 20
+  warn "retrying doctor in 30s..."
+  sleep 30
+  run_doctor
+fi
+if doctor_locked; then
+  if ls "$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
+fi
 log "Doctor step done."
 
 # Plugins must not be world-writable or OpenClaw blocks them