Add DuckDB early initialization for crash recovery

- Add InitializeIfNeeded() to trigger DuckDB initialization at startup
  when duckdb_mode=ON or DuckDB data files exist
- Add DataFilesExist() to check for duckdb.db or WAL files
- Call InitializeIfNeeded() in mysqld_main() after CreateInstance()
  to ensure WAL replay happens at startup rather than on first request
- Fix merge conflicts in wiki documentation files

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: zongzhi.czz

sql/duckdb/duckdb_manager.cc

@@ -35,6 +35,8 @@ this program; if not, write to the Free Software Foundation, Inc.,
 #include "sql/duckdb/duckdb_timezone.h"
 #include "sql/mysqld.h"
 
+#include <sys/stat.h>
+
 namespace myduck {
 
 DuckdbManager *DuckdbManager::m_instance = nullptr;
@@ -146,4 +148,53 @@ std::shared_ptr<duckdb::Connection> DuckdbManager::CreateConnection() {
   return connection;
 }
 
+bool DuckdbManager::DataFilesExist() {
+  char path[FN_REFLEN];
+  struct stat st;
+
+  // Check for main database file: duckdb.db
+  fn_format(path, DUCKDB_FILE_NAME, mysql_real_data_home, "", MYF(0));
+  if (stat(path, &st) == 0 && S_ISREG(st.st_mode)) {
+    return true;
+  }
+
+  // Check for WAL file: duckdb.db.wal
+  fn_format(path, DUCKDB_FILE_NAME, mysql_real_data_home, ".wal", MYF(0));
+  if (stat(path, &st) == 0 && S_ISREG(st.st_mode)) {
+    return true;
+  }
+
+  return false;
+}
+
+bool DuckdbManager::InitializeIfNeeded() {
+  if (m_instance == nullptr) {
+    return true;
+  }
+
+  // Check if we need to initialize DuckDB at startup:
+  // 1. DuckDB mode is ON, or
+  // 2. DuckDB data files exist (need crash recovery)
+  bool need_init = global_mode_on() || DataFilesExist();
+
+  if (!need_init) {
+    return false;
+  }
+
+  if (DataFilesExist()) {
+    LogErr(INFORMATION_LEVEL, ER_DUCKDB,
+           "DuckDB data files detected, initializing for crash recovery.");
+  }
+
+  // Trigger initialization (which includes crash recovery via WAL replay)
+  bool ret = m_instance->Initialize();
+  if (ret) {
+    LogErr(ERROR_LEVEL, ER_DUCKDB,
+           "DuckdbManager::InitializeIfNeeded failed.");
+    return true;
+  }
+
+  return false;
+}
+
 }  // namespace myduck

sql/duckdb/duckdb_manager.h

@@ -50,7 +50,18 @@ class DuckdbManager {
 
   static std::shared_ptr<duckdb::Connection> CreateConnection();
 
-private:
+  /**
+   * Initialize DuckDB if needed at startup.
+   * This should be called after CreateInstance() to trigger crash recovery
+   * when:
+   * 1. DuckDB mode is ON (global_mode == DUCKDB_ON), or
+   * 2. DuckDB data files exist in the data directory
+   *
+   * @return true on error, false on success
+   */
+  static bool InitializeIfNeeded();
+
+ private:
   static DuckdbManager *m_instance;
 
   DuckdbManager();
@@ -59,6 +70,14 @@ class DuckdbManager {
 
   bool Initialize();
 
+  /**
+   * Check if DuckDB data files exist in the data directory.
+   * Looks for duckdb.db file or WAL files.
+   *
+   * @return true if DuckDB data files exist
+   */
+  static bool DataFilesExist();
+
   duckdb::DuckDB *m_database = nullptr;
 
   std::mutex m_mutex;

sql/mysqld.cc

@@ -7918,6 +7918,11 @@ int mysqld_main(int argc, char **argv)
 
   if (myduck::DuckdbManager::CreateInstance()) unireg_abort(MYSQLD_ABORT_EXIT);
 
+  // Initialize DuckDB early if needed for crash recovery
+  // This ensures WAL replay happens at startup rather than on first request
+  if (myduck::DuckdbManager::InitializeIfNeeded())
+    unireg_abort(MYSQLD_ABORT_EXIT);
+
   if (init_server_components()) unireg_abort(MYSQLD_ABORT_EXIT);
 
   if (!server_id_supplied)

wiki/duckdb/how-to-setup-duckdb-node-en.md

@@ -73,7 +73,6 @@ mkdir -p \
 cat > "$DIR/alisql.cnf" <<EOF2
 [mysqld]
 # --- Paths ---
-basedir =
 datadir = ${DIR}/data/dbs
 socket  = ${DIR}/run/mysql.sock
 pid-file = ${DIR}/run/mysqld.pid

wiki/duckdb/how-to-setup-duckdb-node-zh.md

@@ -81,7 +81,6 @@ mkdir -p \
 cat > "$DIR/alisql.cnf" <<EOF2
 [mysqld]
 # --- Paths ---
-basedir =
 datadir = ${DIR}/data/dbs
 socket  = ${DIR}/run/mysql.sock
 pid-file = ${DIR}/run/mysqld.pid