#gStore-weekly | gstore源码解析（六）：备份与恢复机制

上一章我们分析了gstore安全机制系列的源码，接下来将解析gstore的备份与恢复机制。

1.1 简介

gstore的备份和恢复可通过两种方式进行：第一种是通过本地命令执行；第二种就是调用API接口，基于API接口的方式我们可以很方便的实现定时备份的功能，在gstore可视化管理工具workbench中就集成了此功能；接下来会逐一进行解析。

1.2 备份

数据库的备份是针对具体的某一个库进行全量的数据备份，备份的原理就是复制当前数据库文件夹到备份目录下，并重命名为带时间戳后缀的文件夹。

1.2.1 本地备份

本地备份的执行命令和参数说明如下：

$ bin/gbackup -db dbname -p backupfilepath
# dbname:数据库名称
# backupfilepath（optional):备份数据库路径（相对路径，默认为gStore目录下的./backups)

通过以上命令可知，在执行本地命令时需要指定备份的数据库名称，默认备份路径为gStore目录下的backups，在进行本地数据备份时建议先关闭所有API Server；，部分关键代码在Main/gbackup.cpp中，gbackup::Main函数的实现如下所示：

int main(int argc, char * argv[]) {
  Util util;
  string db_name, backup_path;
    // 判断参数是否合法
  if (argc < 2) {
    cout << "Invalid arguments! Input \"bin/gbackup -h\" for help." << endl;
    return 0;
  } else if (argc == 2) {
        // 当本地命令输入为bin/gbackup -h[--help]，则输出帮助信息
    ......
  } else {
        // 读取参数-db或--database
    db_name = Util::getArgValue(argc, argv, "db", "database");
        // 读取参数-p或--path
    backup_path= Util::getArgValue(argc, argv, "p", "path");
        // 如果备份路径参数为空，则赋值为默认备份路径./backups
    if (backup_path.empty()) {
      backup_path = DEFALUT_BACKUP_PATH;
    }
    int len = db_name.length();
        // 校验数据库名称是否合法
    if (db_name.length() > 3 && db_name.substr(len - 3, 3) == ".db") {
      cout<<"your database name can not end with .db! Input \"bin/gbackup -h\" for help."<<endl;
      return -1;
    }
        // 判断为系统库时返回错误信息
    if (db_name == "system") {
      cout << "Your database's name can not be system." << endl;
      return -1;
    }
        // 校验备份路径的合法性
    if (backup_path == "." || backup_path == "./") {
      cout << "Backup Path Can not be root, Backup Failed!" << endl;
      return 0;
    }
    if (backup_path[backup_path.length() - 1] == '/') 
            backup_path = backup_path.substr(0, backup_path.length() - 1);
        // 拼装数据库源路径：db_name.db
    string db_path = db_name + ".db";
    if (backup_path == "") backup_path = DEFALUT_BACKUP_PATH;
    long tv_begin = Util::get_cur_time();
        // 执行复制
    int ret = copy(db_path, backup_path);
    if (ret == 1) {
      cout << "Database Name Error, Backup Failed!" << endl;
    } else {
            // 复制成功后，重命名为db_name.db_时间戳
      string timestamp = Util::get_timestamp();
      backup_path = backup_path + "/" + db_path;
      string _backup_path = backup_path + "_" + timestamp;
      string sys_cmd = "mv " + backup_path + " " + _backup_path;
      system(sys_cmd.c_str());
      long tv_end = Util::get_cur_time();
      cout << "DB:" << db_name << " Backup Successfully! Used " << (tv_end - tv_begin) << " ms"<<endl;  
    }
    return 0;
  }}

Main函数中调用的gbackup::copy函数如下：

/**
 * copy dir
 * @param src_path 
 * @param dest_path 
 * @return int 
 */int copy(string src_path, string dest_path){
  string sys_cmd;
    // 校验源路径是否存在，不存在则返回
  if(!boost::filesystem::exists(src_path)){
    //check the source path
    cout << "Source Path Error, Please Check It Again!" << endl;
    return 1;
  }
    // 判断目标路径是否存在，不存在则创建改路径
  if(!boost::filesystem::exists(dest_path)){
    //check the destnation path 
    sys_cmd = "mkdir -p ./" + dest_path;
    system(sys_cmd.c_str());
  }
    // 执行复制命令
  sys_cmd = "cp -r " + src_path + " ./" +  dest_path ;
  system(sys_cmd.c_str());
    return 0;// success 
    }

1.2.2 接口备份

接口备份的前提需要先启动API Server，我们以ghttp服务为例，接口请求参数如下所示：

{
    "operation":"backup", // 操作名称
    "username": "root", // 用户名
    "password": "123456", //密码
    "db_name": "demo", //备份数据库名称
    "backup_path": "" //备份目录，若不传或为空时，默认为gStore根目录下的backups目录
    }

ghttp服务在接口收到备份请求时，处理逻辑过程与本地备份类似，部分核心代码为ghttp::backup_thread_new函数：

/**
 * @description: backup a database 
 * @Author:liwenjie
 * @param {const} shared_ptr
 * @param {string} db_name:the operation database name
 * @param {string} backup_path: the backup path
 * @return {*}
 */void backup_thread_new(const shared_ptr<HttpServer::Response>& response,string db_name,string backup_path,string ip) {
  string error="";
  string operation="backup";
    // 校验参数db_name是否合法
  error=checkparamValue("db_name",db_name);
  if (error.empty() == false) {
    sendResponseMsg(1003, error, response,ip,operation);
    return;
  }
    // 校验数据库是否存在
  if (checkdbexist(db_name) == false){
    error = "the database [" + db_name + "] not built yet.";
    sendResponseMsg(1004, error, response,ip,operation);
    return;
  }
  ......
  //begin backup database
  string path=backup_path;
    // 如果备份路径为空，则赋值为默认备份路径./backups
  if(path == "") path = BACKUP_PATH;
    // 校验备份路径是否合法
  if(path == "." ){
    cout << "Backup Path Can not be root or empty, Backup Failed!" << endl;
    string error = "Failed to backup the database. Backup Path Can not be root or empty.";
    sendResponseMsg(1003,error,response,ip,operation);
    pthread_rwlock_unlock(&(it_already_build->second->db_lock));
    return;
  }
  if(path[path.length() - 1] == '/') path = path.substr(0, path.length() - 1);
  cout<<"backup path:"<<path<<endl;
  string db_path = db_name + ".db";
  pthread_rwlock_wrlock(&databases_map_lock);
    // 复制文件
  int ret = copy(db_path, path);
  pthread_rwlock_unlock(&databases_map_lock);
  string timestamp="";
  if(ret == 1){
    cout << "Database Folder Misssing, Backup Failed!" << endl;
    string error = "Failed to backup the database. Database Folder Misssing.";
    sendResponseMsg(1005,error,response,ip,operation);
    pthread_rwlock_unlock(&(it_already_build->second->db_lock));
    return;
  } else{
    string time = Util::get_date_time();
    timestamp = Util::get_timestamp();
    cout << "Time:" + time << endl;
    cout << "DB:" + db_name + " Backup done!" << endl;
  }
    // 复制成功后，重命名为db_name.db_时间戳
  path = path + "/" + db_path;
  string _path = path + "_" + timestamp;
  string sys_cmd = "mv " + path + " " + _path;
  cout<<"backup command:"<<sys_cmd<<endl;
  system(sys_cmd.c_str());
  cout << "database backup done." << endl;
  string success = "Database backup successfully.";
  //返回数据
  ......
  *response << "HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: " << resJson.length()  << "\r\n\r\n" << resJson;
  pthread_rwlock_unlock(&(it_already_build->second->db_lock));
    // 记录操作日志
  writeIpAccessLog(ip,operation,"backup "+db_name+" database successfully!",0);

1.3 还原

数据库的还原也是针对具体的某一个库进行数据还原，其原理是从备份路径下复制完整的文件到gStore根目录，在删除当前数据库文件后重命名备份文件为数据库文件。

1.3.1 本地还原

还原的执行命令和参数说明如下：

$ bin/grestore -db dbname -p path 
# dbname: 数据库名称（不需要带.db)
# path: 备份文件路径（相对gStore根目录而言的相对路径，且需要到文件夹）

通过以上命令可知，在执行本地命令时需要指定还原的数据库名称，备份文件路径；部分关键代码在Main/grestore.cpp中，grestore::Main函数的实现如下所示：

int main(int argc, char * argv[]) {
  Util util;
  string db_name, backup_path, db_path, path;
  if (argc < 2) {
    cout << "Invalid arguments! Input \"bin/grestore -h\" for help." << endl;
    return 0;
  } else if (argc == 2) {
        // 当本地命令输入为bin/grestore -h[--help]，则输出帮助信息
    ......
  } else {
        // 读取参数-db或--database，并校验参数的合法性
    db_name = Util::getArgValue(argc, argv, "db", "database");
    ......
        // 读取参数-p或--path，并校验参数的合法性
    backup_path = Util::getArgValue(argc, argv, "p", "path");
    ......
        // 加载system库并询问是否存在待还原的库，若不存在则向system库中增加还原库的信息
    Database system_db("system");
    system_db.load();
    string sparql = "ASK WHERE{<" + db_name + "> <database_status> \"already_built\".}";
    ......
        // 复制备份文件到gStore根目录
    int ret = copy(backup_path, DEFALUT_BUILD_PATH);
    ......
        // 删除当前库文件
    db_path = db_name + ".db";
    string sys_cmd = "rm -rf " + db_path;
    system(sys_cmd.c_str());
        // 重命名备份库文件为db_name.db
    path = Util::get_folder_name(backup_path, db_name);
    sys_cmd = "mv " + path + ' ' + db_path;
    system(sys_cmd.c_str());
    return 0;
  }}

1.3.2 接口还原

接口还原的前提也需要先启动API Server，我们以ghttp服务为例，接口请求参数如下所示：

{
    "operation":"restore", // 操作名称
    "username": "root", // 用户名
    "password": "123456", //密码
    "db_name": "demo", //还原数据库名称
    "backup_path": "" //备份文件的路径
    }
ghttp服务在接口收到还原请求时，处理逻辑与本地还原类似，部分核心代码为ghttp::restore_thread_new函数：

/**
 * @description:restore the database
 * @Author:liwenjie
 * @param {const} shared_ptr
 * @param {string} db_name the operation database name
 * @param {string} backup_path the backup path
 * @param {string} username the operation username
 * @param {string} ip the request ip address
 * @return {*}
 */void restore_thread_new(const shared_ptr<HttpServer::Response>& response,string db_name,string backup_path,string username,string ip) {
  string error="";
  string operation="restore";
    // 校验参数db_name的合法性
  error=checkparamValue("db_name",db_name);
  ......
    // 校验参数backup_path的合法性
  string path=backup_path;
  ......
  string database=db_name;
  cout << "restore "<< database << endl;


  pthread_rwlock_rdlock(&already_build_map_lock);
  std::map<std::string, struct DBInfo *>::iterator it_already_build = already_build.find(db_name);
    // 判断还原的数据库是否存在，若不存在则向system库中增加还原库的信息
  if(it_already_build == already_build.end())
  {
    ......
  }
  ......
  // 删除当前库文件
  string sys_cmd = "rm -rf " + db_name + ".db";
  system(sys_cmd.c_str());
  // 复制备份文件到gStore根目录
  pthread_rwlock_wrlock(&databases_map_lock);
  int ret  = copy(path, DB_PATH);
  pthread_rwlock_unlock(&databases_map_lock);
  // 判断复制是否成功
  if (ret == 1) {
    string error = "Failed to restore the database. Backup Path Error";
    sendResponseMsg(1005,error,response,ip,operation);
    cout << "Backup Path Error, Restore Failed!" << endl;
    pthread_rwlock_unlock(&(it_already_build->second->db_lock));
    return;
  } else {
    string time = Util::get_date_time();
    cout << "Time:" + time << endl;
    cout << "DB:" + db_name + " Restore done!" << endl;
  }
  // 重命名为db_name.db
  pthread_rwlock_unlock(&(it_already_build->second->db_lock));
  path = Util::get_folder_name(path, db_name);
  sys_cmd = "mv " + path + " " + db_name + ".db";
  system(sys_cmd.c_str());
  // 返回数据
  cout << "database restore done." << endl;
  string success = "Database ["+db_name+"] restore successfully.";
  sendResponseMsg(0,success,response,ip,operation);
  }

1.4 小结

本章节介绍了备份与恢复的机制和两种实现方式，建议在阅读的同时结合源码Main/gbackup.cpp、Main/grestore.cpp、Main/ghttp.cpp一起分析，会更容易理解。接下来我们拟将分析gstore的RDF解析机制。