TeamTalk梳理概括

即时通讯重点概括

展开聊聊单聊消息流转流程

• 消息如何封装

1. 怎么解决半包、粘包问题？
2. 消息流转流程介绍下？

• 消息序号（msg_id ）在哪里生成以及生成方式
• 怎么保证数据的不丢失以及重复包？

1. 接收端收到数据后（收到消息区别于阅读消息）如何应答？
2. 消息发送后服务器怎么应答?
3. 消息发送时的seq有什么作用（业务层的ack机制）

展开聊聊群聊消息流转流程

• 如何推送群聊
• 群消息计数器（msg_id ）
• 群会话如何更新(每有一个人发送消息,则其他人都需要更新会话消息)

群成员管理

• 如何创建群
• 如何删除群
• 怎么使用redis管理群成员
  消息未读计数是怎么实现的？
• 服务器怎么保留消息未读计数（redis 单聊和群聊机制不同）
• 客户端的未读消息计数从何而来
• 客户端未读消息计数清0时向服务器发送了什么？服务器又是怎么清除未读消息计数（单聊和群聊机制不同）

数据库

• 数据库表设计（表达笼统）
• 密码存储方式
• 未读消息如何体现
• 聊天消息分表问题（单聊和群聊消息表）
• 最近会话表

MySQL连接池设计

• 为什么使用连接池
• 连接池设置多大合适?

redis连接池设计

• 为什么使用连接池
• 连接池设置多大合适?

文件传输原理

• 在线传输和离线传输有什么区别

实时性

• Http（登录、图片服务）
• Socket
• websocket
  展开聊聊登录流程

并发能力

• 如何做到百万并发
• 如何做到千万并发

db_proxy_server reactor响应处理流程

1. 数据入口 reactor CProxyConn:: HandlePduBuf
2. 怎么初始化epoll+线程池
3. 任务封装
4. 把任务放入线程池
5. 执行任务
6. 把要回应的数据放入回复列表CProxyConn::SendResponsePdulist
7. epoll所在线程读取回复列表的数据发给请求端

单聊消息

消息如何封装？如何保证对端完整解析一帧消息？协议格式？

1. 答：消息封装采用包头（Header）+包体（Body）的格式。包头自定义格式如下代码所示，包体采用protobuf序列化。

typedef struct {uint32_t    length;        // the whole pdu lengthuint16_t    version;       // pdu version numberuint16_t    flag;          // not useduint16_t    service_id;    //uint16_t    command_id;    //uint16_t    seq_num;       // 包序号uint16_t    reversed;      // 保留
} PduHeader_t;

3. 答：

• 采用tcp保证数据传输可靠性
• 通过包头的 length 字段标记一帧消息的长度
• 通过service id 和 command id区分不同的命令（比如登录、退出等）
• 解决数据TCP粘包（包头长度字段）、半包（放入网络库的缓冲区）问题

void CImConn::OnRead()
{for (;;){uint32_t free_buf_len = m_in_buf.GetAllocSize() - m_in_buf.GetWriteOffset();if (free_buf_len < READ_BUF_SIZE)m_in_buf.Extend(READ_BUF_SIZE);int ret = netlib_recv(m_handle, m_in_buf.GetBuffer() + m_in_buf.GetWriteOffset(), READ_BUF_SIZE);if (ret <= 0)break;m_recv_bytes += ret;m_in_buf.IncWriteOffset(ret);m_last_recv_tick = get_tick_count();}CImPdu* pPdu = NULL;try{while ( ( pPdu = CImPdu::ReadPdu(m_in_buf.GetBuffer(), m_in_buf.GetWriteOffset()) ) ){uint32_t pdu_len = pPdu->GetLength();HandlePdu(pPdu);m_in_buf.Read(NULL, pdu_len);delete pPdu;pPdu = NULL;
//                        ++g_recv_pkt_cnt;}} catch (CPduException& ex) {log("!!!catch exception, sid=%u, cid=%u, err_code=%u, err_msg=%s, close the connection ",ex.GetServiceId(), ex.GetCommandId(), ex.GetErrorCode(), ex.GetErrorMsg());if (pPdu) {delete pPdu;pPdu = NULL;}OnClose();}
}

message IMMsgData{//cmd id:                0x0301required uint32 from_user_id = 1;    //消息发送方required uint32 to_session_id = 2;    //消息接受方required uint32 msg_id = 3;required uint32 create_time = 4; required IM.BaseDefine.MsgType msg_type = 5;required bytes msg_data = 6;optional bytes attach_data = 20;
}message IMMsgDataAck{//cmd id:                0x0302required uint32 user_id = 1;    //发送此信令的用户idrequired uint32 session_id = 2;                                required uint32 msg_id = 3;required IM.BaseDefine.SessionType session_type = 4;
}

消息序号（msg_id ）为什么使用redis生成？

5. 消息ID（msg_id ）的作用是防止消息乱序。
6. 消息ID（msg_id ）为什么这么设计？
   答：msg_id 存储在 unread 连接池所在的redis数据库。单聊 msg_id 的 key涉及到 nRelateId。nRelateId 从关系表（IMRelationShip ：两个用户id的映射关系）中获取。

/***  获取会话关系ID*  对于群组，必须把nUserBId设置为群ID**  @param nUserAId  <#nUserAId description#>*  @param nUserBId  <#nUserBId description#>*  @param bAdd      <#bAdd description#>*  @param nStatus 0 获取未被删除会话，1获取所有。*/
uint32_t CRelationModel::getRelationId(uint32_t nUserAId, uint32_t nUserBId, bool bAdd)
{uint32_t nRelationId = INVALID_VALUE;if (nUserAId == 0 || nUserBId == 0) {log("invalied user id:%u->%u", nUserAId, nUserBId);return nRelationId;}CDBManager* pDBManager = CDBManager::getInstance();CDBConn* pDBConn = pDBManager->GetDBConn("teamtalk_slave");if (pDBConn){uint32_t nBigId = nUserAId > nUserBId ? nUserAId : nUserBId;uint32_t nSmallId = nUserAId > nUserBId ? nUserBId : nUserAId;string strSql = "select id from IMRelationShip where smallId=" + int2string(nSmallId) + " and bigId="+ int2string(nBigId) + " and status = 0";CResultSet* pResultSet = pDBConn->ExecuteQuery(strSql.c_str());if (pResultSet){while (pResultSet->Next()){nRelationId = pResultSet->GetInt("id");}delete pResultSet;}else{log("there is no result for sql:%s", strSql.c_str());}pDBManager->RelDBConn(pDBConn);if (nRelationId == INVALID_VALUE && bAdd){nRelationId = addRelation(nSmallId, nBigId);}}else{log("no db connection for teamtalk_slave");}return nRelationId;
}

7. 群聊和单聊msg_id 的区别：key设置不同。

uint32_t CMessageModel::getMsgId(uint32_t nRelateId)
{uint32_t nMsgId = 0;CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("unread");if(pCacheConn){string strKey = "msg_id_" + int2string(nRelateId);nMsgId = pCacheConn->incrBy(strKey, 1);pCacheManager->RelCacheConn(pCacheConn);}return nMsgId;
}/***  获取一个群组的msgId，自增，通过redis控制*  @param nGroupId 群Id*  @return 返回msgId*/
uint32_t CGroupMessageModel::getMsgId(uint32_t nGroupId)
{uint32_t nMsgId = 0;CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("unread");if(pCacheConn){// TODOstring strKey = "group_msg_id_" + int2string(nGroupId);nMsgId = pCacheConn->incrBy(strKey, 1);pCacheManager->RelCacheConn(pCacheConn);}else{log("no cache connection for unread");}return nMsgId;
}

展开聊聊单聊消息流转流程

答：两个用户A给B发消息，用户A把聊天消息封装好以后发送给MsgServer；同时把消息进行持久化，将聊天消息发给这个 DBProxy（数据库代理服务），存储消息成功后，DBProxyServer组包应答MsgServer，MsgServer收到回复后组包应答Client A。如果 A 和 B 两个用户不在同一个 MsgServer 上，那么会通过这个 RouteServer 去中转Pdu包数据（广播给所有的MsgServer，MsgServer再广播给Client B），B收到消息后应答MsgServer，至此，流程结束。然后如果是一些热点数据，我们同时也会写Redis。
群聊消息

展开聊聊群聊消息流转流程

群聊消息流转

void CGroupChat::HandleGroupMessage(CImPdu* pPdu)
{IM::Message::IMMsgData msg;CHECK_PB_PARSE_MSG(msg.ParseFromArray(pPdu->GetBodyData(), pPdu->GetBodyLength()));uint32_t from_user_id = msg.from_user_id();uint32_t to_group_id = msg.to_session_id();string msg_data = msg.msg_data();uint32_t msg_id = msg.msg_id();if (msg_id == 0) {log("HandleGroupMsg, write db failed, %u->%u. ", from_user_id, to_group_id);return;}uint8_t msg_type = msg.msg_type();CDbAttachData attach_data((uchar_t*)msg.attach_data().c_str(), msg.attach_data().length());log("HandleGroupMsg, %u->%u, msg id=%u. ", from_user_id, to_group_id, msg_id);CMsgConn* pFromConn = CImUserManager::GetInstance()->GetMsgConnByHandle(from_user_id,attach_data.GetHandle());if (pFromConn){//接收反馈IM::Message::IMMsgDataAck msg2;msg2.set_user_id(from_user_id);msg2.set_session_id(to_group_id);msg2.set_msg_id(msg_id);msg2.set_session_type(::IM::BaseDefine::SESSION_TYPE_GROUP);CImPdu pdu;pdu.SetPBMsg(&msg2);pdu.SetServiceId(SID_MSG);pdu.SetCommandId(CID_MSG_DATA_ACK);pdu.SetSeqNum(pPdu->GetSeqNum());pFromConn->SendPdu(&pdu);}CRouteServConn* pRouteConn = get_route_serv_conn();if (pRouteConn){pRouteConn->SendPdu(pPdu);}// 服务器没有群的信息，向DB服务器请求群信息，并带上消息作为附件，返回时在发送该消息给其他群成员//IM::BaseDefine::GroupVersionInfo group_version_info;CPduAttachData pduAttachData(ATTACH_TYPE_HANDLE_AND_PDU, attach_data.GetHandle(), pPdu->GetBodyLength(), pPdu->GetBodyData());IM::Group::IMGroupInfoListReq msg3;msg3.set_user_id(from_user_id);IM::BaseDefine::GroupVersionInfo* group_version_info = msg3.add_group_version_list();group_version_info->set_group_id(to_group_id);group_version_info->set_version(0);msg3.set_attach_data(pduAttachData.GetBuffer(), pduAttachData.GetLength());CImPdu pdu;pdu.SetPBMsg(&msg3);pdu.SetServiceId(SID_GROUP);pdu.SetCommandId(CID_GROUP_INFO_REQUEST);CDBServConn* pDbConn = get_db_serv_conn();if(pDbConn){pDbConn->SendPdu(&pdu);}
}
void CGroupChat::HandleGroupInfoResponse(CImPdu* pPdu)
{IM::Group::IMGroupInfoListRsp msg;CHECK_PB_PARSE_MSG(msg.ParseFromArray(pPdu->GetBodyData(), pPdu->GetBodyLength()));uint32_t user_id = msg.user_id();uint32_t group_cnt = msg.group_info_list_size();CPduAttachData pduAttachData((uchar_t*)msg.attach_data().c_str(), msg.attach_data().length());log("HandleGroupInfoResponse, user_id=%u, group_cnt=%u. ", user_id, group_cnt);//此处是查询成员时使用，主要用于群消息从数据库获得msg_id后进行发送,一般此时group_cnt = 1if (pduAttachData.GetPduLength() > 0 && group_cnt > 0){IM::BaseDefine::GroupInfo group_info = msg.group_info_list(0);uint32_t group_id = group_info.group_id();log("GroupInfoRequest is send by server, group_id=%u ", group_id);std::set<uint32_t> group_member_set;for (uint32_t i = 0; i < group_info.group_member_list_size(); i++){uint32_t member_user_id = group_info.group_member_list(i);group_member_set.insert(member_user_id);}if (group_member_set.find(user_id) == group_member_set.end()){log("user_id=%u is not in group, group_id=%u. ", user_id, group_id);return;}IM::Message::IMMsgData msg2;CHECK_PB_PARSE_MSG(msg2.ParseFromArray(pduAttachData.GetPdu(), pduAttachData.GetPduLength()));CImPdu pdu;pdu.SetPBMsg(&msg2);pdu.SetServiceId(SID_MSG);pdu.SetCommandId(CID_MSG_DATA);//Push相关IM::Server::IMGroupGetShieldReq msg3;msg3.set_group_id(group_id);msg3.set_attach_data(pdu.GetBodyData(), pdu.GetBodyLength());for (uint32_t i = 0; i < group_info.group_member_list_size(); i++){uint32_t member_user_id = group_info.group_member_list(i);msg3.add_user_id(member_user_id);CImUser* pToImUser = CImUserManager::GetInstance()->GetImUserById(member_user_id);if (pToImUser){CMsgConn* pFromConn = NULL;if( member_user_id == user_id ){uint32_t reqHandle = pduAttachData.GetHandle();if(reqHandle != 0)pFromConn = CImUserManager::GetInstance()->GetMsgConnByHandle(user_id, reqHandle);}pToImUser->BroadcastData(pdu.GetBuffer(), pdu.GetLength(), pFromConn);}}CImPdu pdu2;pdu2.SetPBMsg(&msg3);pdu2.SetServiceId(SID_OTHER);pdu2.SetCommandId(CID_OTHER_GET_SHIELD_REQ);CDBServConn* pDbConn = get_db_serv_conn();if (pDbConn){pDbConn->SendPdu(&pdu2);}}else if (pduAttachData.GetPduLength() == 0){//正常获取群信息的返回CMsgConn* pConn = CImUserManager::GetInstance()->GetMsgConnByHandle(user_id, pduAttachData.GetHandle());if (pConn){msg.clear_attach_data();pPdu->SetPBMsg(&msg);pConn->SendPdu(pPdu);}}
}

同步群组聊天信息：群会话如何更新(每有一个人发送消息,则其他人都需要更新会话消息)
分析：
如果和单聊类似，实时更新会有大量操作数据库的成本。
某个群成员发消息时，存储消息成功后，会更新群的最新发言时间。
后续优化，将session放在redis中查询不方便？？？所以要分库分表？？？

void CGroupModel::updateGroupChat(uint32_t nGroupId)
{CDBManager* pDBManager = CDBManager::getInstance();CDBConn* pDBConn = pDBManager->GetDBConn("teamtalk_master");if(pDBConn){uint32_t nNow = (uint32_t)time(NULL);string strSql = "update IMGroup set lastChated=" + int2string(nNow) + " where id=" + int2string(nGroupId);pDBConn->ExecuteUpdate(strSql.c_str());pDBManager->RelDBConn(pDBConn);}else{log("no db connection for teamtalk_master");}
}

CSyncCenter类,群会话的更新-独立的线程

CSyncCenter :: doSyncGroupChat

1. 根据时间节点将需要更新会话的群id和最近的聊天时间读取出来放到map
2. 保存当前时间到CSyncCenter
3. 根据群id从redis读取群成员,然后遍历群成员更新会话信息。

/***  开启内网数据同步以及群组聊天记录同步*/
void CSyncCenter::startSync()
{
#ifdef _WIN32(void)CreateThread(NULL, 0, doSyncGroupChat, NULL, 0, &m_nGroupChatThreadId);
#else(void)pthread_create(&m_nGroupChatThreadId, NULL, doSyncGroupChat, NULL);
#endif
}
//谈取更新的时间大于之前更新的时间点,把对应的群id-nLastChat读取出来放到mapChangedGroup存储
/***  同步群组聊天信息**  @param arg NULL**  @return NULL*/
void* CSyncCenter::doSyncGroupChat(void* arg)
{m_bSyncGroupChatRuning = true;CDBManager* pDBManager = CDBManager::getInstance();map<uint32_t, uint32_t> mapChangedGroup;do{mapChangedGroup.clear();CDBConn* pDBConn = pDBManager->GetDBConn("teamtalk_slave");if(pDBConn){string strSql = "select id, lastChated from IMGroup where status=0 and lastChated >=" + int2string(m_pInstance->getLastUpdateGroup());CResultSet* pResult = pDBConn->ExecuteQuery(strSql.c_str());if(pResult){while (pResult->Next()) {uint32_t nGroupId = pResult->GetInt("id");uint32_t nLastChat = pResult->GetInt("lastChated");if(nLastChat != 0){mapChangedGroup[nGroupId] = nLastChat;}}delete pResult;}pDBManager->RelDBConn(pDBConn);}else{log("no db connection for teamtalk_slave");}m_pInstance->updateLastUpdateGroup(time(NULL));for (auto it=mapChangedGroup.begin(); it!=mapChangedGroup.end(); ++it){uint32_t nGroupId =it->first;list<uint32_t> lsUsers;uint32_t nUpdate = it->second;// 读取该群的群成员CGroupModel::getInstance()->getGroupUser(nGroupId, lsUsers);//遍历群成员，更新Sessionfor (auto it1=lsUsers.begin(); it1!=lsUsers.end(); ++it1){uint32_t nUserId = *it1;uint32_t nSessionId = INVALID_VALUE;nSessionId = CSessionModel::getInstance()->getSessionId(nUserId, nGroupId, IM::BaseDefine::SESSION_TYPE_GROUP, true);if(nSessionId != INVALID_VALUE){CSessionModel::getInstance()->updateSession(nSessionId, nUpdate);}else{CSessionModel::getInstance()->addSession(nUserId, nGroupId, IM::BaseDefine::SESSION_TYPE_GROUP);}}}
//    } while (!m_pInstance->m_pCondSync->waitTime(5*1000));} while (m_pInstance->m_bSyncGroupChatWaitting && !(m_pInstance->m_pCondGroupChat->waitTime(5*1000)));
//    } while(m_pInstance->m_bSyncGroupChatWaitting);m_bSyncGroupChatRuning = false;return NULL;
}

怎么保证数据的不丢失以及重复包？

5. 包头的seq_num字段（包序号），未回复消息列表
6. 业务层的ack机制，收到数据会回复ack

消息未读计数是怎么实现的？

服务器怎么保留消息未读计数（单聊和群聊）
7. 单聊和群聊消息未读计数机制为什么不同？
答：加入该群很大，有1000人，999个人的未读消息计数都+1，效率低下。
8. 单聊消息未读计数机制

• key设计：“unread_” + int2string(nToId)；field：int2string(nFromId)，value：自增1
• 发送消息时，将消息写入 mysql 消息表成功后，更新redis。

void CMessageModel::incMsgCount(uint32_t nFromId, uint32_t nToId)
{CacheManager* pCacheManager = CacheManager::getInstance();// increase message countCacheConn* pCacheConn = pCacheManager->GetCacheConn("unread");if (pCacheConn) {pCacheConn->hincrBy("unread_" + int2string(nToId), int2string(nFromId), 1);pCacheManager->RelCacheConn(pCacheConn);} else {log("no cache connection to increase unread count: %d->%d", nFromId, nToId);}
}
9. 群聊消息未读计数机制
- 群总的消息计数key设计：int2string(nGroupId) + _im_group_msg;
- field：count
- 群内某个成员已经读取的消息计数key设计：
int2string(nUserId) + "_" + int2string(nGroupId) + _im_user_group
- field：count
所以：群内某个成员未读消息计数 = 群总消息数量 - 该成员已经读取的消息数量#define     GROUP_TOTAL_MSG_COUNTER_REDIS_KEY_SUFFIX    "_im_group_msg"
#define     GROUP_USER_MSG_COUNTER_REDIS_KEY_SUFFIX     "_im_user_group"
#define     GROUP_COUNTER_SUBKEY_COUNTER_FIELD          "count"
增加群消息计数
/***  增加群消息计数*  @param nUserId  用户Id*  @param nGroupId 群组Id*  @return 成功返回true，失败返回false*/
bool CGroupMessageModel::incMessageCount(uint32_t nUserId, uint32_t nGroupId)
{bool bRet = false;CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("unread");if (pCacheConn){string strGroupKey = int2string(nGroupId) + GROUP_TOTAL_MSG_COUNTER_REDIS_KEY_SUFFIX;pCacheConn->hincrBy(strGroupKey, GROUP_COUNTER_SUBKEY_COUNTER_FIELD, 1);map<string, string> mapGroupCount;bool bRet = pCacheConn->hgetAll(strGroupKey, mapGroupCount);if(bRet){string strUserKey = int2string(nUserId) + "_" + int2string(nGroupId) + GROUP_USER_MSG_COUNTER_REDIS_KEY_SUFFIX;string strReply = pCacheConn->hmset(strUserKey, mapGroupCount);if(!strReply.empty()){bRet = true;}else{log("hmset %s failed !", strUserKey.c_str());}}else{log("hgetAll %s failed!", strGroupKey.c_str());}pCacheManager->RelCacheConn(pCacheConn);}else{log("no cache connection for unread");}return bRet;
}
获取用户群未读消息计数
/***  获取用户群未读消息计数*  @param nUserId       用户Id*  @param nTotalCnt     总条数*  @param lsUnreadCount 每个会话的未读信息包含了条数，最后一个消息的Id，最后一个消息的类型，最后一个消息的类容*/
void CGroupMessageModel::getUnreadMsgCount(uint32_t nUserId, uint32_t &nTotalCnt, list<IM::BaseDefine::UnreadInfo>& lsUnreadCount)
{list<uint32_t> lsGroupId;CGroupModel::getInstance()->getUserGroupIds(nUserId, lsGroupId, 0);uint32_t nCount = 0;CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("unread");if (pCacheConn){for(auto it=lsGroupId.begin(); it!=lsGroupId.end(); ++it){uint32_t nGroupId = *it;string strGroupKey = int2string(nGroupId) + GROUP_TOTAL_MSG_COUNTER_REDIS_KEY_SUFFIX;string strGroupCnt = pCacheConn->hget(strGroupKey, GROUP_COUNTER_SUBKEY_COUNTER_FIELD);if(strGroupCnt.empty()){
//                log("hget %s : count failed !", strGroupKey.c_str());continue;}uint32_t nGroupCnt = (uint32_t)(atoi(strGroupCnt.c_str()));string strUserKey = int2string(nUserId) + "_" + int2string(nGroupId) + GROUP_USER_MSG_COUNTER_REDIS_KEY_SUFFIX;string strUserCnt = pCacheConn->hget(strUserKey, GROUP_COUNTER_SUBKEY_COUNTER_FIELD);uint32_t nUserCnt = ( strUserCnt.empty() ? 0 : ((uint32_t)atoi(strUserCnt.c_str())) );if(nGroupCnt >= nUserCnt) {nCount = nGroupCnt - nUserCnt;}if(nCount > 0){IM::BaseDefine::UnreadInfo cUnreadInfo;cUnreadInfo.set_session_id(nGroupId);cUnreadInfo.set_session_type(IM::BaseDefine::SESSION_TYPE_GROUP);cUnreadInfo.set_unread_cnt(nCount);nTotalCnt += nCount;string strMsgData;uint32_t nMsgId;IM::BaseDefine::MsgType nType;uint32_t nFromId;getLastMsg(nGroupId, nMsgId, strMsgData, nType, nFromId);if(IM::BaseDefine::MsgType_IsValid(nType)){cUnreadInfo.set_latest_msg_id(nMsgId);cUnreadInfo.set_latest_msg_data(strMsgData);cUnreadInfo.set_latest_msg_type(nType);cUnreadInfo.set_latest_msg_from_user_id(nFromId);lsUnreadCount.push_back(cUnreadInfo);}else{log("invalid msgType. userId=%u, groupId=%u, msgType=%u, msgId=%u", nUserId, nGroupId, nType, nMsgId);}}}pCacheManager->RelCacheConn(pCacheConn);}else{log("no cache connection for unread");}
}
清除未读消息
单聊和群聊清除未读消息都是调用如下函数
1. 单聊直接删掉key
2. 群聊将该成员的已读消息数量设置成群总消息数量。
m_handler_map.insert(make_pair(uint32_t(CID_MSG_READ_ACK), DB_PROXY::clearUnreadMsgCounter));
void CUserModel::clearUserCounter(uint32_t nUserId, uint32_t nPeerId, IM::BaseDefine::SessionType nSessionType)
{if(IM::BaseDefine::SessionType_IsValid(nSessionType)){CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("unread");if (pCacheConn){// Clear P2P msg Counterif(nSessionType == IM::BaseDefine::SESSION_TYPE_SINGLE){int nRet = pCacheConn->hdel("unread_" + int2string(nUserId), int2string(nPeerId));if(!nRet){log("hdel failed %d->%d", nPeerId, nUserId);}}// Clear Group msg Counterelse if(nSessionType == IM::BaseDefine::SESSION_TYPE_GROUP){string strGroupKey = int2string(nPeerId) + GROUP_TOTAL_MSG_COUNTER_REDIS_KEY_SUFFIX;map<string, string> mapGroupCount;bool bRet = pCacheConn->hgetAll(strGroupKey, mapGroupCount);if(bRet){string strUserKey = int2string(nUserId) + "_" + int2string(nPeerId) + GROUP_USER_MSG_COUNTER_REDIS_KEY_SUFFIX;string strReply = pCacheConn->hmset(strUserKey, mapGroupCount);if(strReply.empty()) {log("hmset %s failed !", strUserKey.c_str());}}else{log("hgetall %s failed!", strGroupKey.c_str());}}pCacheManager->RelCacheConn(pCacheConn);}else{log("no cache connection for unread");}}else{log("invalid sessionType. userId=%u, fromId=%u, sessionType=%u", nUserId, nPeerId, nSessionType);}
}

群成员管理

1. 为什么使用 redis 管理群成员？
   答：发群消息需要通知群成员，群成员很多从redis获取，提高效率。
2. 群成员管理使用redis 缓存设计，以hash为存储结构，

• key 为 “group_member_”+int2string(nGroupId);
• field 为 userId（用户id）
• Value 为 创建时间：int2string(nCreated)

3. 加入成员:insertNewMember，插入mysql数据库的同时也插入redis缓存

    /***  修改群成员，增加或删除**  @param pPdu      收到的packet包指针*  @param conn_uuid 该包过来的socket 描述符*/void modifyMember(CImPdu* pPdu, uint32_t conn_uuid){IM::Group::IMGroupChangeMemberReq msg;IM::Group::IMGroupChangeMemberRsp msgResp;if(msg.ParseFromArray(pPdu->GetBodyData(), pPdu->GetBodyLength())){uint32_t nUserId = msg.user_id();uint32_t nGroupId = msg.group_id();IM::BaseDefine::GroupModifyType nType = msg.change_type();if (IM::BaseDefine::GroupModifyType_IsValid(nType) &&CGroupModel::getInstance()->isValidateGroupId(nGroupId)) {CImPdu* pPduRes = new CImPdu;uint32_t nCnt = msg.member_id_list_size();set<uint32_t> setUserId;for(uint32_t i=0; i<nCnt;++i){setUserId.insert(msg.member_id_list(i));}list<uint32_t> lsCurUserId;bool bRet = CGroupModel::getInstance()->modifyGroupMember(nUserId, nGroupId, nType, setUserId, lsCurUserId);msgResp.set_user_id(nUserId);msgResp.set_group_id(nGroupId);msgResp.set_change_type(nType);msgResp.set_result_code(bRet?0:1);if(bRet){for(auto it=setUserId.begin(); it!=setUserId.end(); ++it){msgResp.add_chg_user_id_list(*it);}for(auto it=lsCurUserId.begin(); it!=lsCurUserId.end(); ++it){msgResp.add_cur_user_id_list(*it);}}log("userId=%u, groupId=%u, result=%u, changeCount:%u, currentCount=%u",nUserId, nGroupId,  bRet?0:1, msgResp.chg_user_id_list_size(), msgResp.cur_user_id_list_size());msgResp.set_attach_data(msg.attach_data());pPduRes->SetPBMsg(&msgResp);pPduRes->SetSeqNum(pPdu->GetSeqNum());pPduRes->SetServiceId(IM::BaseDefine::SID_GROUP);pPduRes->SetCommandId(IM::BaseDefine::CID_GROUP_CHANGE_MEMBER_RESPONSE);CProxyConn::AddResponsePdu(conn_uuid, pPduRes);}else{log("invalid groupModifyType or groupId. userId=%u, groupId=%u, groupModifyType=%u", nUserId, nGroupId, nType);}}else{log("parse pb failed");}}
bool CGroupModel::insertNewMember(uint32_t nGroupId, set<uint32_t>& setUsers)
{bool bRet = false;uint32_t nUserCnt = (uint32_t)setUsers.size();if(nGroupId != INVALID_VALUE &&  nUserCnt > 0){CDBManager* pDBManager = CDBManager::getInstance();CDBConn* pDBConn = pDBManager->GetDBConn("teamtalk_slave");if (pDBConn){uint32_t nCreated = (uint32_t)time(NULL);// 获取 已经存在群里的用户string strClause;bool bFirst = true;for (auto it=setUsers.begin(); it!=setUsers.end(); ++it){if(bFirst){bFirst = false;strClause = int2string(*it);}else{strClause += ("," + int2string(*it));}}string strSql = "select userId from IMGroupMember where groupId=" + int2string(nGroupId) + " and userId in (" + strClause + ")";CResultSet* pResult = pDBConn->ExecuteQuery(strSql.c_str());set<uint32_t> setHasUser;if(pResult){while (pResult->Next()) {setHasUser.insert(pResult->GetInt("userId"));}delete pResult;}else{log("no result for sql:%s", strSql.c_str());}pDBManager->RelDBConn(pDBConn);pDBConn = pDBManager->GetDBConn("teamtalk_master");if (pDBConn){CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("group_member");if (pCacheConn){// 设置已经存在群中人的状态if (!setHasUser.empty()){strClause.clear();bFirst = true;for (auto it=setHasUser.begin(); it!=setHasUser.end(); ++it) {if(bFirst){bFirst = false;strClause = int2string(*it);}else{strClause += ("," + int2string(*it));}}strSql = "update IMGroupMember set status=0, updated="+int2string(nCreated)+" where groupId=" + int2string(nGroupId) + " and userId in (" + strClause + ")";pDBConn->ExecuteUpdate(strSql.c_str());}strSql = "insert into IMGroupMember(`groupId`, `userId`, `status`, `created`, `updated`) values\(?,?,?,?,?)";//插入新成员auto it = setUsers.begin();uint32_t nStatus = 0;uint32_t nIncMemberCnt = 0;for (;it != setUsers.end();){uint32_t nUserId = *it;if(setHasUser.find(nUserId) == setHasUser.end()){CPrepareStatement* pStmt = new CPrepareStatement();if (pStmt->Init(pDBConn->GetMysql(), strSql)){uint32_t index = 0;pStmt->SetParam(index++, nGroupId);pStmt->SetParam(index++, nUserId);pStmt->SetParam(index++, nStatus);pStmt->SetParam(index++, nCreated);pStmt->SetParam(index++, nCreated);pStmt->ExecuteUpdate();++nIncMemberCnt;delete pStmt;}else{setUsers.erase(it++);delete pStmt;continue;}}++it;}if(nIncMemberCnt != 0){strSql = "update IMGroup set userCnt=userCnt+" + int2string(nIncMemberCnt) + " where id="+int2string(nGroupId);pDBConn->ExecuteUpdate(strSql.c_str());}//更新一份到redis中string strKey = "group_member_"+int2string(nGroupId);for(auto it = setUsers.begin(); it!=setUsers.end(); ++it){pCacheConn->hset(strKey, int2string(*it), int2string(nCreated));}pCacheManager->RelCacheConn(pCacheConn);bRet = true;}else{log("no cache connection");}pDBManager->RelDBConn(pDBConn);}else{log("no db connection for teamtalk_master");}}else{log("no db connection for teamtalk_slave");}}return bRet;
}

4. 删除成员:removeMember；从mysql数据库删除的同时,也从redis缓存删除

bool CGroupModel::removeMember(uint32_t nGroupId, set<uint32_t> &setUser, list<uint32_t>& lsCurUserId)
{if(setUser.size() <= 0){return true;}bool bRet = false;CDBManager* pDBManager = CDBManager::getInstance();CDBConn* pDBConn = pDBManager->GetDBConn("teamtalk_master");if(pDBConn){CacheManager* pCacheManager = CacheManager::getInstance();CacheConn* pCacheConn = pCacheManager->GetCacheConn("group_member");if (pCacheConn){string strClause ;bool bFirst = true;for(auto it= setUser.begin(); it!=setUser.end();++it){if (bFirst) {bFirst = false;strClause = int2string(*it);}else{strClause += ("," + int2string(*it));}}string strSql = "update IMGroupMember set status=1 where  groupId =" + int2string(nGroupId) + " and userId in(" + strClause + ")";pDBConn->ExecuteUpdate(strSql.c_str());//从redis中删除成员string strKey = "group_member_"+ int2string(nGroupId);for (auto it=setUser.begin(); it!=setUser.end(); ++it) {string strField = int2string(*it);pCacheConn->hdel(strKey, strField);}pCacheManager->RelCacheConn(pCacheConn);bRet = true;}else{log("no cache connection");}pDBManager->RelDBConn(pDBConn);if (bRet){getGroupUser(nGroupId,lsCurUserId);}}else{log("no db connection for teamtalk_master");}return bRet;
}

redis连接池设计
6. 为什么使用连接池？
答：对象复用，减小频繁创建链接释放链接的开销时间。
7. CacheInstances=unread,group_set,token,sync,group_member 5 个连接池
8. 为什么分开不同的db redis？
答：方便扩展。
9. pool_name的意义？
答：抽象，不必关注redis是否分布式。

class CacheManager {
public:virtual ~CacheManager();static CacheManager* getInstance();int Init();CacheConn* GetCacheConn(const char* pool_name);void RelCacheConn(CacheConn* pCacheConn);
private:CacheManager();private:static CacheManager*         s_cache_manager;map<string, CachePool*>        m_cache_pool_map;
};
int CacheManager::Init()
{CConfigFileReader config_file("dbproxyserver.conf");//CacheInstances=unread,group_set,token,sync,group_memberchar* cache_instances = config_file.GetConfigName("CacheInstances");if (!cache_instances) {log("not configure CacheIntance");return 1;}char host[64];char port[64];char db[64];char maxconncnt[64];CStrExplode instances_name(cache_instances, ',');for (uint32_t i = 0; i < instances_name.GetItemCnt(); i++) {char* pool_name = instances_name.GetItem(i);//printf("%s", pool_name);snprintf(host, 64, "%s_host", pool_name);snprintf(port, 64, "%s_port", pool_name);snprintf(db, 64, "%s_db", pool_name);snprintf(maxconncnt, 64, "%s_maxconncnt", pool_name);char* cache_host = config_file.GetConfigName(host);char* str_cache_port = config_file.GetConfigName(port);char* str_cache_db = config_file.GetConfigName(db);char* str_max_conn_cnt = config_file.GetConfigName(maxconncnt);if (!cache_host || !str_cache_port || !str_cache_db || !str_max_conn_cnt) {log("not configure cache instance: %s", pool_name);return 2;}CachePool* pCachePool = new CachePool(pool_name, cache_host, atoi(str_cache_port),atoi(str_cache_db), atoi(str_max_conn_cnt));if (pCachePool->Init()) {log("Init cache pool failed");return 3;}m_cache_pool_map.insert(make_pair(pool_name, pCachePool));}return 0;
}