merge with develop branch.

This commit is contained in:
Shuduo Sang 2020-12-10 00:09:42 +08:00
commit dc6112bd3e
219 changed files with 7525 additions and 4369 deletions

View File

@ -13,7 +13,6 @@ ENDIF ()
SET(TD_ACCOUNT FALSE)
SET(TD_ADMIN FALSE)
SET(TD_GRANT FALSE)
SET(TD_SYNC TRUE)
SET(TD_MQTT TRUE)
SET(TD_TSDB_PLUGINS FALSE)

42
Jenkinsfile vendored
View File

@ -175,7 +175,47 @@ pipeline {
}
}
stage('arm64_build'){
agent{label 'arm64'}
steps{
sh '''
cd ${WK}
git fetch
git checkout develop
git pull
cd ${WKC}
git fetch
git checkout develop
git pull
git submodule update
cd ${WKC}/packaging
./release.sh -v cluster -c aarch64 -n 2.0.0.0 -m 2.0.0.0
'''
}
}
stage('arm32_build'){
agent{label 'arm32'}
steps{
catchError(buildResult: 'SUCCESS', stageResult: 'FAILURE') {
sh '''
cd ${WK}
git fetch
git checkout develop
git pull
cd ${WKC}
git fetch
git checkout develop
git pull
git submodule update
cd ${WKC}/packaging
./release.sh -v cluster -c aarch32 -n 2.0.0.0 -m 2.0.0.0
'''
}
}
}
}
}

View File

@ -13,10 +13,6 @@ IF (TD_GRANT)
ADD_DEFINITIONS(-D_GRANT)
ENDIF ()
IF (TD_SYNC)
ADD_DEFINITIONS(-D_SYNC)
ENDIF ()
IF (TD_MQTT)
ADD_DEFINITIONS(-D_MQTT)
ENDIF ()

View File

@ -47,11 +47,6 @@ IF (${MQTT} MATCHES "false")
MESSAGE(STATUS "build without mqtt module")
ENDIF ()
IF (${SYNC} MATCHES "false")
SET(TD_SYNC FALSE)
MESSAGE(STATUS "build without sync module")
ENDIF ()
IF (${RANDOM_FILE_FAIL} MATCHES "true")
SET(TD_RANDOM_FILE_FAIL TRUE)
MESSAGE(STATUS "build with random-file-fail enabled")

View File

@ -32,7 +32,7 @@ ELSEIF (TD_WINDOWS)
#INSTALL(TARGETS taos RUNTIME DESTINATION driver)
#INSTALL(TARGETS shell RUNTIME DESTINATION .)
IF (TD_MVN_INSTALLED)
INSTALL(FILES ${LIBRARY_OUTPUT_PATH}/taos-jdbcdriver-2.0.13-dist.jar DESTINATION connector/jdbc)
INSTALL(FILES ${LIBRARY_OUTPUT_PATH}/taos-jdbcdriver-2.0.15-dist.jar DESTINATION connector/jdbc)
ENDIF ()
ELSEIF (TD_DARWIN)
SET(TD_MAKE_INSTALL_SH "${TD_COMMUNITY_DIR}/packaging/tools/make_install.sh")

View File

@ -4,7 +4,7 @@ PROJECT(TDengine)
IF (DEFINED VERNUMBER)
SET(TD_VER_NUMBER ${VERNUMBER})
ELSE ()
SET(TD_VER_NUMBER "2.0.8.0")
SET(TD_VER_NUMBER "2.0.9.0")
ENDIF ()
IF (DEFINED VERCOMPATIBLE)

View File

@ -44,6 +44,8 @@ CREATE TABLE d1001 USING meters TAGS ("Beijing.Chaoyang", 2);
```
其中d1001是表名meters是超级表的表名后面紧跟标签Location的具体标签值”Beijing.Chaoyang"标签groupId的具体标签值2。虽然在创建表时需要指定标签值但可以事后修改。详细细则请见 TAOS SQL。
**注意:**目前 TDengine 没有从技术层面限制使用一个 database dbA的超级表作为模板建立另一个 database dbB的子表后续会禁止这种用法不建议使用这种方法建表。
TDengine建议将数据采集点的全局唯一ID作为表名(比如设备序列号。但对于有的场景并没有唯一的ID可以将多个ID组合成一个唯一的ID。不建议将具有唯一性的ID作为标签值。
**自动建表**:在某些特殊场景中,用户在写数据时并不确定某个数据采集点的表是否存在,此时可在写入数据时使用自动建表语法来创建不存在的表,若该表已存在则不会建立新表。比如:

View File

@ -95,6 +95,7 @@ TDengine系统后台服务由taosd提供可以在配置文件taos.cfg里修
- logKeepDays日志文件的最长保存时间。大于0时日志文件会被重命名为taosdlog.xxx其中xxx为日志文件最后修改的时间戳单位为秒。默认值0天。
- maxSQLLength单条SQL语句允许最长限制。默认值65380字节。
- telemetryReporting: 是否允许 TDengine 采集和上报基本使用信息0表示不允许1表示允许。 默认值1。
- stream: 是否启用连续查询流计算功能0表示不允许1表示允许。 默认值1。
- queryBufferSize: 为所有并发查询占用保留的内存大小。计算规则可以根据实际应用可能的最大并发数和表的数字相乘,再乘 170 。单位为字节。
**注意:**对于端口TDengine会使用从serverPort起13个连续的TCP和UDP端口号请务必在防火墙打开。因此如果是缺省配置需要打开从6030都6042共13个端口而且必须TCP和UDP都打开。

View File

@ -37,7 +37,7 @@ fqdn h1.taosdata.com
// 配置本数据节点的端口号缺省是6030
serverPort 6030
// 服务端节点数为偶数的时候,需要配置请参考《Arbitrator的使用》的部分
// 使用场景请参考《Arbitrator的使用》的部分
arbitrator ha.taosdata.com:6042
```

View File

@ -157,7 +157,7 @@ C/C++的API类似于MySQL的C API。应用程序使用时需要包含TDengine
获取最近一次API调用失败的原因返回值为错误代码。
**注意**:对于单个数据库连接在同一时刻只能有一个线程使用该连接调用API否则会有未定义的行为出现并可能导致客户端crash。客户端应用可以通过建立多个连接进行多线程的数据写入或查询处理
**注意**:对于每个数据库应用2.0及以上版本 TDengine 推荐只建立一个连接。同时在应用中将该连接 (TAOS*) 结构体传递到不同的线程共享使用。基于 TAOS 结构体发出的查询、写入等操作具有多线程安全性。C 语言的连接器可以按照需求动态建立面向数据库的新连接(该过程对用户不可见),同时建议只有在程序最后退出的时候才调用 taos_close 关闭连接
### 异步查询API

View File

@ -36,16 +36,20 @@
5. ping服务器FQDN如果没有反应请检查你的网络DNS设置或客户端所在计算机的系统hosts文件
6. 检查防火墙设置确认TCP/UDP 端口6030-6042 是打开的
6. 检查防火墙设置Ubuntu 使用 ufw statusCentOS 使用 firewall-cmd --list-port确认TCP/UDP 端口6030-6042 是打开的
7. 对于Linux上的JDBCODBC, Python, Go等接口类似连接, 确保*libtaos.so*在目录*/usr/local/taos/driver*里, 并且*/usr/local/taos/driver*在系统库函数搜索路径*LD_LIBRARY_PATH*里
8. 对于windows上的JDBC, ODBC, Python, Go等连接确保*C:\TDengine\driver\taos.dll*在你的系统库函数搜索目录里 (建议*taos.dll*放在目录 *C:\Windows\System32*)
9. 如果仍不能排除连接故障请使用命令行工具nc来分别判断指定端口的TCP和UDP连接是否通畅
检查UDP端口连接是否工作`nc -vuz {hostIP} {port} `
检查服务器侧TCP端口连接是否工作`nc -l {port}`
检查客户端侧TCP端口连接是否工作`nc {hostIP} {port}`
9. 如果仍不能排除连接故障
* Linux 系统请使用命令行工具nc来分别判断指定端口的TCP和UDP连接是否通畅
检查UDP端口连接是否工作`nc -vuz {hostIP} {port} `
检查服务器侧TCP端口连接是否工作`nc -l {port}`
检查客户端侧TCP端口连接是否工作`nc {hostIP} {port}`
* Windows 系统请使用 PowerShell 命令 Net-TestConnection -ComputerName {fqdn} -Port {port} 检测服务段端口是否访问
10. 也可以使用taos程序内嵌的网络连通检测功能来验证服务器和客户端之间指定的端口连接是否通畅包括TCP和UDP[TDengine 内嵌网络检测工具使用指南](https://www.taosdata.com/blog/2020/09/08/1816.html)。
@ -101,7 +105,20 @@ Connection = DriverManager.getConnection(url, properties);
<version>2.0.4</version>
</dependency>
```
## 14. 怎么报告问题?
## 14. taos connect failed, reason: invalid timestamp
常见原因是服务器和客户端时间没有校准可以通过和时间服务器同步的方式Linux 下使用 ntpdate 命令Windows 在系统时间设置中选择自动同步)校准。
## 15. 表名显示不全
由于 taos shell 在终端中显示宽度有限,有可能比较长的表名显示不全,如果按照显示的不全的表名进行相关操作会发生 Table does not exist 错误。解决方法可以是通过修改 taos.cfg 文件中的设置项 maxBinaryDisplayWidth 或者直接输入命令 set max_binary_display_width 100。或者在命令结尾使用 \G 参数来调整结果的显示方式。
## 16. 怎么报告问题?
如果 FAQ 中的信息不能够帮到您,需要 TDengine 技术团队的技术支持与协助,请将以下两个目录中内容打包:
1. /var/log/taos
2. /etc/taos

View File

@ -236,7 +236,7 @@
# httpDebugFlag 131
# debug flag for monitor
# monitorDebugFlag 131
# monDebugFlag 131
# debug flag for query
# qDebugflag 131
@ -251,7 +251,7 @@
# cqDebugFlag 131
# enable/disable recording the SQL in taos client
# tscEnableRecordSql 0
# enableRecordSql 0
# generate core file when service crash
# enableCoreFile 1
@ -260,4 +260,10 @@
# maxBinaryDisplayWidth 30
# enable/disable telemetry reporting
# telemetryReporting 1
# telemetryReporting 1
# enable/disable stream (continuous query)
# stream 1
# only 50% CPU resources will be used in query processing
# halfCoresForQuery 0

View File

@ -1,6 +1,6 @@
name: tdengine
base: core18
version: '2.0.8.0'
version: '2.0.9.0'
icon: snap/gui/t-dengine.svg
summary: an open-source big data platform designed and optimized for IoT.
description: |
@ -72,7 +72,7 @@ parts:
- usr/bin/taosd
- usr/bin/taos
- usr/bin/taosdemo
- usr/lib/libtaos.so.2.0.8.0
- usr/lib/libtaos.so.2.0.9.0
- usr/lib/libtaos.so.1
- usr/lib/libtaos.so

View File

@ -10,9 +10,7 @@ ADD_SUBDIRECTORY(client)
ADD_SUBDIRECTORY(query)
ADD_SUBDIRECTORY(kit)
ADD_SUBDIRECTORY(plugins)
IF (TD_SYNC)
ADD_SUBDIRECTORY(sync)
ENDIF ()
ADD_SUBDIRECTORY(sync)
ADD_SUBDIRECTORY(balance)
ADD_SUBDIRECTORY(mnode)
ADD_SUBDIRECTORY(vnode)

57
src/balance/inc/bnInt.h Normal file
View File

@ -0,0 +1,57 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_BALANCE_INT_H
#define TDENGINE_BALANCE_INT_H
#ifdef __cplusplus
extern "C" {
#endif
#include "mnodeInt.h"
#include "mnodeDef.h"
#include "mnodeDnode.h"
typedef struct {
int32_t size;
int32_t maxSize;
SDnodeObj **list;
} SBnDnodes;
typedef struct {
void * timer;
bool stop;
pthread_mutex_t mutex;
pthread_cond_t cond;
pthread_t thread;
} SBnThread;
typedef struct {
pthread_mutex_t mutex;
} SBnMgmt;
int32_t bnInit();
void bnCleanUp();
bool bnStart();
void bnCheckStatus();
void bnCheckModules();
extern SBnDnodes tsBnDnodes;
extern void *tsMnodeTmr;
#ifdef __cplusplus
}
#endif
#endif

34
src/balance/inc/bnScore.h Normal file
View File

@ -0,0 +1,34 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_BALANCE_SCORE_H
#define TDENGINE_BALANCE_SCORE_H
#ifdef __cplusplus
extern "C" {
#endif
#include "bnInt.h"
void bnInitDnodes();
void bnCleanupDnodes();
void bnAccquireDnodes();
void bnReleaseDnodes();
float bnTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extraVnode);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -0,0 +1,33 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_BALANCE_THREAD_H
#define TDENGINE_BALANCE_THREAD_H
#ifdef __cplusplus
extern "C" {
#endif
#include "bnInt.h"
int32_t bnInitThread();
void bnCleanupThread();
void bnNotify();
void bnStartTimer(int64_t mseconds);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -15,17 +15,13 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "tutil.h"
#include "tbalance.h"
#include "tref.h"
#include "tsync.h"
#include "ttimer.h"
#include "tglobal.h"
#include "tdataformat.h"
#include "dnode.h"
#include "mnode.h"
#include "mnodeDef.h"
#include "mnodeInt.h"
#include "mnodeDnode.h"
#include "bnInt.h"
#include "bnScore.h"
#include "bnThread.h"
#include "mnodeDb.h"
#include "mnodeMnode.h"
#include "mnodeSdb.h"
@ -33,36 +29,20 @@
#include "mnodeUser.h"
#include "mnodeVgroup.h"
/*
* once sdb work as mater, then tsAccessSquence reset to zero
* increase tsAccessSquence every balance interval
*/
extern void * tsMnodeTmr;
static void * tsBalanceTimer = NULL;
static int32_t tsBalanceDnodeListSize = 0;
static SDnodeObj ** tsBalanceDnodeList = NULL;
static int32_t tsBalanceDnodeListMallocSize = 16;
static pthread_mutex_t tsBalanceMutex;
extern int64_t tsDnodeRid;
extern int64_t tsSdbRid;
static SBnMgmt tsBnMgmt;
static void bnMonitorDnodeModule();
static void balanceStartTimer(int64_t mseconds);
static void balanceInitDnodeList();
static void balanceCleanupDnodeList();
static void balanceAccquireDnodeList();
static void balanceReleaseDnodeList();
static void balanceMonitorDnodeModule();
static float balanceTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extraVnode);
static int32_t balanceGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn);
static int32_t balanceRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn);
static void balanceLock() {
pthread_mutex_lock(&tsBalanceMutex);
static void bnLock() {
pthread_mutex_lock(&tsBnMgmt.mutex);
}
static void balanceUnLock() {
pthread_mutex_unlock(&tsBalanceMutex);
static void bnUnLock() {
pthread_mutex_unlock(&tsBnMgmt.mutex);
}
static bool balanceCheckFree(SDnodeObj *pDnode) {
static bool bnCheckFree(SDnodeObj *pDnode) {
if (pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) {
mError("dnode:%d, status:%s not available", pDnode->dnodeId, mnodeGetDnodeStatusStr(pDnode->status));
return false;
@ -86,7 +66,7 @@ static bool balanceCheckFree(SDnodeObj *pDnode) {
return true;
}
static void balanceDiscardVnode(SVgObj *pVgroup, SVnodeGid *pVnodeGid) {
static void bnDiscardVnode(SVgObj *pVgroup, SVnodeGid *pVnodeGid) {
mDebug("vgId:%d, dnode:%d is dropping", pVgroup->vgId, pVnodeGid->dnodeId);
SDnodeObj *pDnode = mnodeGetDnode(pVnodeGid->dnodeId);
@ -111,27 +91,26 @@ static void balanceDiscardVnode(SVgObj *pVgroup, SVnodeGid *pVnodeGid) {
mnodeUpdateVgroup(pVgroup);
}
static void balanceSwapVnodeGid(SVnodeGid *pVnodeGid1, SVnodeGid *pVnodeGid2) {
static void bnSwapVnodeGid(SVnodeGid *pVnodeGid1, SVnodeGid *pVnodeGid2) {
// SVnodeGid tmp = *pVnodeGid1;
// *pVnodeGid1 = *pVnodeGid2;
// *pVnodeGid2 = tmp;
}
int32_t balanceAllocVnodes(SVgObj *pVgroup) {
int32_t bnAllocVnodes(SVgObj *pVgroup) {
static int32_t randIndex = 0;
int32_t dnode = 0;
int32_t vnodes = 0;
balanceLock();
balanceAccquireDnodeList();
bnLock();
bnAccquireDnodes();
mDebug("db:%s, try alloc %d vnodes to vgroup, dnodes total:%d, avail:%d", pVgroup->dbName, pVgroup->numOfVnodes,
mnodeGetDnodesNum(), tsBalanceDnodeListSize);
mnodeGetDnodesNum(), tsBnDnodes.size);
for (int32_t i = 0; i < pVgroup->numOfVnodes; ++i) {
for (; dnode < tsBalanceDnodeListSize; ++dnode) {
SDnodeObj *pDnode = tsBalanceDnodeList[dnode];
if (balanceCheckFree(pDnode)) {
for (; dnode < tsBnDnodes.size; ++dnode) {
SDnodeObj *pDnode = tsBnDnodes.list[dnode];
if (bnCheckFree(pDnode)) {
SVnodeGid *pVnodeGid = pVgroup->vnodeGid + i;
pVnodeGid->dnodeId = pDnode->dnodeId;
pVnodeGid->pDnode = pDnode;
@ -148,8 +127,8 @@ int32_t balanceAllocVnodes(SVgObj *pVgroup) {
}
if (vnodes != pVgroup->numOfVnodes) {
balanceReleaseDnodeList();
balanceUnLock();
bnReleaseDnodes();
bnUnLock();
mDebug("db:%s, need vnodes:%d, but alloc:%d", pVgroup->dbName, pVgroup->numOfVnodes, vnodes);
@ -179,33 +158,33 @@ int32_t balanceAllocVnodes(SVgObj *pVgroup) {
if (pVgroup->numOfVnodes == 1) {
} else if (pVgroup->numOfVnodes == 2) {
if (randIndex++ % 2 == 0) {
balanceSwapVnodeGid(pVgroup->vnodeGid, pVgroup->vnodeGid + 1);
bnSwapVnodeGid(pVgroup->vnodeGid, pVgroup->vnodeGid + 1);
}
} else {
int32_t randVal = randIndex++ % 6;
if (randVal == 1) { // 1, 0, 2
balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1);
bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1);
} else if (randVal == 2) { // 1, 2, 0
balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1);
balanceSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2);
bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1);
bnSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2);
} else if (randVal == 3) { // 2, 1, 0
balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2);
bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2);
} else if (randVal == 4) { // 2, 0, 1
balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2);
balanceSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2);
bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2);
bnSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2);
}
if (randVal == 5) { // 0, 2, 1
balanceSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2);
bnSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2);
} else {
} // 0, 1, 2
}
balanceReleaseDnodeList();
balanceUnLock();
bnReleaseDnodes();
bnUnLock();
return TSDB_CODE_SUCCESS;
}
static bool balanceCheckVgroupReady(SVgObj *pVgroup, SVnodeGid *pRmVnode) {
static bool bnCheckVgroupReady(SVgObj *pVgroup, SVnodeGid *pRmVnode) {
if (pVgroup->lbTime + 5 * tsStatusInterval > tsAccessSquence) {
return false;
}
@ -232,7 +211,7 @@ static bool balanceCheckVgroupReady(SVgObj *pVgroup, SVnodeGid *pRmVnode) {
* desc: remove one vnode from vgroup
* all vnodes in vgroup should in ready state, except the balancing one
**/
static int32_t balanceRemoveVnode(SVgObj *pVgroup) {
static int32_t bnRemoveVnode(SVgObj *pVgroup) {
if (pVgroup->numOfVnodes <= 1) return -1;
SVnodeGid *pRmVnode = NULL;
@ -274,17 +253,17 @@ static int32_t balanceRemoveVnode(SVgObj *pVgroup) {
pSelVnode = pRmVnode;
}
if (!balanceCheckVgroupReady(pVgroup, pSelVnode)) {
if (!bnCheckVgroupReady(pVgroup, pSelVnode)) {
mDebug("vgId:%d, is not ready", pVgroup->vgId);
return -1;
} else {
mDebug("vgId:%d, is ready, discard dnode:%d", pVgroup->vgId, pSelVnode->dnodeId);
balanceDiscardVnode(pVgroup, pSelVnode);
bnDiscardVnode(pVgroup, pSelVnode);
return TSDB_CODE_SUCCESS;
}
}
static bool balanceCheckDnodeInVgroup(SDnodeObj *pDnode, SVgObj *pVgroup) {
static bool bnCheckDnodeInVgroup(SDnodeObj *pDnode, SVgObj *pVgroup) {
for (int32_t i = 0; i < pVgroup->numOfVnodes; ++i) {
SVnodeGid *pGid = &pVgroup->vnodeGid[i];
if (pGid->dnodeId == 0) break;
@ -299,13 +278,13 @@ static bool balanceCheckDnodeInVgroup(SDnodeObj *pDnode, SVgObj *pVgroup) {
/**
* desc: add vnode to vgroup, find a new one if dest dnode is null
**/
static int32_t balanceAddVnode(SVgObj *pVgroup, SDnodeObj *pSrcDnode, SDnodeObj *pDestDnode) {
static int32_t bnAddVnode(SVgObj *pVgroup, SDnodeObj *pSrcDnode, SDnodeObj *pDestDnode) {
if (pDestDnode == NULL) {
for (int32_t i = 0; i < tsBalanceDnodeListSize; ++i) {
SDnodeObj *pDnode = tsBalanceDnodeList[i];
for (int32_t i = 0; i < tsBnDnodes.size; ++i) {
SDnodeObj *pDnode = tsBnDnodes.list[i];
if (pDnode == pSrcDnode) continue;
if (balanceCheckDnodeInVgroup(pDnode, pVgroup)) continue;
if (!balanceCheckFree(pDnode)) continue;
if (bnCheckDnodeInVgroup(pDnode, pVgroup)) continue;
if (!bnCheckFree(pDnode)) continue;
pDestDnode = pDnode;
mDebug("vgId:%d, add vnode to dnode:%d", pVgroup->vgId, pDnode->dnodeId);
@ -333,25 +312,25 @@ static int32_t balanceAddVnode(SVgObj *pVgroup, SDnodeObj *pSrcDnode, SDnodeObj
return TSDB_CODE_SUCCESS;
}
static bool balanceMonitorBalance() {
if (tsBalanceDnodeListSize < 2) return false;
static bool bnMonitorBalance() {
if (tsBnDnodes.size < 2) return false;
for (int32_t src = tsBalanceDnodeListSize - 1; src >= 0; --src) {
SDnodeObj *pDnode = tsBalanceDnodeList[src];
mDebug("%d-dnode:%d, state:%s, score:%.1f, numOfCores:%d, openVnodes:%d", tsBalanceDnodeListSize - src - 1,
for (int32_t src = tsBnDnodes.size - 1; src >= 0; --src) {
SDnodeObj *pDnode = tsBnDnodes.list[src];
mDebug("%d-dnode:%d, state:%s, score:%.1f, numOfCores:%d, openVnodes:%d", tsBnDnodes.size - src - 1,
pDnode->dnodeId, mnodeGetDnodeStatusStr(pDnode->status), pDnode->score, pDnode->numOfCores,
pDnode->openVnodes);
}
float scoresDiff = tsBalanceDnodeList[tsBalanceDnodeListSize - 1]->score - tsBalanceDnodeList[0]->score;
float scoresDiff = tsBnDnodes.list[tsBnDnodes.size - 1]->score - tsBnDnodes.list[0]->score;
if (scoresDiff < 0.01) {
mDebug("all dnodes:%d is already balanced, scoresDiff:%f", tsBalanceDnodeListSize, scoresDiff);
mDebug("all dnodes:%d is already balanced, scoresDiff:%f", tsBnDnodes.size, scoresDiff);
return false;
}
for (int32_t src = tsBalanceDnodeListSize - 1; src > 0; --src) {
SDnodeObj *pSrcDnode = tsBalanceDnodeList[src];
float srcScore = balanceTryCalcDnodeScore(pSrcDnode, -1);
for (int32_t src = tsBnDnodes.size - 1; src > 0; --src) {
SDnodeObj *pSrcDnode = tsBnDnodes.list[src];
float srcScore = bnTryCalcDnodeScore(pSrcDnode, -1);
if (tsEnableBalance == 0 && pSrcDnode->status != TAOS_DN_STATUS_DROPPING) {
continue;
}
@ -362,19 +341,19 @@ static bool balanceMonitorBalance() {
pIter = mnodeGetNextVgroup(pIter, &pVgroup);
if (pVgroup == NULL) break;
if (balanceCheckDnodeInVgroup(pSrcDnode, pVgroup)) {
if (bnCheckDnodeInVgroup(pSrcDnode, pVgroup)) {
for (int32_t dest = 0; dest < src; dest++) {
SDnodeObj *pDestDnode = tsBalanceDnodeList[dest];
if (balanceCheckDnodeInVgroup(pDestDnode, pVgroup)) continue;
SDnodeObj *pDestDnode = tsBnDnodes.list[dest];
if (bnCheckDnodeInVgroup(pDestDnode, pVgroup)) continue;
float destScore = balanceTryCalcDnodeScore(pDestDnode, 1);
float destScore = bnTryCalcDnodeScore(pDestDnode, 1);
if (srcScore + 0.0001 < destScore) continue;
if (!balanceCheckFree(pDestDnode)) continue;
if (!bnCheckFree(pDestDnode)) continue;
mDebug("vgId:%d, balance from dnode:%d to dnode:%d, srcScore:%.1f:%.1f, destScore:%.1f:%.1f",
pVgroup->vgId, pSrcDnode->dnodeId, pDestDnode->dnodeId, pSrcDnode->score,
srcScore, pDestDnode->score, destScore);
balanceAddVnode(pVgroup, pSrcDnode, pDestDnode);
bnAddVnode(pVgroup, pSrcDnode, pDestDnode);
mnodeDecVgroupRef(pVgroup);
mnodeCancelGetNextVgroup(pIter);
return true;
@ -392,7 +371,7 @@ static bool balanceMonitorBalance() {
// 1. reset balanceAccessSquence to zero
// 2. reset state of dnodes to offline
// 3. reset lastAccess of dnodes to zero
void balanceReset() {
void bnReset() {
void * pIter = NULL;
SDnodeObj *pDnode = NULL;
while (1) {
@ -413,7 +392,7 @@ void balanceReset() {
tsAccessSquence = 0;
}
static int32_t balanceMonitorVgroups() {
static int32_t bnMonitorVgroups() {
void * pIter = NULL;
SVgObj *pVgroup = NULL;
bool hasUpdatingVgroup = false;
@ -429,11 +408,11 @@ static int32_t balanceMonitorVgroups() {
if (vgReplica > dbReplica) {
mInfo("vgId:%d, replica:%d numOfVnodes:%d, try remove one vnode", pVgroup->vgId, dbReplica, vgReplica);
hasUpdatingVgroup = true;
code = balanceRemoveVnode(pVgroup);
code = bnRemoveVnode(pVgroup);
} else if (vgReplica < dbReplica) {
mInfo("vgId:%d, replica:%d numOfVnodes:%d, try add one vnode", pVgroup->vgId, dbReplica, vgReplica);
hasUpdatingVgroup = true;
code = balanceAddVnode(pVgroup, NULL, NULL);
code = bnAddVnode(pVgroup, NULL, NULL);
}
mnodeDecVgroupRef(pVgroup);
@ -446,7 +425,7 @@ static int32_t balanceMonitorVgroups() {
return hasUpdatingVgroup;
}
static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) {
static bool bnMonitorDnodeDropping(SDnodeObj *pDnode) {
mDebug("dnode:%d, in dropping state", pDnode->dnodeId);
void * pIter = NULL;
@ -456,7 +435,7 @@ static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) {
pIter = mnodeGetNextVgroup(pIter, &pVgroup);
if (pVgroup == NULL) break;
hasThisDnode = balanceCheckDnodeInVgroup(pDnode, pVgroup);
hasThisDnode = bnCheckDnodeInVgroup(pDnode, pVgroup);
mnodeDecVgroupRef(pVgroup);
if (hasThisDnode) {
@ -474,7 +453,7 @@ static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) {
return false;
}
static bool balanceMontiorDropping() {
static bool bnMontiorDropping() {
void *pIter = NULL;
SDnodeObj *pDnode = NULL;
@ -499,7 +478,7 @@ static bool balanceMontiorDropping() {
}
if (pDnode->status == TAOS_DN_STATUS_DROPPING) {
bool ret = balanceMonitorDnodeDropping(pDnode);
bool ret = bnMonitorDnodeDropping(pDnode);
mnodeDecDnodeRef(pDnode);
mnodeCancelGetNextDnode(pIter);
return ret;
@ -509,33 +488,31 @@ static bool balanceMontiorDropping() {
return false;
}
static bool balanceStart() {
bool bnStart() {
if (!sdbIsMaster()) return false;
balanceLock();
bnLock();
bnAccquireDnodes();
balanceAccquireDnodeList();
bnMonitorDnodeModule();
balanceMonitorDnodeModule();
bool updateSoon = balanceMontiorDropping();
bool updateSoon = bnMontiorDropping();
if (!updateSoon) {
updateSoon = balanceMonitorVgroups();
updateSoon = bnMonitorVgroups();
}
if (!updateSoon) {
updateSoon = balanceMonitorBalance();
updateSoon = bnMonitorBalance();
}
balanceReleaseDnodeList();
balanceUnLock();
bnReleaseDnodes();
bnUnLock();
return updateSoon;
}
static void balanceSetVgroupOffline(SDnodeObj* pDnode) {
static void bnSetVgroupOffline(SDnodeObj* pDnode) {
void *pIter = NULL;
while (1) {
SVgObj *pVgroup;
@ -551,10 +528,13 @@ static void balanceSetVgroupOffline(SDnodeObj* pDnode) {
}
}
static void balanceCheckDnodeAccess() {
void bnCheckStatus() {
void * pIter = NULL;
SDnodeObj *pDnode = NULL;
void *dnodeSdb = taosAcquireRef(tsSdbRid, tsDnodeRid);
if (dnodeSdb == NULL) return;
while (1) {
pIter = mnodeGetNextDnode(pIter, &pDnode);
if (pDnode == NULL) break;
@ -564,84 +544,41 @@ static void balanceCheckDnodeAccess() {
pDnode->offlineReason = TAOS_DN_OFF_STATUS_MSG_TIMEOUT;
mInfo("dnode:%d, set to offline state, access seq:%d last seq:%d laststat:%d", pDnode->dnodeId, tsAccessSquence,
pDnode->lastAccess, pDnode->status);
balanceSetVgroupOffline(pDnode);
bnSetVgroupOffline(pDnode);
}
}
mnodeDecDnodeRef(pDnode);
}
taosReleaseRef(tsSdbRid, tsDnodeRid);
}
static void balanceProcessBalanceTimer(void *handle, void *tmrId) {
if (!sdbIsMaster()) return;
tsBalanceTimer = NULL;
tsAccessSquence ++;
balanceCheckDnodeAccess();
bool updateSoon = false;
if (handle == NULL) {
if (tsAccessSquence % tsBalanceInterval == 0) {
mDebug("balance function is scheduled by timer");
updateSoon = balanceStart();
}
} else {
int64_t mseconds = (int64_t)handle;
mDebug("balance function is scheduled by event for %" PRId64 " mseconds arrived", mseconds);
updateSoon = balanceStart();
}
if (updateSoon) {
balanceStartTimer(1000);
} else {
taosTmrReset(balanceProcessBalanceTimer, tsStatusInterval * 1000, NULL, tsMnodeTmr, &tsBalanceTimer);
}
}
static void balanceStartTimer(int64_t mseconds) {
taosTmrReset(balanceProcessBalanceTimer, mseconds, (void *)mseconds, tsMnodeTmr, &tsBalanceTimer);
}
void balanceSyncNotify() {
void bnCheckModules() {
if (sdbIsMaster()) {
balanceLock();
balanceAccquireDnodeList();
balanceMonitorDnodeModule();
balanceReleaseDnodeList();
balanceUnLock();
bnLock();
bnAccquireDnodes();
bnMonitorDnodeModule();
bnReleaseDnodes();
bnUnLock();
}
}
void balanceAsyncNotify() {
balanceStartTimer(500);
}
int32_t balanceInit() {
mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_SCORES, balanceGetScoresMeta);
mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_SCORES, balanceRetrieveScores);
mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_SCORES, mnodeCancelGetNextDnode);
pthread_mutex_init(&tsBalanceMutex, NULL);
balanceInitDnodeList();
balanceStartTimer(2000);
mDebug("balance start fp:%p initialized", balanceProcessBalanceTimer);
balanceReset();
int32_t bnInit() {
pthread_mutex_init(&tsBnMgmt.mutex, NULL);
bnInitDnodes();
bnInitThread();
bnReset();
return 0;
}
void balanceCleanUp() {
if (tsBalanceTimer != NULL) {
taosTmrStopA(&tsBalanceTimer);
pthread_mutex_destroy(&tsBalanceMutex);
tsBalanceTimer = NULL;
mDebug("stop balance timer");
}
balanceCleanupDnodeList();
void bnCleanUp() {
bnCleanupThread();
bnCleanupDnodes();
pthread_mutex_destroy(&tsBnMgmt.mutex);
}
int32_t balanceDropDnode(SDnodeObj *pDnode) {
int32_t bnDropDnode(SDnodeObj *pDnode) {
int32_t totalFreeVnodes = 0;
void * pIter = NULL;
SDnodeObj *pTempDnode = NULL;
@ -650,7 +587,7 @@ int32_t balanceDropDnode(SDnodeObj *pDnode) {
pIter = mnodeGetNextDnode(pIter, &pTempDnode);
if (pTempDnode == NULL) break;
if (pTempDnode != pDnode && balanceCheckFree(pTempDnode)) {
if (pTempDnode != pDnode && bnCheckFree(pTempDnode)) {
totalFreeVnodes += (TSDB_MAX_VNODES - pTempDnode->openVnodes);
}
@ -665,298 +602,17 @@ int32_t balanceDropDnode(SDnodeObj *pDnode) {
pDnode->status = TAOS_DN_STATUS_DROPPING;
mnodeUpdateDnode(pDnode);
balanceStartTimer(1100);
bnStartTimer(1100);
return TSDB_CODE_SUCCESS;
}
static int32_t balanceCalcCpuScore(SDnodeObj *pDnode) {
if (pDnode->cpuAvgUsage < 80)
return 0;
else if (pDnode->cpuAvgUsage < 90)
return 10;
else
return 50;
}
static int32_t balanceCalcMemoryScore(SDnodeObj *pDnode) {
if (pDnode->memoryAvgUsage < 80)
return 0;
else if (pDnode->memoryAvgUsage < 90)
return 10;
else
return 50;
}
static int32_t balanceCalcDiskScore(SDnodeObj *pDnode) {
if (pDnode->diskAvgUsage < 80)
return 0;
else if (pDnode->diskAvgUsage < 90)
return 10;
else
return 50;
}
static int32_t balanceCalcBandwidthScore(SDnodeObj *pDnode) {
if (pDnode->bandwidthUsage < 30)
return 0;
else if (pDnode->bandwidthUsage < 80)
return 10;
else
return 50;
}
static float balanceCalcModuleScore(SDnodeObj *pDnode) {
if (pDnode->numOfCores <= 0) return 0;
if (pDnode->isMgmt) {
return (float)tsMnodeEqualVnodeNum / pDnode->numOfCores;
}
return 0;
}
static float balanceCalcVnodeScore(SDnodeObj *pDnode, int32_t extra) {
if (pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) return 100000000;
if (pDnode->numOfCores <= 0) return 0;
return (float)(pDnode->openVnodes + extra) / pDnode->numOfCores;
}
/**
* calc singe score, such as cpu/memory/disk/bandwitdh/vnode
* 1. get the score config
* 2. if the value is out of range, use border data
* 3. otherwise use interpolation method
**/
void balanceCalcDnodeScore(SDnodeObj *pDnode) {
pDnode->score = balanceCalcCpuScore(pDnode) + balanceCalcMemoryScore(pDnode) + balanceCalcDiskScore(pDnode) +
balanceCalcBandwidthScore(pDnode) + balanceCalcModuleScore(pDnode) +
balanceCalcVnodeScore(pDnode, 0) + pDnode->customScore;
}
float balanceTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extra) {
int32_t systemScore = balanceCalcCpuScore(pDnode) + balanceCalcMemoryScore(pDnode) + balanceCalcDiskScore(pDnode) +
balanceCalcBandwidthScore(pDnode);
float moduleScore = balanceCalcModuleScore(pDnode);
float vnodeScore = balanceCalcVnodeScore(pDnode, extra);
float score = systemScore + moduleScore + vnodeScore + pDnode->customScore;
return score;
}
static void balanceInitDnodeList() {
tsBalanceDnodeList = calloc(tsBalanceDnodeListMallocSize, sizeof(SDnodeObj *));
}
static void balanceCleanupDnodeList() {
if (tsBalanceDnodeList != NULL) {
free(tsBalanceDnodeList);
tsBalanceDnodeList = NULL;
}
}
static void balanceCheckDnodeListSize(int32_t dnodesNum) {
if (tsBalanceDnodeListMallocSize <= dnodesNum) {
tsBalanceDnodeListMallocSize = dnodesNum * 2;
tsBalanceDnodeList = realloc(tsBalanceDnodeList, tsBalanceDnodeListMallocSize * sizeof(SDnodeObj *));
}
}
void balanceAccquireDnodeList() {
int32_t dnodesNum = mnodeGetDnodesNum();
balanceCheckDnodeListSize(dnodesNum);
void * pIter = NULL;
SDnodeObj *pDnode = NULL;
int32_t dnodeIndex = 0;
while (1) {
if (dnodeIndex >= dnodesNum) {
mnodeCancelGetNextDnode(pIter);
break;
}
pIter = mnodeGetNextDnode(pIter, &pDnode);
if (pDnode == NULL) break;
if (pDnode->status == TAOS_DN_STATUS_OFFLINE) {
mnodeDecDnodeRef(pDnode);
continue;
}
balanceCalcDnodeScore(pDnode);
int32_t orderIndex = dnodeIndex;
for (; orderIndex > 0; --orderIndex) {
if (pDnode->score > tsBalanceDnodeList[orderIndex - 1]->score) {
break;
}
tsBalanceDnodeList[orderIndex] = tsBalanceDnodeList[orderIndex - 1];
}
tsBalanceDnodeList[orderIndex] = pDnode;
dnodeIndex++;
}
tsBalanceDnodeListSize = dnodeIndex;
}
void balanceReleaseDnodeList() {
for (int32_t i = 0; i < tsBalanceDnodeListSize; ++i) {
SDnodeObj *pDnode = tsBalanceDnodeList[i];
if (pDnode != NULL) {
mnodeDecDnodeRef(pDnode);
}
}
}
static int32_t balanceGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn) {
SUserObj *pUser = mnodeGetUserFromConn(pConn);
if (pUser == NULL) return 0;
if (strcmp(pUser->pAcct->user, "root") != 0) {
mnodeDecUserRef(pUser);
return TSDB_CODE_MND_NO_RIGHTS;
}
int32_t cols = 0;
SSchema *pSchema = pMeta->schema;
pShow->bytes[cols] = 2;
pSchema[cols].type = TSDB_DATA_TYPE_SMALLINT;
strcpy(pSchema[cols].name, "id");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "system scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "custom scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "module scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "vnode scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "total scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_INT;
strcpy(pSchema[cols].name, "open vnodes");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_INT;
strcpy(pSchema[cols].name, "cpu cores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 18 + VARSTR_HEADER_SIZE;
pSchema[cols].type = TSDB_DATA_TYPE_BINARY;
strcpy(pSchema[cols].name, "balance state");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pMeta->numOfColumns = htons(cols);
pShow->numOfColumns = cols;
pShow->offset[0] = 0;
for (int32_t i = 1; i < cols; ++i) {
pShow->offset[i] = pShow->offset[i - 1] + pShow->bytes[i - 1];
}
pShow->numOfRows = mnodeGetDnodesNum();
pShow->rowSize = pShow->offset[cols - 1] + pShow->bytes[cols - 1];
pShow->pIter = NULL;
mnodeDecUserRef(pUser);
return 0;
}
static int32_t balanceRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn) {
int32_t numOfRows = 0;
SDnodeObj *pDnode = NULL;
char * pWrite;
int32_t cols = 0;
while (numOfRows < rows) {
pShow->pIter = mnodeGetNextDnode(pShow->pIter, &pDnode);
if (pDnode == NULL) break;
int32_t systemScore = balanceCalcCpuScore(pDnode) + balanceCalcMemoryScore(pDnode) + balanceCalcDiskScore(pDnode) +
balanceCalcBandwidthScore(pDnode);
float moduleScore = balanceCalcModuleScore(pDnode);
float vnodeScore = balanceCalcVnodeScore(pDnode, 0);
cols = 0;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int16_t *)pWrite = pDnode->dnodeId;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = systemScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = pDnode->customScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = (int32_t)moduleScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = (int32_t)vnodeScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = (int32_t)(vnodeScore + moduleScore + pDnode->customScore + systemScore);
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int32_t *)pWrite = pDnode->openVnodes;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int32_t *)pWrite = pDnode->numOfCores;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
STR_TO_VARSTR(pWrite, mnodeGetDnodeStatusStr(pDnode->status));
cols++;
numOfRows++;
mnodeDecDnodeRef(pDnode);
}
mnodeVacuumResult(data, pShow->numOfColumns, numOfRows, rows, pShow);
pShow->numOfReads += numOfRows;
return numOfRows;
}
static void balanceMonitorDnodeModule() {
static void bnMonitorDnodeModule() {
int32_t numOfMnodes = mnodeGetMnodesNum();
if (numOfMnodes >= tsNumOfMnodes) return;
for (int32_t i = 0; i < tsBalanceDnodeListSize; ++i) {
SDnodeObj *pDnode = tsBalanceDnodeList[i];
for (int32_t i = 0; i < tsBnDnodes.size; ++i) {
SDnodeObj *pDnode = tsBnDnodes.list[i];
if (pDnode == NULL) break;
if (pDnode->isMgmt || pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) {
@ -980,7 +636,7 @@ static void balanceMonitorDnodeModule() {
}
}
int32_t balanceAlterDnode(struct SDnodeObj *pSrcDnode, int32_t vnodeId, int32_t dnodeId) {
int32_t bnAlterDnode(struct SDnodeObj *pSrcDnode, int32_t vnodeId, int32_t dnodeId) {
if (!sdbIsMaster()) {
mError("dnode:%d, failed to alter vgId:%d to dnode:%d, for self not master", pSrcDnode->dnodeId, vnodeId, dnodeId);
return TSDB_CODE_MND_DNODE_NOT_EXIST;
@ -1004,29 +660,29 @@ int32_t balanceAlterDnode(struct SDnodeObj *pSrcDnode, int32_t vnodeId, int32_t
return TSDB_CODE_MND_DNODE_NOT_EXIST;
}
balanceLock();
balanceAccquireDnodeList();
bnLock();
bnAccquireDnodes();
int32_t code = TSDB_CODE_SUCCESS;
if (!balanceCheckDnodeInVgroup(pSrcDnode, pVgroup)) {
if (!bnCheckDnodeInVgroup(pSrcDnode, pVgroup)) {
mError("dnode:%d, failed to alter vgId:%d to dnode:%d, vgroup not in dnode:%d", pSrcDnode->dnodeId, vnodeId,
dnodeId, pSrcDnode->dnodeId);
code = TSDB_CODE_MND_VGROUP_NOT_IN_DNODE;
} else if (balanceCheckDnodeInVgroup(pDestDnode, pVgroup)) {
} else if (bnCheckDnodeInVgroup(pDestDnode, pVgroup)) {
mError("dnode:%d, failed to alter vgId:%d to dnode:%d, vgroup already in dnode:%d", pSrcDnode->dnodeId, vnodeId,
dnodeId, dnodeId);
code = TSDB_CODE_MND_VGROUP_ALREADY_IN_DNODE;
} else if (!balanceCheckFree(pDestDnode)) {
} else if (!bnCheckFree(pDestDnode)) {
mError("dnode:%d, failed to alter vgId:%d to dnode:%d, for dnode:%d not free", pSrcDnode->dnodeId, vnodeId, dnodeId,
dnodeId);
code = TSDB_CODE_MND_DNODE_NOT_FREE;
} else {
code = balanceAddVnode(pVgroup, pSrcDnode, pDestDnode);
code = bnAddVnode(pVgroup, pSrcDnode, pDestDnode);
mInfo("dnode:%d, alter vgId:%d to dnode:%d, result:%s", pSrcDnode->dnodeId, vnodeId, dnodeId, tstrerror(code));
}
balanceReleaseDnodeList();
balanceUnLock();
bnReleaseDnodes();
bnUnLock();
mnodeDecVgroupRef(pVgroup);
mnodeDecDnodeRef(pDestDnode);

312
src/balance/src/bnScore.c Normal file
View File

@ -0,0 +1,312 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "tglobal.h"
#include "mnodeShow.h"
#include "mnodeUser.h"
#include "bnScore.h"
SBnDnodes tsBnDnodes;
static int32_t bnGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn);
static int32_t bnRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn);
static int32_t bnCalcCpuScore(SDnodeObj *pDnode) {
if (pDnode->cpuAvgUsage < 80)
return 0;
else if (pDnode->cpuAvgUsage < 90)
return 10;
else
return 50;
}
static int32_t bnCalcMemoryScore(SDnodeObj *pDnode) {
if (pDnode->memoryAvgUsage < 80)
return 0;
else if (pDnode->memoryAvgUsage < 90)
return 10;
else
return 50;
}
static int32_t bnCalcDiskScore(SDnodeObj *pDnode) {
if (pDnode->diskAvgUsage < 80)
return 0;
else if (pDnode->diskAvgUsage < 90)
return 10;
else
return 50;
}
static int32_t bnCalcBandScore(SDnodeObj *pDnode) {
if (pDnode->bandwidthUsage < 30)
return 0;
else if (pDnode->bandwidthUsage < 80)
return 10;
else
return 50;
}
static float bnCalcModuleScore(SDnodeObj *pDnode) {
if (pDnode->numOfCores <= 0) return 0;
if (pDnode->isMgmt) {
return (float)tsMnodeEqualVnodeNum / pDnode->numOfCores;
}
return 0;
}
static float bnCalcVnodeScore(SDnodeObj *pDnode, int32_t extra) {
if (pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) return 100000000;
if (pDnode->numOfCores <= 0) return 0;
return (float)(pDnode->openVnodes + extra) / pDnode->numOfCores;
}
/**
* calc singe score, such as cpu/memory/disk/bandwitdh/vnode
* 1. get the score config
* 2. if the value is out of range, use border data
* 3. otherwise use interpolation method
**/
static void bnCalcDnodeScore(SDnodeObj *pDnode) {
pDnode->score = bnCalcCpuScore(pDnode) + bnCalcMemoryScore(pDnode) + bnCalcDiskScore(pDnode) +
bnCalcBandScore(pDnode) + bnCalcModuleScore(pDnode) + bnCalcVnodeScore(pDnode, 0) +
pDnode->customScore;
}
float bnTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extra) {
int32_t systemScore = bnCalcCpuScore(pDnode) + bnCalcMemoryScore(pDnode) + bnCalcDiskScore(pDnode) +
bnCalcBandScore(pDnode);
float moduleScore = bnCalcModuleScore(pDnode);
float vnodeScore = bnCalcVnodeScore(pDnode, extra);
float score = systemScore + moduleScore + vnodeScore + pDnode->customScore;
return score;
}
void bnInitDnodes() {
mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_SCORES, bnGetScoresMeta);
mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_SCORES, bnRetrieveScores);
mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_SCORES, mnodeCancelGetNextDnode);
memset(&tsBnDnodes, 0, sizeof(SBnDnodes));
tsBnDnodes.maxSize = 16;
tsBnDnodes.list = calloc(tsBnDnodes.maxSize, sizeof(SDnodeObj *));
}
void bnCleanupDnodes() {
if (tsBnDnodes.list != NULL) {
free(tsBnDnodes.list);
tsBnDnodes.list = NULL;
}
}
static void bnCheckDnodesSize(int32_t dnodesNum) {
if (tsBnDnodes.maxSize <= dnodesNum) {
tsBnDnodes.maxSize = dnodesNum * 2;
tsBnDnodes.list = realloc(tsBnDnodes.list, tsBnDnodes.maxSize * sizeof(SDnodeObj *));
}
}
void bnAccquireDnodes() {
int32_t dnodesNum = mnodeGetDnodesNum();
bnCheckDnodesSize(dnodesNum);
void * pIter = NULL;
SDnodeObj *pDnode = NULL;
int32_t dnodeIndex = 0;
while (1) {
if (dnodeIndex >= dnodesNum) {
mnodeCancelGetNextDnode(pIter);
break;
}
pIter = mnodeGetNextDnode(pIter, &pDnode);
if (pDnode == NULL) break;
if (pDnode->status == TAOS_DN_STATUS_OFFLINE) {
mnodeDecDnodeRef(pDnode);
continue;
}
bnCalcDnodeScore(pDnode);
int32_t orderIndex = dnodeIndex;
for (; orderIndex > 0; --orderIndex) {
if (pDnode->score > tsBnDnodes.list[orderIndex - 1]->score) {
break;
}
tsBnDnodes.list[orderIndex] = tsBnDnodes.list[orderIndex - 1];
}
tsBnDnodes.list[orderIndex] = pDnode;
dnodeIndex++;
}
tsBnDnodes.size = dnodeIndex;
}
void bnReleaseDnodes() {
for (int32_t i = 0; i < tsBnDnodes.size; ++i) {
SDnodeObj *pDnode = tsBnDnodes.list[i];
if (pDnode != NULL) {
mnodeDecDnodeRef(pDnode);
}
}
}
static int32_t bnGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn) {
SUserObj *pUser = mnodeGetUserFromConn(pConn);
if (pUser == NULL) return 0;
if (strcmp(pUser->pAcct->user, "root") != 0) {
mnodeDecUserRef(pUser);
return TSDB_CODE_MND_NO_RIGHTS;
}
int32_t cols = 0;
SSchema *pSchema = pMeta->schema;
pShow->bytes[cols] = 2;
pSchema[cols].type = TSDB_DATA_TYPE_SMALLINT;
strcpy(pSchema[cols].name, "id");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "system scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "custom scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "module scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "vnode scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_FLOAT;
strcpy(pSchema[cols].name, "total scores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_INT;
strcpy(pSchema[cols].name, "open vnodes");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 4;
pSchema[cols].type = TSDB_DATA_TYPE_INT;
strcpy(pSchema[cols].name, "cpu cores");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pShow->bytes[cols] = 18 + VARSTR_HEADER_SIZE;
pSchema[cols].type = TSDB_DATA_TYPE_BINARY;
strcpy(pSchema[cols].name, "balance state");
pSchema[cols].bytes = htons(pShow->bytes[cols]);
cols++;
pMeta->numOfColumns = htons(cols);
pShow->numOfColumns = cols;
pShow->offset[0] = 0;
for (int32_t i = 1; i < cols; ++i) {
pShow->offset[i] = pShow->offset[i - 1] + pShow->bytes[i - 1];
}
pShow->numOfRows = mnodeGetDnodesNum();
pShow->rowSize = pShow->offset[cols - 1] + pShow->bytes[cols - 1];
pShow->pIter = NULL;
mnodeDecUserRef(pUser);
return 0;
}
static int32_t bnRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn) {
int32_t numOfRows = 0;
SDnodeObj *pDnode = NULL;
char * pWrite;
int32_t cols = 0;
while (numOfRows < rows) {
pShow->pIter = mnodeGetNextDnode(pShow->pIter, &pDnode);
if (pDnode == NULL) break;
int32_t systemScore = bnCalcCpuScore(pDnode) + bnCalcMemoryScore(pDnode) + bnCalcDiskScore(pDnode) + bnCalcBandScore(pDnode);
float moduleScore = bnCalcModuleScore(pDnode);
float vnodeScore = bnCalcVnodeScore(pDnode, 0);
cols = 0;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int16_t *)pWrite = pDnode->dnodeId;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = systemScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = pDnode->customScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = (int32_t)moduleScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = (int32_t)vnodeScore;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(float *)pWrite = (int32_t)(vnodeScore + moduleScore + pDnode->customScore + systemScore);
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int32_t *)pWrite = pDnode->openVnodes;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int32_t *)pWrite = pDnode->numOfCores;
cols++;
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
STR_TO_VARSTR(pWrite, mnodeGetDnodeStatusStr(pDnode->status));
cols++;
numOfRows++;
mnodeDecDnodeRef(pDnode);
}
mnodeVacuumResult(data, pShow->numOfColumns, numOfRows, rows, pShow);
pShow->numOfReads += numOfRows;
return numOfRows;
}

132
src/balance/src/bnThread.c Normal file
View File

@ -0,0 +1,132 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "ttimer.h"
#include "tglobal.h"
#include "mnodeSdb.h"
#include "bnThread.h"
static SBnThread tsBnThread;
static void *bnThreadFunc(void *arg) {
while (1) {
pthread_mutex_lock(&tsBnThread.mutex);
if (tsBnThread.stop) {
pthread_mutex_unlock(&tsBnThread.mutex);
break;
}
pthread_cond_wait(&tsBnThread.cond, &tsBnThread.mutex);
bool updateSoon = bnStart();
bnStartTimer(updateSoon ? 1000 : -1);
pthread_mutex_unlock(&(tsBnThread.mutex));
}
mDebug("balance thread is stopped");
return NULL;
}
int32_t bnInitThread() {
memset(&tsBnThread, 0, sizeof(SBnThread));
tsBnThread.stop = false;
pthread_mutex_init(&tsBnThread.mutex, NULL);
pthread_cond_init(&tsBnThread.cond, NULL);
pthread_attr_t thattr;
pthread_attr_init(&thattr);
pthread_attr_setdetachstate(&thattr, PTHREAD_CREATE_JOINABLE);
int32_t ret = pthread_create(&tsBnThread.thread, &thattr, bnThreadFunc, NULL);
pthread_attr_destroy(&thattr);
if (ret != 0) {
mError("failed to create balance thread since %s", strerror(errno));
return -1;
}
bnStartTimer(2000);
mDebug("balance thread is created");
return 0;
}
void bnCleanupThread() {
mDebug("balance thread will be cleanup");
if (tsBnThread.timer != NULL) {
taosTmrStopA(&tsBnThread.timer);
tsBnThread.timer = NULL;
mDebug("stop balance timer");
}
pthread_mutex_lock(&tsBnThread.mutex);
tsBnThread.stop = true;
pthread_cond_signal(&tsBnThread.cond);
pthread_mutex_unlock(&(tsBnThread.mutex));
pthread_join(tsBnThread.thread, NULL);
pthread_cond_destroy(&tsBnThread.cond);
pthread_mutex_destroy(&tsBnThread.mutex);
}
static void bnPostSignal() {
if (tsBnThread.stop) return;
pthread_mutex_lock(&tsBnThread.mutex);
pthread_cond_signal(&tsBnThread.cond);
pthread_mutex_unlock(&(tsBnThread.mutex));
}
/*
* once sdb work as mater, then tsAccessSquence reset to zero
* increase tsAccessSquence every balance interval
*/
static void bnProcessTimer(void *handle, void *tmrId) {
if (!sdbIsMaster()) return;
if (tsBnThread.stop) return;
tsBnThread.timer = NULL;
tsAccessSquence++;
bnCheckStatus();
bnStartTimer(-1);
if (handle == NULL) {
if (tsAccessSquence % tsBalanceInterval == 0) {
mDebug("balance function is scheduled by timer");
bnPostSignal();
}
} else {
int64_t mseconds = (int64_t)handle;
mDebug("balance function is scheduled by event for %" PRId64 " mseconds arrived", mseconds);
bnPostSignal();
}
}
void bnStartTimer(int64_t mseconds) {
if (tsBnThread.stop) return;
bool updateSoon = (mseconds != -1);
if (updateSoon) {
taosTmrReset(bnProcessTimer, mseconds, (void *)mseconds, tsMnodeTmr, &tsBnThread.timer);
} else {
taosTmrReset(bnProcessTimer, tsStatusInterval * 1000, NULL, tsMnodeTmr, &tsBnThread.timer);
}
}
void bnNotify() {
bnStartTimer(500);
}

View File

@ -92,7 +92,7 @@ typedef struct SVgroupTableInfo {
} SVgroupTableInfo;
static FORCE_INLINE SQueryInfo* tscGetQueryInfoDetail(SSqlCmd* pCmd, int32_t subClauseIndex) {
assert(pCmd != NULL && subClauseIndex >= 0 && subClauseIndex < TSDB_MAX_UNION_CLAUSE);
assert(pCmd != NULL && subClauseIndex >= 0);
if (pCmd->pQueryInfo == NULL || subClauseIndex >= pCmd->numOfClause) {
return NULL;
@ -278,7 +278,7 @@ bool hasMoreClauseToTry(SSqlObj* pSql);
void tscTryQueryNextVnode(SSqlObj *pSql, __async_cb_func_t fp);
void tscAsyncQuerySingleRowForNextVnode(void *param, TAOS_RES *tres, int numOfRows);
void tscTryQueryNextClause(SSqlObj* pSql, __async_cb_func_t fp);
int tscSetMgmtEpSetFromCfg(const char *first, const char *second);
int tscSetMgmtEpSetFromCfg(const char *first, const char *second, SRpcCorEpSet *corEpSet);
bool tscSetSqlOwner(SSqlObj* pSql);
void tscClearSqlOwner(SSqlObj* pSql);

View File

@ -246,11 +246,14 @@ typedef struct SQueryInfo {
int16_t fillType; // final result fill type
int16_t numOfTables;
STableMetaInfo **pTableMetaInfo;
struct STSBuf * tsBuf;
struct STSBuf *tsBuf;
int64_t * fillVal; // default value for fill
char * msg; // pointer to the pCmd->payload to keep error message temporarily
int64_t clauseLimit; // limit for current sub clause
int64_t prjOffset; // offset value in the original sql expression, only applied at client side
int64_t tableLimit; // table limit in case of super table projection query + global order + limit
int32_t udColumnId; // current user-defined constant output field column id, monotonically decreases from TSDB_UD_COLUMN_INDEX
int16_t resColumnId; // result column id
} SQueryInfo;
@ -336,6 +339,7 @@ typedef struct STscObj {
int64_t hbrid;
struct SSqlObj * sqlList;
struct SSqlStream *streamList;
SRpcCorEpSet *tscCorMgmtEpSet;
void* pDnodeConn;
pthread_mutex_t mutex;
T_REF_DECLARE()
@ -378,6 +382,7 @@ typedef struct SSqlObj {
typedef struct SSqlStream {
SSqlObj *pSql;
const char* dstTable;
uint32_t streamId;
char listed;
bool isProject;
@ -404,6 +409,8 @@ typedef struct SSqlStream {
struct SSqlStream *prev, *next;
} SSqlStream;
void tscSetStreamDestTable(SSqlStream* pStream, const char* dstTable);
int32_t tscInitRpc(const char *user, const char *secret, void** pDnodeConn);
void tscInitMsgsFp();
@ -515,7 +522,6 @@ extern int tsInsertHeadSize;
extern int tscNumOfThreads;
extern int tscRefId;
extern SRpcCorEpSet tscMgmtEpSet;
extern int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo);

View File

@ -96,7 +96,7 @@ void taos_query_a(TAOS *taos, const char *sqlstr, __async_cb_func_t fp, void *pa
return;
}
taosNotePrintTsc(sqlstr);
nPrintTsc(sqlstr);
SSqlObj *pSql = (SSqlObj *)calloc(1, sizeof(SSqlObj));
if (pSql == NULL) {

View File

@ -54,8 +54,8 @@
#define DO_UPDATE_TAG_COLUMNS(ctx, ts) \
do { \
for (int32_t i = 0; i < (ctx)->tagInfo.numOfTagCols; ++i) { \
SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[i]; \
for (int32_t _i = 0; _i < (ctx)->tagInfo.numOfTagCols; ++_i) { \
SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[_i]; \
if (__ctx->functionId == TSDB_FUNC_TS_DUMMY) { \
__ctx->tag.i64Key = (ts); \
__ctx->tag.nType = TSDB_DATA_TYPE_BIGINT; \
@ -66,8 +66,8 @@
#define DO_UPDATE_TAG_COLUMNS_WITHOUT_TS(ctx) \
do { \
for (int32_t i = 0; i < (ctx)->tagInfo.numOfTagCols; ++i) { \
SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[i]; \
for (int32_t _i = 0; _i < (ctx)->tagInfo.numOfTagCols; ++_i) { \
SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[_i]; \
aAggs[TSDB_FUNC_TAG].xFunction(__ctx); \
} \
} while (0);
@ -305,7 +305,7 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI
} else if (functionId == TSDB_FUNC_FIRST || functionId == TSDB_FUNC_LAST) {
*type = (int16_t)dataType;
*bytes = (int16_t)dataBytes;
*interBytes = dataBytes;
*interBytes = (int16_t)(dataBytes + sizeof(SFirstLastInfo));
} else if (functionId == TSDB_FUNC_SPREAD) {
*type = (int16_t)TSDB_DATA_TYPE_DOUBLE;
*bytes = sizeof(double);
@ -426,8 +426,7 @@ static void count_function_f(SQLFunctionCtx *pCtx, int32_t index) {
}
SET_VAL(pCtx, 1, 1);
*((int64_t *)pCtx->aOutputBuf) += 1;
*((int64_t *)pCtx->aOutputBuf) += pCtx->size;
// do not need it actually
SResultRowCellInfo *pInfo = GET_RES_INFO(pCtx);
@ -1170,8 +1169,8 @@ static int32_t minmax_merge_impl(SQLFunctionCtx *pCtx, int32_t bytes, char *outp
if ((*(int32_t *)output < v) ^ isMin) {
*(int32_t *)output = v;
for (int32_t i = 0; i < pCtx->tagInfo.numOfTagCols; ++i) {
SQLFunctionCtx *__ctx = pCtx->tagInfo.pTagCtxList[i];
for (int32_t j = 0; j < pCtx->tagInfo.numOfTagCols; ++j) {
SQLFunctionCtx *__ctx = pCtx->tagInfo.pTagCtxList[j];
aAggs[TSDB_FUNC_TAG].xFunction(__ctx);
}
@ -1680,16 +1679,35 @@ static void last_function_f(SQLFunctionCtx *pCtx, int32_t index) {
if (pCtx->hasNull && isNull(pData, pCtx->inputType)) {
return;
}
SET_VAL(pCtx, 1, 1);
memcpy(pCtx->aOutputBuf, pData, pCtx->inputBytes);
TSKEY ts = pCtx->ptsList[index];
DO_UPDATE_TAG_COLUMNS(pCtx, ts);
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
pResInfo->hasResult = DATA_SET_FLAG;
pResInfo->complete = true; // set query completed
// the scan order is not the required order, ignore it
if (pCtx->order != pCtx->param[0].i64Key) {
return;
}
if (pCtx->order == TSDB_ORDER_DESC) {
SET_VAL(pCtx, 1, 1);
memcpy(pCtx->aOutputBuf, pData, pCtx->inputBytes);
TSKEY ts = pCtx->ptsList[index];
DO_UPDATE_TAG_COLUMNS(pCtx, ts);
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
pResInfo->hasResult = DATA_SET_FLAG;
pResInfo->complete = true; // set query completed
} else { // in case of ascending order check, all data needs to be checked
SResultRowCellInfo* pResInfo = GET_RES_INFO(pCtx);
TSKEY ts = pCtx->ptsList[index];
char* buf = GET_ROWCELL_INTERBUF(pResInfo);
if (pResInfo->hasResult != DATA_SET_FLAG || (*(TSKEY*)buf) < ts) {
pResInfo->hasResult = DATA_SET_FLAG;
memcpy(pCtx->aOutputBuf, pData, pCtx->inputBytes);
*(TSKEY*)buf = ts;
DO_UPDATE_TAG_COLUMNS(pCtx, ts);
}
}
}
static void last_data_assign_impl(SQLFunctionCtx *pCtx, char *pData, int32_t index) {
@ -1712,7 +1730,7 @@ static void last_data_assign_impl(SQLFunctionCtx *pCtx, char *pData, int32_t ind
static void last_dist_function(SQLFunctionCtx *pCtx) {
/*
* 1. for scan data in asc order, no need to check data
* 1. for scan data is not the required order
* 2. for data blocks that are not loaded, no need to check data
*/
if (pCtx->order != pCtx->param[0].i64Key) {
@ -2448,7 +2466,7 @@ static bool percentile_function_setup(SQLFunctionCtx *pCtx) {
static void percentile_function(SQLFunctionCtx *pCtx) {
int32_t notNullElems = 0;
SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx);
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
SPercentileInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo);
// the first stage, only acquire the min/max value
@ -2549,12 +2567,14 @@ static void percentile_finalizer(SQLFunctionCtx *pCtx) {
double v = pCtx->param[0].nType == TSDB_DATA_TYPE_INT ? pCtx->param[0].i64Key : pCtx->param[0].dKey;
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
tMemBucket * pMemBucket = ((SPercentileInfo *)GET_ROWCELL_INTERBUF(pResInfo))->pMemBucket;
if (pMemBucket->total > 0) { // check for null
*(double *)pCtx->aOutputBuf = getPercentile(pMemBucket, v);
} else {
SPercentileInfo* ppInfo = (SPercentileInfo *) GET_ROWCELL_INTERBUF(pResInfo);
tMemBucket * pMemBucket = ppInfo->pMemBucket;
if (pMemBucket == NULL || pMemBucket->total == 0) { // check for null
assert(ppInfo->numOfElems == 0);
setNull(pCtx->aOutputBuf, pCtx->outputType, pCtx->outputBytes);
} else {
*(double *)pCtx->aOutputBuf = getPercentile(pMemBucket, v);
}
tMemBucketDestroy(pMemBucket);
@ -3632,114 +3652,119 @@ static bool twa_function_setup(SQLFunctionCtx *pCtx) {
return true;
}
static int32_t twa_function_impl(SQLFunctionCtx* pCtx, int32_t index, int32_t size) {
static int32_t twa_function_impl(SQLFunctionCtx* pCtx, int32_t tsIndex, int32_t index, int32_t size) {
int32_t notNullElems = 0;
TSKEY *primaryKey = pCtx->ptsList;
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
STwaInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo);
int32_t i = index;
int32_t step = GET_FORWARD_DIRECTION_FACTOR(pCtx->order);
if (pCtx->start.key != INT64_MIN) {
assert(pCtx->start.key < primaryKey[index] && pInfo->lastKey == INT64_MIN);
assert((pCtx->start.key < primaryKey[tsIndex + i] && pCtx->order == TSDB_ORDER_ASC) ||
(pCtx->start.key > primaryKey[tsIndex + i] && pCtx->order == TSDB_ORDER_DESC));
pInfo->lastKey = primaryKey[index];
GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, 0));
assert(pInfo->lastKey == INT64_MIN);
pInfo->lastKey = primaryKey[tsIndex + i];
GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index));
pInfo->dOutput += ((pInfo->lastValue + pCtx->start.val) / 2) * (pInfo->lastKey - pCtx->start.key);
pInfo->hasResult = DATA_SET_FLAG;
pInfo->win.skey = pCtx->start.key;
pInfo->win.skey = pCtx->start.key;
notNullElems++;
i += 1;
i += step;
} else if (pInfo->lastKey == INT64_MIN) {
pInfo->lastKey = primaryKey[index];
GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, 0));
pInfo->lastKey = primaryKey[tsIndex + i];
GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index));
pInfo->hasResult = DATA_SET_FLAG;
pInfo->win.skey = pInfo->lastKey;
pInfo->win.skey = pInfo->lastKey;
notNullElems++;
i += 1;
i += step;
}
// calculate the value of
switch(pCtx->inputType) {
case TSDB_DATA_TYPE_TINYINT: {
int8_t *val = (int8_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
int8_t *val = (int8_t*) GET_INPUT_CHAR_INDEX(pCtx, 0);
for (; i < size && i >= 0; i += step) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i] - pInfo->lastKey);
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i];
pInfo->lastKey = primaryKey[i + tsIndex];
}
break;
}
case TSDB_DATA_TYPE_SMALLINT: {
int16_t *val = (int16_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
int16_t *val = (int16_t*) GET_INPUT_CHAR_INDEX(pCtx, 0);
for (; i < size && i >= 0; i += step) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i] - pInfo->lastKey);
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i];
pInfo->lastKey = primaryKey[i + tsIndex];
}
break;
}
case TSDB_DATA_TYPE_INT: {
int32_t *val = (int32_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
int32_t *val = (int32_t*) GET_INPUT_CHAR_INDEX(pCtx, 0);
for (; i < size && i >= 0; i += step) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i] - pInfo->lastKey);
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i];
pInfo->lastKey = primaryKey[i + tsIndex];
}
break;
}
case TSDB_DATA_TYPE_BIGINT: {
int64_t *val = (int64_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
int64_t *val = (int64_t*) GET_INPUT_CHAR_INDEX(pCtx, 0);
for (; i < size && i >= 0; i += step) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i] - pInfo->lastKey);
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey);
pInfo->lastValue = (double) val[i];
pInfo->lastKey = primaryKey[i];
pInfo->lastKey = primaryKey[i + tsIndex];
}
break;
}
case TSDB_DATA_TYPE_FLOAT: {
float *val = (float*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
float *val = (float*) GET_INPUT_CHAR_INDEX(pCtx, 0);
for (; i < size && i >= 0; i += step) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i] - pInfo->lastKey);
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i];
pInfo->lastKey = primaryKey[i + tsIndex];
}
break;
}
case TSDB_DATA_TYPE_DOUBLE: {
double *val = (double*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
double *val = (double*) GET_INPUT_CHAR_INDEX(pCtx, 0);
for (; i < size && i >= 0; i += step) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i] - pInfo->lastKey);
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i];
pInfo->lastKey = primaryKey[i + tsIndex];
}
break;
}
@ -3764,16 +3789,13 @@ static void twa_function(SQLFunctionCtx *pCtx) {
STwaInfo * pInfo = GET_ROWCELL_INTERBUF(pResInfo);
// skip null value
int32_t i = 0;
int32_t step = GET_FORWARD_DIRECTION_FACTOR(pCtx->order);
int32_t i = (pCtx->order == TSDB_ORDER_ASC)? 0:(pCtx->size - 1);
while (pCtx->hasNull && i < pCtx->size && isNull((char *)data + pCtx->inputBytes * i, pCtx->inputType)) {
i++;
i += step;
}
if (i >= pCtx->size) {
return;
}
int32_t notNullElems = twa_function_impl(pCtx, pCtx->startOffset, pCtx->size);
int32_t notNullElems = twa_function_impl(pCtx, pCtx->startOffset, i, pCtx->size);
SET_VAL(pCtx, notNullElems, 1);
if (notNullElems > 0) {
@ -3791,11 +3813,136 @@ static void twa_function_f(SQLFunctionCtx *pCtx, int32_t index) {
return;
}
int32_t notNullElems = twa_function_impl(pCtx, index, 1);
int32_t notNullElems = 0;
TSKEY *primaryKey = pCtx->ptsList;
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
STwaInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo);
int32_t i = pCtx->startOffset;
int32_t size = pCtx->size;
if (pCtx->start.key != INT64_MIN) {
assert(pInfo->lastKey == INT64_MIN);
pInfo->lastKey = primaryKey[index];
GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index));
pInfo->dOutput += ((pInfo->lastValue + pCtx->start.val) / 2) * (pInfo->lastKey - pCtx->start.key);
pInfo->hasResult = DATA_SET_FLAG;
pInfo->win.skey = pCtx->start.key;
notNullElems++;
i += 1;
} else if (pInfo->lastKey == INT64_MIN) {
pInfo->lastKey = primaryKey[index];
GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index));
pInfo->hasResult = DATA_SET_FLAG;
pInfo->win.skey = pInfo->lastKey;
notNullElems++;
i += 1;
}
// calculate the value of
switch(pCtx->inputType) {
case TSDB_DATA_TYPE_TINYINT: {
int8_t *val = (int8_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i + index];
}
break;
}
case TSDB_DATA_TYPE_SMALLINT: {
int16_t *val = (int16_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i + index];
}
break;
}
case TSDB_DATA_TYPE_INT: {
int32_t *val = (int32_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i + index];
}
break;
}
case TSDB_DATA_TYPE_BIGINT: {
int64_t *val = (int64_t*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey);
pInfo->lastValue = (double) val[i];
pInfo->lastKey = primaryKey[i + index];
}
break;
}
case TSDB_DATA_TYPE_FLOAT: {
float *val = (float*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i + index];
}
break;
}
case TSDB_DATA_TYPE_DOUBLE: {
double *val = (double*) GET_INPUT_CHAR_INDEX(pCtx, index);
for (; i < size; i++) {
if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) {
continue;
}
pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey);
pInfo->lastValue = val[i];
pInfo->lastKey = primaryKey[i + index];
}
break;
}
default: assert(0);
}
// the last interpolated time window value
if (pCtx->end.key != INT64_MIN) {
pInfo->dOutput += ((pInfo->lastValue + pCtx->end.val) / 2) * (pCtx->end.key - pInfo->lastKey);
pInfo->lastValue = pCtx->end.val;
pInfo->lastKey = pCtx->end.key;
}
pInfo->win.ekey = pInfo->lastKey;
SET_VAL(pCtx, notNullElems, 1);
if (notNullElems > 0) {
pResInfo->hasResult = DATA_SET_FLAG;
}
if (pCtx->stableQuery) {
SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx);
memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(STwaInfo));
}
}
@ -4414,11 +4561,11 @@ static void sumrate_finalizer(SQLFunctionCtx *pCtx) {
*
*/
int32_t functionCompatList[] = {
// count, sum, avg, min, max, stddev, percentile, apercentile, first, last
1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
// last_row, top, bottom, spread, twa, leastsqr, ts, ts_dummy, tag_dummy, ts_z
// count, sum, avg, min, max, stddev, percentile, apercentile, first, last
1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
// last_row,top, bottom, spread, twa, leastsqr, ts, ts_dummy, tag_dummy, ts_z
4, -1, -1, 1, 1, 1, 1, 1, 1, -1,
// tag, colprj, tagprj, arithmetic, diff, first_dist, last_dist, interp rate irate
// tag, colprj, tagprj, arithmetic, diff, first_dist, last_dist, interp rate irate
1, 1, 1, 1, -1, 1, 1, 5, 1, 1,
// sum_rate, sum_irate, avg_rate, avg_irate
1, 1, 1, 1,

View File

@ -721,10 +721,16 @@ int32_t tscLocalReducerEnvCreate(SSqlObj *pSql, tExtMemBuffer ***pMemBuffer, tOr
// final result depends on the fields number
memset(pSchema, 0, sizeof(SSchema) * size);
for (int32_t i = 0; i < size; ++i) {
SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, i);
SSchema *p1 = tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, pExpr->colInfo.colIndex);
SSchema p1 = {0};
if (pExpr->colInfo.colIndex != TSDB_TBNAME_COLUMN_INDEX) {
p1 = *tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, pExpr->colInfo.colIndex);
} else {
p1 = tGetTableNameColumnSchema();
}
int32_t inter = 0;
int16_t type = -1;
@ -743,7 +749,8 @@ int32_t tscLocalReducerEnvCreate(SSqlObj *pSql, tExtMemBuffer ***pMemBuffer, tOr
functionId = TSDB_FUNC_LAST;
}
getResultDataInfo(p1->type, p1->bytes, functionId, 0, &type, &bytes, &inter, 0, false);
int32_t ret = getResultDataInfo(p1.type, p1.bytes, functionId, 0, &type, &bytes, &inter, 0, false);
assert(ret == TSDB_CODE_SUCCESS);
}
pSchema[i].type = (uint8_t)type;
@ -1626,7 +1633,7 @@ void tscInitResObjForLocalQuery(SSqlObj *pObj, int32_t numOfRes, int32_t rowLen)
int32_t doArithmeticCalculate(SQueryInfo* pQueryInfo, tFilePage* pOutput, int32_t rowSize, int32_t finalRowSize) {
int32_t maxRowSize = MAX(rowSize, finalRowSize);
char* pbuf = calloc(1, pOutput->num * maxRowSize);
char* pbuf = calloc(1, (size_t)(pOutput->num * maxRowSize));
size_t size = tscNumOfFields(pQueryInfo);
SArithmeticSupport arithSup = {0};
@ -1653,16 +1660,16 @@ int32_t doArithmeticCalculate(SQueryInfo* pQueryInfo, tFilePage* pOutput, int32_
tExprTreeCalcTraverse(arithSup.pArithExpr->pExpr, (int32_t) pOutput->num, pbuf + pOutput->num*offset, &arithSup, TSDB_ORDER_ASC, getArithmeticInputSrc);
} else {
SSqlExpr* pExpr = pSup->pSqlExpr;
memcpy(pbuf + pOutput->num * offset, pExpr->offset * pOutput->num + pOutput->data, pExpr->resBytes * pOutput->num);
memcpy(pbuf + pOutput->num * offset, pExpr->offset * pOutput->num + pOutput->data, (size_t)(pExpr->resBytes * pOutput->num));
}
offset += pSup->field.bytes;
}
memcpy(pOutput->data, pbuf, pOutput->num * offset);
memcpy(pOutput->data, pbuf, (size_t)(pOutput->num * offset));
tfree(pbuf);
tfree(arithSup.data);
return offset;
}
}

View File

@ -630,11 +630,17 @@ int32_t tscAllocateMemIfNeed(STableDataBlocks *pDataBlock, int32_t rowSize, int3
return TSDB_CODE_SUCCESS;
}
static void tsSetBlockInfo(SSubmitBlk *pBlocks, const STableMeta *pTableMeta, int32_t numOfRows) {
static int32_t tsSetBlockInfo(SSubmitBlk *pBlocks, const STableMeta *pTableMeta, int32_t numOfRows) {
pBlocks->tid = pTableMeta->id.tid;
pBlocks->uid = pTableMeta->id.uid;
pBlocks->sversion = pTableMeta->sversion;
pBlocks->numOfRows += numOfRows;
if (pBlocks->numOfRows + numOfRows >= INT16_MAX) {
return TSDB_CODE_TSC_INVALID_SQL;
} else {
pBlocks->numOfRows += numOfRows;
return TSDB_CODE_SUCCESS;
}
}
// data block is disordered, sort it in ascending order
@ -722,7 +728,11 @@ static int32_t doParseInsertStatement(SSqlObj *pSql, void *pTableList, char **st
}
SSubmitBlk *pBlocks = (SSubmitBlk *)(dataBuf->pData);
tsSetBlockInfo(pBlocks, pTableMeta, numOfRows);
code = tsSetBlockInfo(pBlocks, pTableMeta, numOfRows);
if (code != TSDB_CODE_SUCCESS) {
tscInvalidSQLErrMsg(pCmd->payload, "too many rows in sql, total number of rows should be less than 32767", *str);
return code;
}
dataBuf->vgId = pTableMeta->vgroupInfo.vgId;
dataBuf->numOfTables = 1;
@ -1384,7 +1394,10 @@ static int doPackSendDataBlock(SSqlObj *pSql, int32_t numOfRows, STableDataBlock
STableMeta *pTableMeta = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0)->pTableMeta;
SSubmitBlk *pBlocks = (SSubmitBlk *)(pTableDataBlocks->pData);
tsSetBlockInfo(pBlocks, pTableMeta, numOfRows);
code = tsSetBlockInfo(pBlocks, pTableMeta, numOfRows);
if (code != TSDB_CODE_SUCCESS) {
return tscInvalidSQLErrMsg(pCmd->payload, "too many rows in sql, total number of rows should be less than 32767", NULL);
}
if ((code = tscMergeTableDataBlocks(pSql, pCmd->pDataBlocks)) != TSDB_CODE_SUCCESS) {
return code;

View File

@ -39,6 +39,7 @@ void tscInitConnCb(void *param, TAOS_RES *result, int code) {
tscSlowQueryConnInitialized = true;
tscSaveSlowQueryFp(sql, NULL);
}
taos_free_result(result);
}
void tscAddIntoSqlList(SSqlObj *pSql) {
@ -69,6 +70,7 @@ void tscSaveSlowQueryFpCb(void *param, TAOS_RES *result, int code) {
} else {
tscDebug("success to save slow query, code:%d", code);
}
taos_free_result(result);
}
void tscSaveSlowQueryFp(void *handle, void *tmrId) {
@ -260,6 +262,11 @@ int tscBuildQueryStreamDesc(void *pMsg, STscObj *pObj) {
SSqlStream *pStream = pObj->streamList;
while (pStream) {
tstrncpy(pSdesc->sql, pStream->pSql->sqlstr, sizeof(pSdesc->sql));
if (pStream->dstTable == NULL) {
pSdesc->dstTable[0] = 0;
} else {
tstrncpy(pSdesc->dstTable, pStream->dstTable, sizeof(pSdesc->dstTable));
}
pSdesc->streamId = htonl(pStream->streamId);
pSdesc->num = htobe64(pStream->num);

View File

@ -2835,6 +2835,10 @@ static bool functionCompatibleCheck(SQueryInfo* pQueryInfo, bool joinQuery) {
if (functionCompatList[functionId] != factor) {
return false;
} else {
if (factor == -1) { // two functions with the same -1 flag
return false;
}
}
if (functionId == TSDB_FUNC_LAST_ROW && joinQuery) {
@ -3278,7 +3282,12 @@ static int32_t extractColumnFilterInfo(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SC
((pSchema->type == TSDB_DATA_TYPE_BINARY || pSchema->type == TSDB_DATA_TYPE_NCHAR) ? 1 : 0);
if (pColFilter->filterstr) {
if (pExpr->nSQLOptr != TK_EQ && pExpr->nSQLOptr != TK_NE && pExpr->nSQLOptr != TK_LIKE) {
if (pExpr->nSQLOptr != TK_EQ
&& pExpr->nSQLOptr != TK_NE
&& pExpr->nSQLOptr != TK_ISNULL
&& pExpr->nSQLOptr != TK_NOTNULL
&& pExpr->nSQLOptr != TK_LIKE
) {
return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg2);
}
} else {
@ -5103,7 +5112,7 @@ int32_t validateDNodeConfig(tDCLSQL* pOptions) {
const int tokenDebugFlagEnd = 20;
const SDNodeDynConfOption cfgOptions[] = {
{"resetLog", 8}, {"resetQueryCache", 15}, {"balance", 7}, {"monitor", 7},
{"debugFlag", 9}, {"monitorDebugFlag", 16}, {"vDebugFlag", 10}, {"mDebugFlag", 10},
{"debugFlag", 9}, {"monDebugFlag", 12}, {"vDebugFlag", 10}, {"mDebugFlag", 10},
{"cDebugFlag", 10}, {"httpDebugFlag", 13}, {"qDebugflag", 10}, {"sdbDebugFlag", 12},
{"uDebugFlag", 10}, {"tsdbDebugFlag", 13}, {"sDebugflag", 10}, {"rpcDebugFlag", 12},
{"dDebugFlag", 10}, {"mqttDebugFlag", 13}, {"wDebugFlag", 10}, {"tmrDebugFlag", 12},
@ -5307,15 +5316,18 @@ int32_t parseLimitClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t clauseIn
// keep original limitation value in globalLimit
pQueryInfo->clauseLimit = pQueryInfo->limit.limit;
pQueryInfo->prjOffset = pQueryInfo->limit.offset;
pQueryInfo->prjOffset = pQueryInfo->limit.offset;
pQueryInfo->tableLimit = -1;
if (tscOrderedProjectionQueryOnSTable(pQueryInfo, 0)) {
/*
* the limitation/offset value should be removed during retrieve data from virtual node,
* since the global order are done in client side, so the limitation should also
* be done at the client side.
* the offset value should be removed during retrieve data from virtual node, since the
* global order are done in client side, so the offset is applied at the client side
* However, note that the maximum allowed number of result for each table should be less
* than or equal to the value of limit.
*/
if (pQueryInfo->limit.limit > 0) {
pQueryInfo->tableLimit = pQueryInfo->limit.limit + pQueryInfo->limit.offset;
pQueryInfo->limit.limit = -1;
}

View File

@ -26,7 +26,7 @@
#include "ttimer.h"
#include "tlockfree.h"
SRpcCorEpSet tscMgmtEpSet;
///SRpcCorEpSet tscMgmtEpSet;
int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo) = {0};
@ -73,10 +73,11 @@ static void tscSetDnodeEpSet(SSqlObj* pSql, SVgroupInfo* pVgroupInfo) {
assert(hasFqdn);
}
static void tscDumpMgmtEpSet(SRpcEpSet *epSet) {
taosCorBeginRead(&tscMgmtEpSet.version);
*epSet = tscMgmtEpSet.epSet;
taosCorEndRead(&tscMgmtEpSet.version);
static void tscDumpMgmtEpSet(SSqlObj *pSql) {
SRpcCorEpSet *pCorEpSet = pSql->pTscObj->tscCorMgmtEpSet;
taosCorBeginRead(&pCorEpSet->version);
pSql->epSet = pCorEpSet->epSet;
taosCorEndRead(&pCorEpSet->version);
}
static void tscEpSetHtons(SRpcEpSet *s) {
for (int32_t i = 0; i < s->numOfEps; i++) {
@ -94,11 +95,12 @@ bool tscEpSetIsEqual(SRpcEpSet *s1, SRpcEpSet *s2) {
}
return true;
}
void tscUpdateMgmtEpSet(SRpcEpSet *pEpSet) {
void tscUpdateMgmtEpSet(SSqlObj *pSql, SRpcEpSet *pEpSet) {
// no need to update if equal
taosCorBeginWrite(&tscMgmtEpSet.version);
tscMgmtEpSet.epSet = *pEpSet;
taosCorEndWrite(&tscMgmtEpSet.version);
SRpcCorEpSet *pCorEpSet = pSql->pTscObj->tscCorMgmtEpSet;
taosCorBeginWrite(&pCorEpSet->version);
pCorEpSet->epSet = *pEpSet;
taosCorEndWrite(&pCorEpSet->version);
}
static void tscDumpEpSetFromVgroupInfo(SCorVgroupInfo *pVgroupInfo, SRpcEpSet *pEpSet) {
if (pVgroupInfo == NULL) { return;}
@ -133,18 +135,6 @@ static void tscUpdateVgroupInfo(SSqlObj *pObj, SRpcEpSet *pEpSet) {
taosCorEndWrite(&pVgroupInfo->version);
}
void tscPrintMgmtEp() {
SRpcEpSet dump;
tscDumpMgmtEpSet(&dump);
if (dump.numOfEps <= 0) {
tscError("invalid mnode EP list:%d", dump.numOfEps);
} else {
for (int i = 0; i < dump.numOfEps; ++i) {
tscDebug("mnode index:%d %s:%d", i, dump.fqdn[i], dump.port[i]);
}
}
}
void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) {
STscObj *pObj = (STscObj *)param;
if (pObj == NULL) return;
@ -162,7 +152,7 @@ void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) {
SRpcEpSet * epSet = &pRsp->epSet;
if (epSet->numOfEps > 0) {
tscEpSetHtons(epSet);
tscUpdateMgmtEpSet(epSet);
tscUpdateMgmtEpSet(pSql, epSet);
}
pSql->pTscObj->connId = htonl(pRsp->connId);
@ -191,9 +181,16 @@ void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) {
void tscProcessActivityTimer(void *handle, void *tmrId) {
int64_t rid = (int64_t) handle;
STscObj *pObj = taosAcquireRef(tscRefId, rid);
if (pObj == NULL) return;
if (pObj == NULL) {
return;
}
SSqlObj* pHB = taosAcquireRef(tscObjRef, pObj->hbrid);
if (pHB == NULL) {
taosReleaseRef(tscRefId, rid);
return;
}
assert(pHB->self == pObj->hbrid);
pHB->retry = 0;
@ -219,7 +216,7 @@ int tscSendMsgToServer(SSqlObj *pSql) {
// set the mgmt ip list
if (pSql->cmd.command >= TSDB_SQL_MGMT) {
tscDumpMgmtEpSet(&pSql->epSet);
tscDumpMgmtEpSet(pSql);
}
memcpy(pMsg, pSql->cmd.payload, pSql->cmd.payloadLen);
@ -277,7 +274,7 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) {
if (pCmd->command < TSDB_SQL_MGMT) {
tscUpdateVgroupInfo(pSql, pEpSet);
} else {
tscUpdateMgmtEpSet(pEpSet);
tscUpdateMgmtEpSet(pSql, pEpSet);
}
}
}
@ -684,6 +681,7 @@ int tscBuildQueryMsg(SSqlObj *pSql, SSqlInfo *pInfo) {
pQueryMsg->tagNameRelType = htons(pQueryInfo->tagCond.relType);
pQueryMsg->numOfTags = htonl(numOfTags);
pQueryMsg->queryType = htonl(pQueryInfo->type);
pQueryMsg->tableLimit = htobe64(pQueryInfo->tableLimit);
size_t numOfOutput = tscSqlExprNumOfExprs(pQueryInfo);
pQueryMsg->numOfOutput = htons((int16_t)numOfOutput); // this is the stage one output column number
@ -2057,7 +2055,7 @@ int tscProcessConnectRsp(SSqlObj *pSql) {
if (pConnect->epSet.numOfEps > 0) {
tscEpSetHtons(&pConnect->epSet);
tscUpdateMgmtEpSet(&pConnect->epSet);
tscUpdateMgmtEpSet(pSql, &pConnect->epSet);
}
strcpy(pObj->sversion, pConnect->serverVersion);

View File

@ -58,6 +58,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa
terrno = TSDB_CODE_TSC_INVALID_USER_LENGTH;
return NULL;
}
SRpcCorEpSet corMgmtEpSet;
char secretEncrypt[32] = {0};
int secretEncryptLen = 0;
@ -82,11 +83,13 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa
}
secretEncryptLen = outlen;
}
if (ip) {
if (tscSetMgmtEpSetFromCfg(ip, NULL) < 0) return NULL;
if (port) tscMgmtEpSet.epSet.port[0] = port;
}
if (tscSetMgmtEpSetFromCfg(ip, NULL, &corMgmtEpSet) < 0) return NULL;
if (port) corMgmtEpSet.epSet.port[0] = port;
} else {
if (tscSetMgmtEpSetFromCfg(tsFirst, tsSecond, &corMgmtEpSet) < 0) return NULL;
}
void *pDnodeConn = NULL;
if (tscInitRpc(user, secretEncrypt, &pDnodeConn) != 0) {
@ -100,10 +103,20 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa
rpcClose(pDnodeConn);
return NULL;
}
// set up tscObj's mgmtEpSet
pObj->tscCorMgmtEpSet = (SRpcCorEpSet *)malloc(sizeof(SRpcCorEpSet));
if (NULL == pObj->tscCorMgmtEpSet) {
terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
rpcClose(pDnodeConn);
free(pObj->tscCorMgmtEpSet);
free(pObj);
}
memcpy(pObj->tscCorMgmtEpSet, &corMgmtEpSet, sizeof(SRpcCorEpSet));
pObj->signature = pObj;
pObj->pDnodeConn = pDnodeConn;
T_REF_INIT_VAL(pObj, 1);
tstrncpy(pObj->user, user, sizeof(pObj->user));
secretEncryptLen = MIN(secretEncryptLen, sizeof(pObj->pass));
@ -115,6 +128,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa
if (len >= TSDB_DB_NAME_LEN) {
terrno = TSDB_CODE_TSC_INVALID_DB_LENGTH;
rpcClose(pDnodeConn);
free(pObj->tscCorMgmtEpSet);
free(pObj);
return NULL;
}
@ -132,6 +146,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa
if (NULL == pSql) {
terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
rpcClose(pDnodeConn);
free(pObj->tscCorMgmtEpSet);
free(pObj);
return NULL;
}
@ -149,6 +164,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa
terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
rpcClose(pDnodeConn);
free(pSql);
free(pObj->tscCorMgmtEpSet);
free(pObj);
return NULL;
}
@ -276,16 +292,18 @@ void taos_close(TAOS *taos) {
pObj->signature = NULL;
taosTmrStopA(&(pObj->pTimer));
SSqlObj* pHb = (SSqlObj*)taosAcquireRef(tscObjRef, pObj->hbrid);
if (pHb != NULL) {
if (pHb->rpcRid > 0) { // wait for rsp from dnode
rpcCancelRequest(pHb->rpcRid);
pHb->rpcRid = -1;
}
if (pObj->hbrid > 0) {
SSqlObj* pHb = (SSqlObj*)taosAcquireRef(tscObjRef, pObj->hbrid);
if (pHb != NULL) {
if (pHb->rpcRid > 0) { // wait for rsp from dnode
rpcCancelRequest(pHb->rpcRid);
pHb->rpcRid = -1;
}
tscDebug("%p HB is freed", pHb);
taos_free_result(pHb);
taosReleaseRef(tscObjRef, pHb->self);
tscDebug("%p HB is freed", pHb);
taos_free_result(pHb);
taosReleaseRef(tscObjRef, pHb->self);
}
}
int32_t ref = T_REF_DEC(pObj);
@ -326,7 +344,7 @@ TAOS_RES* taos_query_c(TAOS *taos, const char *sqlstr, uint32_t sqlLen, TAOS_RES
return NULL;
}
taosNotePrintTsc(sqlstr);
nPrintTsc(sqlstr);
SSqlObj* pSql = calloc(1, sizeof(SSqlObj));
if (pSql == NULL) {

View File

@ -535,6 +535,10 @@ static void tscCreateStream(void *param, TAOS_RES *res, int code) {
pStream, pTableMetaInfo->name, pStream->interval.interval, pStream->interval.sliding, starttime, pSql->sqlstr);
}
void tscSetStreamDestTable(SSqlStream* pStream, const char* dstTable) {
pStream->dstTable = dstTable;
}
TAOS_STREAM *taos_open_stream(TAOS *taos, const char *sqlstr, void (*fp)(void *param, TAOS_RES *, TAOS_ROW row),
int64_t stime, void *param, void (*callback)(void *)) {
STscObj *pObj = (STscObj *)taos;

View File

@ -157,7 +157,7 @@ static SSub* tscCreateSubscription(STscObj* pObj, const char* topic, const char*
registerSqlObj(pSql);
code = tsParseSql(pSql, false);
code = tsParseSql(pSql, true);
if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) {
tsem_wait(&pSub->sem);
code = pSql->res.code;
@ -168,7 +168,7 @@ static SSub* tscCreateSubscription(STscObj* pObj, const char* topic, const char*
goto fail;
}
if (pSql->cmd.command != TSDB_SQL_SELECT) {
if (pSql->cmd.command != TSDB_SQL_SELECT && pSql->cmd.command != TSDB_SQL_RETRIEVE_EMPTY_RESULT) {
line = __LINE__;
code = TSDB_CODE_TSC_INVALID_SQL;
goto fail;
@ -182,7 +182,7 @@ fail:
if (pSql->self != 0) {
taosReleaseRef(tscObjRef, pSql->self);
} else {
tscFreeSqlObj(pSql);
tscFreeSqlObj(pSql);
}
pSql = NULL;
@ -401,9 +401,11 @@ TAOS_SUB *taos_subscribe(TAOS *taos, int restart, const char* topic, const char
tscLoadSubscriptionProgress(pSub);
}
if (!tscUpdateSubscription(pObj, pSub)) {
taos_unsubscribe(pSub, 1);
return NULL;
if (pSub->pSql->cmd.command == TSDB_SQL_SELECT) {
if (!tscUpdateSubscription(pObj, pSub)) {
taos_unsubscribe(pSub, 1);
return NULL;
}
}
pSub->interval = interval;
@ -417,10 +419,80 @@ TAOS_SUB *taos_subscribe(TAOS *taos, int restart, const char* topic, const char
return pSub;
}
SSqlObj* recreateSqlObj(SSub* pSub) {
SSqlObj* pSql = calloc(1, sizeof(SSqlObj));
if (pSql == NULL) {
return NULL;
}
pSql->signature = pSql;
pSql->pTscObj = pSub->taos;
SSqlCmd* pCmd = &pSql->cmd;
SSqlRes* pRes = &pSql->res;
if (tsem_init(&pSql->rspSem, 0, 0) == -1) {
tscFreeSqlObj(pSql);
return NULL;
}
pSql->param = pSub;
pSql->maxRetry = TSDB_MAX_REPLICA;
pSql->fp = asyncCallback;
pSql->fetchFp = asyncCallback;
pSql->sqlstr = strdup(pSub->pSql->sqlstr);
if (pSql->sqlstr == NULL) {
tscFreeSqlObj(pSql);
return NULL;
}
pRes->qhandle = 0;
pRes->numOfRows = 1;
int code = tscAllocPayload(pCmd, TSDB_DEFAULT_PAYLOAD_SIZE);
if (code != TSDB_CODE_SUCCESS) {
tscFreeSqlObj(pSql);
return NULL;
}
registerSqlObj(pSql);
code = tsParseSql(pSql, true);
if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) {
tsem_wait(&pSub->sem);
code = pSql->res.code;
}
if (code != TSDB_CODE_SUCCESS) {
taosReleaseRef(tscObjRef, pSql->self);
return NULL;
}
if (pSql->cmd.command != TSDB_SQL_SELECT) {
taosReleaseRef(tscObjRef, pSql->self);
return NULL;
}
return pSql;
}
TAOS_RES *taos_consume(TAOS_SUB *tsub) {
SSub *pSub = (SSub *)tsub;
if (pSub == NULL) return NULL;
if (pSub->pSql->cmd.command == TSDB_SQL_RETRIEVE_EMPTY_RESULT) {
SSqlObj* pSql = recreateSqlObj(pSub);
if (pSql == NULL) {
return NULL;
}
if (pSub->pSql->self != 0) {
taosReleaseRef(tscObjRef, pSub->pSql->self);
} else {
tscFreeSqlObj(pSub->pSql);
}
pSub->pSql = pSql;
pSql->pSubscription = pSub;
}
tscSaveSubscriptionProgress(pSub);
SSqlObj *pSql = pSub->pSql;
@ -512,10 +584,13 @@ void taos_unsubscribe(TAOS_SUB *tsub, int keepProgress) {
}
if (pSub->pSql != NULL) {
taos_free_result(pSub->pSql);
if (pSub->pSql->self != 0) {
taosReleaseRef(tscObjRef, pSub->pSql->self);
} else {
tscFreeSqlObj(pSub->pSql);
}
}
tscFreeSqlObj(pSub->pSql);
taosArrayDestroy(pSub->progress);
tsem_destroy(&pSub->sem);
memset(pSub, 0, sizeof(*pSub));

View File

@ -17,6 +17,7 @@
#include "taosmsg.h"
#include "tref.h"
#include "trpc.h"
#include "tnote.h"
#include "tsystem.h"
#include "ttimer.h"
#include "tutil.h"
@ -41,7 +42,6 @@ int tscRefId = -1;
int tscNumOfThreads;
static pthread_once_t tscinit = PTHREAD_ONCE_INIT;
void taosInitNote(int numOfNoteLines, int maxNotes, char* lable);
//void tscUpdateEpSet(void *ahandle, SRpcEpSet *pEpSet);
void tscCheckDiskUsage(void *UNUSED_PARAM(para), void* UNUSED_PARAM(param)) {
@ -78,7 +78,6 @@ int32_t tscInitRpc(const char *user, const char *secretEncrypt, void **pDnodeCon
return 0;
}
void taos_init_imp(void) {
char temp[128] = {0};
@ -104,6 +103,7 @@ void taos_init_imp(void) {
taosReadGlobalCfg();
taosCheckGlobalCfg();
taosInitNotes();
rpcInit();
tscDebug("starting to initialize TAOS client ...");
@ -111,16 +111,6 @@ void taos_init_imp(void) {
}
taosSetCoreDump();
if (tsTscEnableRecordSql != 0) {
taosInitNote(tsNumOfLogLines / 10, 1, (char*)"tsc_note");
}
if (tscSetMgmtEpSetFromCfg(tsFirst, tsSecond) < 0) {
tscError("failed to init mnode EP list");
return;
}
tscInitMsgsFp();
int queueSize = tsMaxConnections*2;
@ -144,7 +134,7 @@ void taos_init_imp(void) {
int64_t refreshTime = 10; // 10 seconds by default
if (tscMetaCache == NULL) {
tscMetaCache = taosCacheInit(TSDB_DATA_TYPE_BINARY, refreshTime, false, tscFreeTableMetaHelper, "tableMeta");
tscObjRef = taosOpenRef(4096, tscFreeRegisteredSqlObj);
tscObjRef = taosOpenRef(40960, tscFreeRegisteredSqlObj);
}
tscRefId = taosOpenRef(200, tscCloseTscObj);

View File

@ -864,7 +864,7 @@ void tscCloseTscObj(void *param) {
rpcClose(pObj->pDnodeConn);
pObj->pDnodeConn = NULL;
}
tfree(pObj->tscCorMgmtEpSet);
pthread_mutex_destroy(&pObj->mutex);
tscDebug("%p DB connection is closed, dnodeConn:%p", pObj, p);
@ -2440,10 +2440,10 @@ char* strdup_throw(const char* str) {
return p;
}
int tscSetMgmtEpSetFromCfg(const char *first, const char *second) {
int tscSetMgmtEpSetFromCfg(const char *first, const char *second, SRpcCorEpSet *corMgmtEpSet) {
corMgmtEpSet->version = 0;
// init mgmt ip set
tscMgmtEpSet.version = 0;
SRpcEpSet *mgmtEpSet = &(tscMgmtEpSet.epSet);
SRpcEpSet *mgmtEpSet = &(corMgmtEpSet->epSet);
mgmtEpSet->numOfEps = 0;
mgmtEpSet->inUse = 0;

View File

@ -35,6 +35,7 @@ extern int32_t tsNumOfMnodes;
extern int32_t tsEnableVnodeBak;
extern int32_t tsEnableTelemetryReporting;
extern char tsEmail[];
extern char tsArbitrator[];
// common
extern int tsRpcTimer;
@ -56,6 +57,7 @@ extern char tsTempDir[];
//query buffer management
extern int32_t tsQueryBufferSize; // maximum allowed usage buffer for each data node during query processing
extern int32_t tsHalfCoresForQuery; // only 50% will be used in query processing
// client
extern int32_t tsTableMetaKeepTimer;
@ -125,6 +127,9 @@ extern char tsMonitorDbName[];
extern char tsInternalPass[];
extern int32_t tsMonitorInterval;
// stream
extern int32_t tsEnableStream;
// internal
extern int32_t tsPrintAuth;
extern int32_t tscEmbedded;
@ -176,7 +181,7 @@ extern int32_t tmrDebugFlag;
extern int32_t sdbDebugFlag;
extern int32_t httpDebugFlag;
extern int32_t mqttDebugFlag;
extern int32_t monitorDebugFlag;
extern int32_t monDebugFlag;
extern int32_t uDebugFlag;
extern int32_t rpcDebugFlag;
extern int32_t odbcDebugFlag;

View File

@ -25,6 +25,8 @@ void extractTableName(const char *tableId, char *name);
char* extractDBName(const char *tableId, char *name);
size_t tableIdPrefix(const char* name, char* prefix, int32_t len);
void extractTableNameFromToken(SStrToken *pToken, SStrToken* pTable);
SSchema tGetTableNameColumnSchema();

View File

@ -107,6 +107,9 @@ int64_t tsMaxRetentWindow = 24 * 3600L; // maximum time window tolerance
// positive value (in MB)
int32_t tsQueryBufferSize = -1;
// only 50% cpu will be used in query processing in dnode
int32_t tsHalfCoresForQuery = 0;
// db parameters
int32_t tsCacheBlockSize = TSDB_DEFAULT_CACHE_BLOCK_SIZE;
int32_t tsBlocksPerVnode = TSDB_DEFAULT_TOTAL_BLOCKS;
@ -161,6 +164,9 @@ char tsMonitorDbName[TSDB_DB_NAME_LEN] = "log";
char tsInternalPass[] = "secretkey";
int32_t tsMonitorInterval = 30; // seconds
// stream
int32_t tsEnableStream = 1;
// internal
int32_t tsPrintAuth = 0;
int32_t tscEmbedded = 0;
@ -197,16 +203,16 @@ int32_t tsVersion = 0;
// log
int32_t tsNumOfLogLines = 10000000;
int32_t mDebugFlag = 135;
int32_t sdbDebugFlag = 135;
int32_t mDebugFlag = 131;
int32_t sdbDebugFlag = 131;
int32_t dDebugFlag = 135;
int32_t vDebugFlag = 135;
int32_t vDebugFlag = 131;
int32_t cDebugFlag = 131;
int32_t jniDebugFlag = 131;
int32_t odbcDebugFlag = 131;
int32_t httpDebugFlag = 131;
int32_t mqttDebugFlag = 131;
int32_t monitorDebugFlag = 131;
int32_t monDebugFlag = 131;
int32_t qDebugFlag = 131;
int32_t rpcDebugFlag = 131;
int32_t uDebugFlag = 131;
@ -214,11 +220,11 @@ int32_t debugFlag = 0;
int32_t sDebugFlag = 135;
int32_t wDebugFlag = 135;
int32_t tsdbDebugFlag = 131;
int32_t cqDebugFlag = 135;
int32_t cqDebugFlag = 131;
int32_t (*monitorStartSystemFp)() = NULL;
void (*monitorStopSystemFp)() = NULL;
void (*monitorExecuteSQLFp)(char *sql) = NULL;
int32_t (*monStartSystemFp)() = NULL;
void (*monStopSystemFp)() = NULL;
void (*monExecuteSQLFp)(char *sql) = NULL;
char *qtypeStr[] = {"rpc", "fwd", "wal", "cq", "query"};
@ -235,7 +241,7 @@ void taosSetAllDebugFlag() {
odbcDebugFlag = debugFlag;
httpDebugFlag = debugFlag;
mqttDebugFlag = debugFlag;
monitorDebugFlag = debugFlag;
monDebugFlag = debugFlag;
qDebugFlag = debugFlag;
rpcDebugFlag = debugFlag;
uDebugFlag = debugFlag;
@ -276,15 +282,15 @@ bool taosCfgDynamicOptions(char *msg) {
if (strncasecmp(cfg->option, "monitor", olen) == 0) {
if (1 == vint) {
if (monitorStartSystemFp) {
(*monitorStartSystemFp)();
if (monStartSystemFp) {
(*monStartSystemFp)();
uInfo("monitor is enabled");
} else {
uError("monitor can't be updated, for monitor not initialized");
}
} else {
if (monitorStopSystemFp) {
(*monitorStopSystemFp)();
if (monStopSystemFp) {
(*monStopSystemFp)();
uInfo("monitor is disabled");
} else {
uError("monitor can't be updated, for monitor not initialized");
@ -307,8 +313,8 @@ bool taosCfgDynamicOptions(char *msg) {
}
if (strncasecmp(option, "resetQueryCache", 15) == 0) {
if (monitorExecuteSQLFp) {
(*monitorExecuteSQLFp)("resetQueryCache");
if (monExecuteSQLFp) {
(*monExecuteSQLFp)("resetQueryCache");
uInfo("resetquerycache is executed");
} else {
uError("resetquerycache can't be executed, for monitor not started");
@ -410,7 +416,7 @@ static void doInitGlobalConfig(void) {
cfg.option = "arbitrator";
cfg.ptr = tsArbitrator;
cfg.valType = TAOS_CFG_VTYPE_STRING;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_CLIENT;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW | TSDB_CFG_CTYPE_B_CLIENT;
cfg.minValue = 0;
cfg.maxValue = 0;
cfg.ptrLength = TSDB_EP_LEN;
@ -881,11 +887,21 @@ static void doInitGlobalConfig(void) {
cfg.unitType = TAOS_CFG_UTYPE_BYTE;
taosInitConfigOption(cfg);
cfg.option = "halfCoresForQuery";
cfg.ptr = &tsHalfCoresForQuery;
cfg.valType = TAOS_CFG_VTYPE_INT32;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW;
cfg.minValue = 0;
cfg.maxValue = 1;
cfg.ptrLength = 1;
cfg.unitType = TAOS_CFG_UTYPE_NONE;
taosInitConfigOption(cfg);
// locale & charset
cfg.option = "timezone";
cfg.ptr = tsTimezone;
cfg.valType = TAOS_CFG_VTYPE_STRING;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_CLIENT;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW | TSDB_CFG_CTYPE_B_CLIENT;
cfg.minValue = 0;
cfg.maxValue = 0;
cfg.ptrLength = tListLen(tsTimezone);
@ -895,7 +911,7 @@ static void doInitGlobalConfig(void) {
cfg.option = "locale";
cfg.ptr = tsLocale;
cfg.valType = TAOS_CFG_VTYPE_STRING;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_CLIENT;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW | TSDB_CFG_CTYPE_B_CLIENT;
cfg.minValue = 0;
cfg.maxValue = 0;
cfg.ptrLength = tListLen(tsLocale);
@ -905,7 +921,7 @@ static void doInitGlobalConfig(void) {
cfg.option = "charset";
cfg.ptr = tsCharset;
cfg.valType = TAOS_CFG_VTYPE_STRING;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_CLIENT;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW | TSDB_CFG_CTYPE_B_CLIENT;
cfg.minValue = 0;
cfg.maxValue = 0;
cfg.ptrLength = tListLen(tsCharset);
@ -1015,6 +1031,16 @@ static void doInitGlobalConfig(void) {
cfg.unitType = TAOS_CFG_UTYPE_NONE;
taosInitConfigOption(cfg);
cfg.option = "stream";
cfg.ptr = &tsEnableStream;
cfg.valType = TAOS_CFG_VTYPE_INT32;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW;
cfg.minValue = 0;
cfg.maxValue = 1;
cfg.ptrLength = 1;
cfg.unitType = TAOS_CFG_UTYPE_NONE;
taosInitConfigOption(cfg);
cfg.option = "httpEnableRecordSql";
cfg.ptr = &tsHttpEnableRecordSql;
cfg.valType = TAOS_CFG_VTYPE_INT32;
@ -1227,8 +1253,8 @@ static void doInitGlobalConfig(void) {
cfg.unitType = TAOS_CFG_UTYPE_NONE;
taosInitConfigOption(cfg);
cfg.option = "monitorDebugFlag";
cfg.ptr = &monitorDebugFlag;
cfg.option = "monDebugFlag";
cfg.ptr = &monDebugFlag;
cfg.valType = TAOS_CFG_VTYPE_INT32;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_LOG;
cfg.minValue = 0;
@ -1277,7 +1303,7 @@ static void doInitGlobalConfig(void) {
cfg.unitType = TAOS_CFG_UTYPE_NONE;
taosInitConfigOption(cfg);
cfg.option = "tscEnableRecordSql";
cfg.option = "enableRecordSql";
cfg.ptr = &tsTscEnableRecordSql;
cfg.valType = TAOS_CFG_VTYPE_INT32;
cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG;

View File

@ -39,6 +39,13 @@ char* extractDBName(const char* tableId, char* name) {
return strncpy(name, &tableId[offset1 + 1], len);
}
size_t tableIdPrefix(const char* name, char* prefix, int32_t len) {
tstrncpy(prefix, name, len);
strcat(prefix, TS_PATH_DELIMITER);
return strlen(prefix);
}
SSchema tGetTableNameColumnSchema() {
SSchema s = {0};
s.bytes = TSDB_TABLE_NAME_LEN - 1 + VARSTR_HEADER_SIZE;
@ -198,4 +205,4 @@ SSchema tscGetTbnameColumnSchema() {
strcpy(s.name, TSQL_TBNAME_L);
return s;
}
}

@ -1 +1 @@
Subproject commit 8c58c512b6acda8bcdfa48fdc7140227b5221766
Subproject commit 050667e5b4d0eafa5387e4283e713559b421203f

@ -1 +1 @@
Subproject commit d598db167eb256fe67409b7bb3d0eb7fffc3ff8c
Subproject commit ec77d9049a719dabfd1a7c1122a209e201861944

View File

@ -8,7 +8,7 @@ IF (TD_MVN_INSTALLED)
ADD_CUSTOM_COMMAND(OUTPUT ${JDBC_CMD_NAME}
POST_BUILD
COMMAND mvn -Dmaven.test.skip=true install -f ${CMAKE_CURRENT_SOURCE_DIR}/pom.xml
COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/target/taos-jdbcdriver-2.0.13-dist.jar ${LIBRARY_OUTPUT_PATH}
COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/target/taos-jdbcdriver-2.0.15-dist.jar ${LIBRARY_OUTPUT_PATH}
COMMAND mvn -Dmaven.test.skip=true clean -f ${CMAKE_CURRENT_SOURCE_DIR}/pom.xml
COMMENT "build jdbc driver")
ADD_CUSTOM_TARGET(${JDBC_TARGET_NAME} ALL WORKING_DIRECTORY ${EXECUTABLE_OUTPUT_PATH} DEPENDS ${JDBC_CMD_NAME})

View File

@ -5,7 +5,7 @@
<groupId>com.taosdata.jdbc</groupId>
<artifactId>taos-jdbcdriver</artifactId>
<version>2.0.13</version>
<version>2.0.15</version>
<packaging>jar</packaging>
<name>JDBCDriver</name>
@ -36,7 +36,6 @@
</developer>
</developers>
<dependencies>
<dependency>
<groupId>commons-logging</groupId>

View File

@ -3,7 +3,7 @@
<modelVersion>4.0.0</modelVersion>
<groupId>com.taosdata.jdbc</groupId>
<artifactId>taos-jdbcdriver</artifactId>
<version>2.0.13</version>
<version>2.0.15</version>
<packaging>jar</packaging>
<name>JDBCDriver</name>
<url>https://github.com/taosdata/TDengine/tree/master/src/connector/jdbc</url>

View File

@ -81,7 +81,7 @@ public class TSDBStatement implements Statement {
}
if (!this.connector.isUpdateQuery(pSql)) {
TSDBResultSet res = new TSDBResultSet(this.connector, resultSetPointer);
TSDBResultSet res = new TSDBResultSet(this.connector, resultSetPointer);
res.setBatchFetch(this.connection.getBatchFetch());
return res;
} else {
@ -125,7 +125,8 @@ public class TSDBStatement implements Statement {
}
public int getMaxFieldSize() throws SQLException {
throw new SQLException(TSDBConstants.UNSUPPORT_METHOD_EXCEPTIONZ_MSG);
return 0;
// throw new SQLException(TSDBConstants.UNSUPPORT_METHOD_EXCEPTIONZ_MSG);
}
public void setMaxFieldSize(int max) throws SQLException {
@ -218,7 +219,8 @@ public class TSDBStatement implements Statement {
}
public int getFetchDirection() throws SQLException {
throw new SQLException(TSDBConstants.UNSUPPORT_METHOD_EXCEPTIONZ_MSG);
return ResultSet.FETCH_FORWARD;
// throw new SQLException(TSDBConstants.UNSUPPORT_METHOD_EXCEPTIONZ_MSG);
}
/*

View File

@ -5,7 +5,7 @@ with open("README.md", "r") as fh:
setuptools.setup(
name="taos",
version="2.0.2",
version="2.0.3",
author="Taosdata Inc.",
author_email="support@taosdata.com",
description="TDengine python client package",

View File

@ -3,7 +3,7 @@ from .connection import TDengineConnection
from .cursor import TDengineCursor
# Globals
apilevel = '2.0'
apilevel = '2.0.3'
threadsafety = 0
paramstyle = 'pyformat'

View File

@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
if num_of_rows > 0:
return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]]
else:
@ -106,6 +107,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
res=[]
if num_of_rows > 0:
for i in range(abs(num_of_rows)):

View File

@ -26,7 +26,7 @@ class TDengineCursor(object):
"""
def __init__(self, connection=None):
self._description = None
self._description = []
self._rowcount = -1
self._connection = None
self._result = None
@ -234,7 +234,7 @@ class TDengineCursor(object):
def _reset_result(self):
"""Reset the result to unused version.
"""
self._description = None
self._description = []
self._rowcount = -1
if self._result is not None:
CTaosInterface.freeResult(self._result)

View File

@ -5,7 +5,7 @@ with open("README.md", "r") as fh:
setuptools.setup(
name="taos",
version="2.0.2",
version="2.0.3",
author="Taosdata Inc.",
author_email="support@taosdata.com",
description="TDengine python client package",

View File

@ -3,7 +3,7 @@ from .connection import TDengineConnection
from .cursor import TDengineCursor
# Globals
apilevel = '2.0'
apilevel = '2.0.3'
threadsafety = 0
paramstyle = 'pyformat'
@ -21,4 +21,4 @@ def connect(*args, **kwargs):
@rtype: TDengineConnector
"""
return TDengineConnection(*args, **kwargs)
return TDengineConnection(*args, **kwargs)

View File

@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
if num_of_rows > 0:
return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]]
else:
@ -106,6 +107,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
res=[]
if num_of_rows > 0:
for i in range(abs(num_of_rows)):

View File

@ -27,7 +27,7 @@ class TDengineCursor(object):
"""
def __init__(self, connection=None):
self._description = None
self._description = []
self._rowcount = -1
self._connection = None
self._result = None
@ -242,7 +242,7 @@ class TDengineCursor(object):
def _reset_result(self):
"""Reset the result to unused version.
"""
self._description = None
self._description = []
self._rowcount = -1
if self._result is not None:
CTaosInterface.freeResult(self._result)

View File

@ -5,7 +5,7 @@ with open("README.md", "r") as fh:
setuptools.setup(
name="taos",
version="2.0.2",
version="2.0.3",
author="Taosdata Inc.",
author_email="support@taosdata.com",
description="TDengine python client package",

View File

@ -3,7 +3,7 @@ from .connection import TDengineConnection
from .cursor import TDengineCursor
# Globals
apilevel = '2.0'
apilevel = '2.0.3'
threadsafety = 0
paramstyle = 'pyformat'

View File

@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
if num_of_rows > 0:
return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]]
else:
@ -106,6 +107,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
res=[]
if num_of_rows > 0:
for i in range(abs(num_of_rows)):

View File

@ -27,7 +27,7 @@ class TDengineCursor(object):
"""
def __init__(self, connection=None):
self._description = None
self._description = []
self._rowcount = -1
self._connection = None
self._result = None
@ -193,7 +193,7 @@ class TDengineCursor(object):
def _reset_result(self):
"""Reset the result to unused version.
"""
self._description = None
self._description = []
self._rowcount = -1
if self._result is not None:
CTaosInterface.freeResult(self._result)

View File

@ -5,7 +5,7 @@ with open("README.md", "r") as fh:
setuptools.setup(
name="taos",
version="2.0.2",
version="2.0.3",
author="Taosdata Inc.",
author_email="support@taosdata.com",
description="TDengine python client package",

View File

@ -3,7 +3,7 @@ from .connection import TDengineConnection
from .cursor import TDengineCursor
# Globals
apilevel = '2.0'
apilevel = '2.0.3'
threadsafety = 0
paramstyle = 'pyformat'

View File

@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
if num_of_rows > 0:
return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]]
else:
@ -108,6 +109,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False):
def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False):
"""Function to convert C binary row to python row
"""
assert(nbytes is not None)
res=[]
if num_of_rows > 0:
for i in range(abs(num_of_rows)):

View File

@ -28,7 +28,7 @@ class TDengineCursor(object):
"""
def __init__(self, connection=None):
self._description = None
self._description = []
self._rowcount = -1
self._connection = None
self._result = None
@ -194,7 +194,7 @@ class TDengineCursor(object):
def _reset_result(self):
"""Reset the result to unused version.
"""
self._description = None
self._description = []
self._rowcount = -1
if self._result is not None:
CTaosInterface.freeResult(self._result)

View File

@ -57,6 +57,7 @@ typedef struct SCqObj {
uint64_t uid;
int32_t tid; // table ID
int32_t rowSize; // bytes of a row
char * dstTable;
char * sqlStr; // SQL string
STSchema * pSchema; // pointer to schema array
void * pStream;
@ -69,6 +70,9 @@ static void cqProcessStreamRes(void *param, TAOS_RES *tres, TAOS_ROW row);
static void cqCreateStream(SCqContext *pContext, SCqObj *pObj);
void *cqOpen(void *ahandle, const SCqCfg *pCfg) {
if (tsEnableStream == 0) {
return NULL;
}
SCqContext *pContext = calloc(sizeof(SCqContext), 1);
if (pContext == NULL) {
terrno = TAOS_SYSTEM_ERROR(errno);
@ -99,6 +103,9 @@ void *cqOpen(void *ahandle, const SCqCfg *pCfg) {
}
void cqClose(void *handle) {
if (tsEnableStream == 0) {
return;
}
SCqContext *pContext = handle;
if (handle == NULL) return;
@ -129,6 +136,9 @@ void cqClose(void *handle) {
}
void cqStart(void *handle) {
if (tsEnableStream == 0) {
return;
}
SCqContext *pContext = handle;
if (pContext->dbConn || pContext->master) return;
@ -147,6 +157,9 @@ void cqStart(void *handle) {
}
void cqStop(void *handle) {
if (tsEnableStream == 0) {
return;
}
SCqContext *pContext = handle;
cInfo("vgId:%d, stop all CQs", pContext->vgId);
if (pContext->dbConn == NULL || pContext->master == 0) return;
@ -173,16 +186,21 @@ void cqStop(void *handle) {
pthread_mutex_unlock(&pContext->mutex);
}
void *cqCreate(void *handle, uint64_t uid, int32_t tid, char *sqlStr, STSchema *pSchema) {
void *cqCreate(void *handle, uint64_t uid, int32_t sid, const char* dstTable, char *sqlStr, STSchema *pSchema) {
if (tsEnableStream == 0) {
return NULL;
}
SCqContext *pContext = handle;
SCqObj *pObj = calloc(sizeof(SCqObj), 1);
if (pObj == NULL) return NULL;
pObj->uid = uid;
pObj->tid = tid;
pObj->sqlStr = malloc(strlen(sqlStr)+1);
strcpy(pObj->sqlStr, sqlStr);
pObj->tid = sid;
if (dstTable != NULL) {
pObj->dstTable = strdup(dstTable);
}
pObj->sqlStr = strdup(sqlStr);
pObj->pSchema = tdDupSchema(pSchema);
pObj->rowSize = schemaTLen(pSchema);
@ -203,6 +221,9 @@ void *cqCreate(void *handle, uint64_t uid, int32_t tid, char *sqlStr, STSchema *
}
void cqDrop(void *handle) {
if (tsEnableStream == 0) {
return;
}
SCqObj *pObj = handle;
SCqContext *pContext = pObj->pContext;
@ -229,6 +250,7 @@ void cqDrop(void *handle) {
cInfo("vgId:%d, id:%d CQ:%s is dropped", pContext->vgId, pObj->tid, pObj->sqlStr);
tdFreeSchema(pObj->pSchema);
free(pObj->dstTable);
free(pObj->sqlStr);
free(pObj);
@ -274,6 +296,7 @@ static void cqCreateStream(SCqContext *pContext, SCqObj *pObj) {
if (pObj->pStream == NULL) {
pObj->pStream = taos_open_stream(pContext->dbConn, pObj->sqlStr, cqProcessStreamRes, 0, pObj, NULL);
if (pObj->pStream) {
tscSetStreamDestTable(pObj->pStream, pObj->dstTable);
pContext->num++;
cInfo("vgId:%d, id:%d CQ:%s is openned", pContext->vgId, pObj->tid, pObj->sqlStr);
} else {

View File

@ -70,7 +70,7 @@ int main(int argc, char *argv[]) {
tdDestroyTSchemaBuilder(&schemaBuilder);
for (int sid =1; sid<10; ++sid) {
cqCreate(pCq, sid, sid, "select avg(speed) from demo.t1 sliding(1s) interval(5s)", pSchema);
cqCreate(pCq, sid, sid, NULL, "select avg(speed) from demo.t1 sliding(1s) interval(5s)", pSchema);
}
tdFreeSchema(pSchema);

View File

@ -12,7 +12,7 @@ AUX_SOURCE_DIRECTORY(src SRC)
IF (TD_LINUX)
ADD_EXECUTABLE(taosd ${SRC})
TARGET_LINK_LIBRARIES(taosd mnode monitor http tsdb twal vnode cJson lz4)
TARGET_LINK_LIBRARIES(taosd mnode monitor http tsdb twal vnode cJson lz4 balance sync)
IF (TD_SOMODE_STATIC)
TARGET_LINK_LIBRARIES(taosd taos_static)
@ -32,10 +32,6 @@ IF (TD_LINUX)
TARGET_LINK_LIBRARIES(taosd mqtt)
ENDIF ()
IF (TD_SYNC)
TARGET_LINK_LIBRARIES(taosd balance sync)
ENDIF ()
SET(PREPARE_ENV_CMD "prepare_env_cmd")
SET(PREPARE_ENV_TARGET "prepare_env_target")
ADD_CUSTOM_COMMAND(OUTPUT ${PREPARE_ENV_CMD}

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitCfg();
void dnodeCleanupCfg();

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitCheck();
void dnodeCleanupCheck();

View File

@ -19,8 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "taosmsg.h"
#include "dnodeInt.h"
int32_t dnodeInitEps();
void dnodeCleanupEps();

View File

@ -19,8 +19,13 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "taoserror.h"
#include "taosmsg.h"
#include "tlog.h"
#include "trpc.h"
#include "tglobal.h"
#include "dnode.h"
#include "vnode.h"
extern int32_t dDebugFlag;
@ -31,6 +36,14 @@ extern int32_t dDebugFlag;
#define dDebug(...) { if (dDebugFlag & DEBUG_DEBUG) { taosPrintLog("DND ", dDebugFlag, __VA_ARGS__); }}
#define dTrace(...) { if (dDebugFlag & DEBUG_TRACE) { taosPrintLog("DND ", dDebugFlag, __VA_ARGS__); }}
typedef enum {
TSDB_RUN_STATUS_INITIALIZE,
TSDB_RUN_STATUS_RUNING,
TSDB_RUN_STATUS_STOPPED
} SRunStatus;
SRunStatus dnodeGetRunStatus();
#ifdef __cplusplus
}
#endif

View File

@ -19,16 +19,19 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "taosmsg.h"
#include "dnodeInt.h"
int32_t dnodeInitMInfos();
void dnodeCleanupMInfos();
void dnodeUpdateMInfos(SMnodeInfos *minfos);
void dnodeUpdateEpSetForPeer(SRpcEpSet *epSet);
void dnodeGetMInfos(SMnodeInfos *minfos);
void dnodeUpdateMInfos(SMInfos *pMinfos);
void dnodeUpdateEpSetForPeer(SRpcEpSet *pEpSet);
void dnodeGetMInfos(SMInfos *pMinfos);
bool dnodeIsMasterEp(char *ep);
void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell);
void dnodeGetEpSetForPeer(SRpcEpSet *epSet);
void dnodeGetEpSetForShell(SRpcEpSet *epSet);
#ifdef __cplusplus
}
#endif

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitMPeer();
void dnodeCleanupMPeer();

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitMRead();
void dnodeCleanupMRead();

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitMWrite();
void dnodeCleanupMWrite();

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitSystem();
void dnodeCleanUpSystem();

View File

@ -19,10 +19,11 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitModules();
void dnodeStartModules();
void dnodeCleanupModules();
bool dnodeStartMnode(SMInfos *pMinfos);
void dnodeProcessModuleStatus(uint32_t moduleStatus);
#ifdef __cplusplus

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitServer();
void dnodeCleanupServer();

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitShell();
void dnodeCleanupShell();

33
src/dnode/inc/dnodeStep.h Normal file
View File

@ -0,0 +1,33 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_DNODE_STEP_H
#define TDENGINE_DNODE_STEP_H
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeStepInit(SStep *pSteps, int32_t stepSize);
void dnodeStepCleanup(SStep *pSteps, int32_t stepSize);
void dnodeReportStep(char *name, char *desc, int8_t finished);
void dnodeSendStartupStep(SRpcMsg *pMsg);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitTelemetry();
void dnodeCleanupTelemetry();

View File

@ -0,0 +1,32 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_DNODE_VMGMT_H
#define TDENGINE_DNODE_VMGMT_H
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitVMgmt();
void dnodeCleanupVMgmt();
void dnodeDispatchToVMgmtQueue(SRpcMsg *rpcMsg);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitVRead();
void dnodeCleanupVRead();

View File

@ -19,6 +19,7 @@
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitVWrite();
void dnodeCleanupVWrite();

View File

@ -0,0 +1,34 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_DNODE_VNODES_H
#define TDENGINE_DNODE_VNODES_H
#ifdef __cplusplus
extern "C" {
#endif
#include "dnodeInt.h"
int32_t dnodeInitVnodes();
void dnodeCleanupVnodes();
int32_t dnodeInitStatusTimer();
void dnodeCleanupStatusTimer();
void dnodeSendStatusMsgToMnode();
#ifdef __cplusplus
}
#endif
#endif

View File

@ -16,9 +16,6 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "cJSON.h"
#include "tglobal.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeCfg.h"
static SDnodeCfg tsCfg = {0};

View File

@ -15,8 +15,6 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "tglobal.h"
#include "dnodeInt.h"
#include "dnodeCheck.h"
typedef struct {

View File

@ -16,10 +16,7 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "cJSON.h"
#include "tglobal.h"
#include "hash.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeEps.h"
static SDnodeEps *tsEps = NULL;
@ -236,7 +233,14 @@ PRASE_EPS_OVER:
dnodeResetEps(eps);
if (eps) free(eps);
#if 0
dnodeUpdateEp(dnodeGetDnodeId(), tsLocalEp, tsLocalFqdn, &tsServerPort);
#else
if (dnodeCheckEpChanged(dnodeGetDnodeId(), tsLocalEp)) {
dError("dnode:%d, localEp is different from %s in dnodeEps.json and need reconfigured", dnodeGetDnodeId(), tsLocalEp);
return -1;
}
#endif
terrno = 0;
return 0;

View File

@ -16,18 +16,15 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "cJSON.h"
#include "tglobal.h"
#include "mnode.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeMInfos.h"
static SMnodeInfos tsMInfos;
static SRpcEpSet tsMEpSet;
static SMInfos tsMInfos;
static SRpcEpSet tsMEpSet;
static pthread_mutex_t tsMInfosMutex;
static void dnodeResetMInfos(SMnodeInfos *minfos);
static void dnodePrintMInfos(SMnodeInfos *minfos);
static void dnodeResetMInfos(SMInfos *minfos);
static void dnodePrintMInfos(SMInfos *minfos);
static int32_t dnodeReadMInfos();
static int32_t dnodeWriteMInfos();
@ -44,14 +41,14 @@ int32_t dnodeInitMInfos() {
void dnodeCleanupMInfos() { pthread_mutex_destroy(&tsMInfosMutex); }
void dnodeUpdateMInfos(SMnodeInfos *minfos) {
if (minfos->mnodeNum <= 0 || minfos->mnodeNum > 3) {
dError("invalid mnode infos, mnodeNum:%d", minfos->mnodeNum);
void dnodeUpdateMInfos(SMInfos *pMinfos) {
if (pMinfos->mnodeNum <= 0 || pMinfos->mnodeNum > 3) {
dError("invalid mnode infos, mnodeNum:%d", pMinfos->mnodeNum);
return;
}
for (int32_t i = 0; i < minfos->mnodeNum; ++i) {
SMnodeInfo *minfo = &minfos->mnodeInfos[i];
for (int32_t i = 0; i < pMinfos->mnodeNum; ++i) {
SMInfo *minfo = &pMinfos->mnodeInfos[i];
minfo->mnodeId = htonl(minfo->mnodeId);
if (minfo->mnodeId <= 0 || strlen(minfo->mnodeEp) <= 5) {
dError("invalid mnode info:%d, mnodeId:%d mnodeEp:%s", i, minfo->mnodeId, minfo->mnodeEp);
@ -60,14 +57,14 @@ void dnodeUpdateMInfos(SMnodeInfos *minfos) {
}
pthread_mutex_lock(&tsMInfosMutex);
if (minfos->mnodeNum != tsMInfos.mnodeNum) {
dnodeResetMInfos(minfos);
if (pMinfos->mnodeNum != tsMInfos.mnodeNum) {
dnodeResetMInfos(pMinfos);
dnodeWriteMInfos();
sdbUpdateAsync();
} else {
int32_t size = sizeof(SMnodeInfos);
if (memcmp(minfos, &tsMInfos, size) != 0) {
dnodeResetMInfos(minfos);
int32_t size = sizeof(SMInfos);
if (memcmp(pMinfos, &tsMInfos, size) != 0) {
dnodeResetMInfos(pMinfos);
dnodeWriteMInfos();
sdbUpdateAsync();
}
@ -99,11 +96,11 @@ bool dnodeIsMasterEp(char *ep) {
return isMaster;
}
void dnodeGetMInfos(SMnodeInfos *minfos) {
void dnodeGetMInfos(SMInfos *pMinfos) {
pthread_mutex_lock(&tsMInfosMutex);
memcpy(minfos, &tsMInfos, sizeof(SMnodeInfos));
memcpy(pMinfos, &tsMInfos, sizeof(SMInfos));
for (int32_t i = 0; i < tsMInfos.mnodeNum; ++i) {
minfos->mnodeInfos[i].mnodeId = htonl(tsMInfos.mnodeInfos[i].mnodeId);
pMinfos->mnodeInfos[i].mnodeId = htonl(tsMInfos.mnodeInfos[i].mnodeId);
}
pthread_mutex_unlock(&tsMInfosMutex);
}
@ -123,15 +120,15 @@ void dnodeGetEpSetForShell(SRpcEpSet *epSet) {
pthread_mutex_unlock(&tsMInfosMutex);
}
static void dnodePrintMInfos(SMnodeInfos *minfos) {
dInfo("print mnode infos, mnodeNum:%d inUse:%d", minfos->mnodeNum, minfos->inUse);
for (int32_t i = 0; i < minfos->mnodeNum; i++) {
dInfo("mnode index:%d, %s", minfos->mnodeInfos[i].mnodeId, minfos->mnodeInfos[i].mnodeEp);
static void dnodePrintMInfos(SMInfos *pMinfos) {
dInfo("print minfos, mnodeNum:%d inUse:%d", pMinfos->mnodeNum, pMinfos->inUse);
for (int32_t i = 0; i < pMinfos->mnodeNum; i++) {
dInfo("mnode index:%d, %s", pMinfos->mnodeInfos[i].mnodeId, pMinfos->mnodeInfos[i].mnodeEp);
}
}
static void dnodeResetMInfos(SMnodeInfos *minfos) {
if (minfos == NULL) {
static void dnodeResetMInfos(SMInfos *pMinfos) {
if (pMinfos == NULL) {
tsMEpSet.numOfEps = 1;
taosGetFqdnPortFromEp(tsFirst, tsMEpSet.fqdn[0], &tsMEpSet.port[0]);
@ -142,10 +139,10 @@ static void dnodeResetMInfos(SMnodeInfos *minfos) {
return;
}
if (minfos->mnodeNum == 0) return;
if (pMinfos->mnodeNum == 0) return;
int32_t size = sizeof(SMnodeInfos);
memcpy(&tsMInfos, minfos, size);
int32_t size = sizeof(SMInfos);
memcpy(&tsMInfos, pMinfos, size);
tsMEpSet.inUse = tsMInfos.inUse;
tsMEpSet.numOfEps = tsMInfos.mnodeNum;
@ -153,16 +150,17 @@ static void dnodeResetMInfos(SMnodeInfos *minfos) {
taosGetFqdnPortFromEp(tsMInfos.mnodeInfos[i].mnodeEp, tsMEpSet.fqdn[i], &tsMEpSet.port[i]);
}
dnodePrintMInfos(minfos);
dnodePrintMInfos(pMinfos);
}
static int32_t dnodeReadMInfos() {
int32_t len = 0;
int32_t maxLen = 2000;
char * content = calloc(1, maxLen + 1);
cJSON * root = NULL;
FILE * fp = NULL;
SMnodeInfos minfos = {0};
int32_t len = 0;
int32_t maxLen = 2000;
char * content = calloc(1, maxLen + 1);
cJSON * root = NULL;
FILE * fp = NULL;
SMInfos minfos = {0};
bool nodeChanged = false;
char file[TSDB_FILENAME_LEN + 20] = {0};
sprintf(file, "%s/mnodeEpSet.json", tsDnodeDir);
@ -221,14 +219,19 @@ static int32_t dnodeReadMInfos() {
dError("failed to read mnodeEpSet.json, nodeId not found");
goto PARSE_MINFOS_OVER;
}
minfos.mnodeInfos[i].mnodeId = nodeId->valueint;
cJSON *nodeEp = cJSON_GetObjectItem(nodeInfo, "nodeEp");
if (!nodeEp || nodeEp->type != cJSON_String || nodeEp->valuestring == NULL) {
dError("failed to read mnodeEpSet.json, nodeName not found");
goto PARSE_MINFOS_OVER;
}
strncpy(minfos.mnodeInfos[i].mnodeEp, nodeEp->valuestring, TSDB_EP_LEN);
SMInfo *pMinfo = &minfos.mnodeInfos[i];
pMinfo->mnodeId = nodeId->valueint;
tstrncpy(pMinfo->mnodeEp, nodeEp->valuestring, TSDB_EP_LEN);
bool changed = dnodeCheckEpChanged(pMinfo->mnodeId, pMinfo->mnodeEp);
if (changed) nodeChanged = changed;
}
dInfo("read file %s successed", file);
@ -241,10 +244,15 @@ PARSE_MINFOS_OVER:
terrno = 0;
for (int32_t i = 0; i < minfos.mnodeNum; ++i) {
SMnodeInfo *mInfo = &minfos.mnodeInfos[i];
SMInfo *mInfo = &minfos.mnodeInfos[i];
dnodeUpdateEp(mInfo->mnodeId, mInfo->mnodeEp, NULL, NULL);
}
dnodeResetMInfos(&minfos);
if (nodeChanged) {
dnodeWriteMInfos();
}
return 0;
}
@ -286,3 +294,25 @@ static int32_t dnodeWriteMInfos() {
dInfo("successed to write %s", file);
return 0;
}
void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) {
SRpcConnInfo connInfo = {0};
rpcGetConnInfo(rpcMsg->handle, &connInfo);
SRpcEpSet epSet = {0};
if (forShell) {
dnodeGetEpSetForShell(&epSet);
} else {
dnodeGetEpSetForPeer(&epSet);
}
dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType],
taosIpStr(connInfo.clientIp), connInfo.user, epSet.numOfEps, epSet.inUse);
for (int32_t i = 0; i < epSet.numOfEps; ++i) {
dDebug("mnode index:%d %s:%d", i, epSet.fqdn[i], epSet.port[i]);
epSet.port[i] = htons(epSet.port[i]);
}
rpcSendRedirectRsp(rpcMsg->handle, &epSet);
}

View File

@ -15,16 +15,11 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tutil.h"
#include "tqueue.h"
#include "twal.h"
#include "tglobal.h"
#include "mnode.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeMgmt.h"
#include "dnodeVMgmt.h"
#include "dnodeMInfos.h"
#include "dnodeMWrite.h"
typedef struct {

View File

@ -15,16 +15,11 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tutil.h"
#include "tqueue.h"
#include "twal.h"
#include "tglobal.h"
#include "mnode.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeMgmt.h"
#include "dnodeVMgmt.h"
#include "dnodeMInfos.h"
#include "dnodeMRead.h"
typedef struct {

View File

@ -15,17 +15,11 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tutil.h"
#include "ttimer.h"
#include "tqueue.h"
#include "twal.h"
#include "tglobal.h"
#include "mnode.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeMgmt.h"
#include "dnodeVMgmt.h"
#include "dnodeMInfos.h"
#include "dnodeMWrite.h"
typedef struct {
@ -127,7 +121,7 @@ void dnodeDispatchToMWriteQueue(SRpcMsg *pMsg) {
dnodeSendRedirectMsg(pMsg, true);
} else {
SMnodeMsg *pWrite = mnodeCreateMsg(pMsg);
dDebug("msg:%p, app:%p type:%s is put into mwrite queue:%p", pWrite, pWrite->rpcMsg.ahandle,
dTrace("msg:%p, app:%p type:%s is put into mwrite queue:%p", pWrite, pWrite->rpcMsg.ahandle,
taosMsg[pWrite->rpcMsg.msgType], tsMWriteQueue);
taosWriteQitem(tsMWriteQueue, TAOS_QTYPE_RPC, pWrite);
}
@ -136,7 +130,7 @@ void dnodeDispatchToMWriteQueue(SRpcMsg *pMsg) {
}
static void dnodeFreeMWriteMsg(SMnodeMsg *pWrite) {
dDebug("msg:%p, app:%p type:%s is freed from mwrite queue:%p", pWrite, pWrite->rpcMsg.ahandle,
dTrace("msg:%p, app:%p type:%s is freed from mwrite queue:%p", pWrite, pWrite->rpcMsg.ahandle,
taosMsg[pWrite->rpcMsg.msgType], tsMWriteQueue);
mnodeCleanupMsg(pWrite);

View File

@ -16,15 +16,14 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taos.h"
#include "tutil.h"
#include "tnote.h"
#include "ttimer.h"
#include "tconfig.h"
#include "tglobal.h"
#include "tfile.h"
#include "twal.h"
#include "trpc.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeMgmt.h"
// #include "tfs.h"
#include "tsync.h"
#include "dnodeStep.h"
#include "dnodePeer.h"
#include "dnodeModule.h"
#include "dnodeEps.h"
@ -33,50 +32,47 @@
#include "dnodeCheck.h"
#include "dnodeVRead.h"
#include "dnodeVWrite.h"
#include "dnodeVMgmt.h"
#include "dnodeVnodes.h"
#include "dnodeMRead.h"
#include "dnodeMWrite.h"
#include "dnodeMPeer.h"
#include "dnodeShell.h"
#include "dnodeTelemetry.h"
void *tsDnodeTmr = NULL;
static SRunStatus tsRunStatus = TSDB_RUN_STATUS_STOPPED;
static int32_t dnodeInitStorage();
static void dnodeCleanupStorage();
static void dnodeSetRunStatus(SRunStatus status);
static void dnodeCheckDataDirOpenned(char *dir);
static int32_t dnodeInitComponents();
static void dnodeCleanupComponents(int32_t stepId);
static int dnodeCreateDir(const char *dir);
typedef struct {
const char *const name;
int32_t (*init)();
void (*cleanup)();
} SDnodeComponent;
static const SDnodeComponent tsDnodeComponents[] = {
{"tfile", tfInit, tfCleanup},
{"rpc", rpcInit, rpcCleanup},
{"storage", dnodeInitStorage, dnodeCleanupStorage},
{"dnodecfg", dnodeInitCfg, dnodeCleanupCfg},
{"dnodeeps", dnodeInitEps, dnodeCleanupEps},
{"globalcfg" ,taosCheckGlobalCfg, NULL},
{"mnodeinfos",dnodeInitMInfos, dnodeCleanupMInfos},
{"wal", walInit, walCleanUp},
{"check", dnodeInitCheck, dnodeCleanupCheck}, // NOTES: dnodeInitCheck must be behind the dnodeinitStorage component !!!
{"vread", dnodeInitVRead, dnodeCleanupVRead},
{"vwrite", dnodeInitVWrite, dnodeCleanupVWrite},
{"mread", dnodeInitMRead, dnodeCleanupMRead},
{"mwrite", dnodeInitMWrite, dnodeCleanupMWrite},
{"mpeer", dnodeInitMPeer, dnodeCleanupMPeer},
{"client", dnodeInitClient, dnodeCleanupClient},
{"server", dnodeInitServer, dnodeCleanupServer},
{"mgmt", dnodeInitMgmt, dnodeCleanupMgmt},
{"modules", dnodeInitModules, dnodeCleanupModules},
{"mgmt-tmr", dnodeInitMgmtTimer, dnodeCleanupMgmtTimer},
{"shell", dnodeInitShell, dnodeCleanupShell},
{"telemetry", dnodeInitTelemetry, dnodeCleanupTelemetry},
static SStep tsDnodeSteps[] = {
{"dnode-tfile", tfInit, tfCleanup},
{"dnode-rpc", rpcInit, rpcCleanup},
{"dnode-globalcfg", taosCheckGlobalCfg, NULL},
{"dnode-storage", dnodeInitStorage, dnodeCleanupStorage},
{"dnode-cfg", dnodeInitCfg, dnodeCleanupCfg},
{"dnode-eps", dnodeInitEps, dnodeCleanupEps},
{"dnode-minfos", dnodeInitMInfos, dnodeCleanupMInfos},
{"dnode-wal", walInit, walCleanUp},
{"dnode-sync", syncInit, syncCleanUp},
{"dnode-check", dnodeInitCheck, dnodeCleanupCheck}, // NOTES: dnodeInitCheck must be behind the dnodeinitStorage component !!!
{"dnode-vread", dnodeInitVRead, dnodeCleanupVRead},
{"dnode-vwrite", dnodeInitVWrite, dnodeCleanupVWrite},
{"dnode-vmgmt", dnodeInitVMgmt, dnodeCleanupVMgmt},
{"dnode-mread", dnodeInitMRead, dnodeCleanupMRead},
{"dnode-mwrite", dnodeInitMWrite, dnodeCleanupMWrite},
{"dnode-mpeer", dnodeInitMPeer, dnodeCleanupMPeer},
{"dnode-client", dnodeInitClient, dnodeCleanupClient},
{"dnode-server", dnodeInitServer, dnodeCleanupServer},
{"dnode-vnodes", dnodeInitVnodes, dnodeCleanupVnodes},
{"dnode-modules", dnodeInitModules, dnodeCleanupModules},
{"dnode-shell", dnodeInitShell, dnodeCleanupShell},
{"dnode-statustmr", dnodeInitStatusTimer,dnodeCleanupStatusTimer},
{"dnode-telemetry", dnodeInitTelemetry, dnodeCleanupTelemetry},
};
static int dnodeCreateDir(const char *dir) {
@ -87,24 +83,31 @@ static int dnodeCreateDir(const char *dir) {
return 0;
}
static void dnodeCleanupComponents(int32_t stepId) {
for (int32_t i = stepId; i >= 0; i--) {
if (tsDnodeComponents[i].cleanup) {
(*tsDnodeComponents[i].cleanup)();
}
}
static void dnodeCleanupComponents() {
int32_t stepSize = sizeof(tsDnodeSteps) / sizeof(SStep);
dnodeStepCleanup(tsDnodeSteps, stepSize);
}
static int32_t dnodeInitComponents() {
int32_t code = 0;
for (int32_t i = 0; i < sizeof(tsDnodeComponents) / sizeof(tsDnodeComponents[0]); i++) {
if (tsDnodeComponents[i].init() != 0) {
dnodeCleanupComponents(i);
code = -1;
break;
}
int32_t stepSize = sizeof(tsDnodeSteps) / sizeof(SStep);
return dnodeStepInit(tsDnodeSteps, stepSize);
}
static int32_t dnodeInitTmr() {
tsDnodeTmr = taosTmrInit(100, 200, 60000, "DND-DM");
if (tsDnodeTmr == NULL) {
dError("failed to init dnode timer");
return -1;
}
return 0;
}
static void dnodeCleanupTmr() {
if (tsDnodeTmr != NULL) {
taosTmrCleanUp(tsDnodeTmr);
tsDnodeTmr = NULL;
}
return code;
}
int32_t dnodeInitSystem() {
@ -115,6 +118,8 @@ int32_t dnodeInitSystem() {
taosInitGlobalCfg();
taosReadGlobalLogCfg();
taosSetCoreDump();
taosInitNotes();
dnodeInitTmr();
signal(SIGPIPE, SIG_IGN);
if (dnodeCreateDir(tsLogDir) < 0) {
@ -140,7 +145,6 @@ int32_t dnodeInitSystem() {
return -1;
}
dnodeStartModules();
dnodeSetRunStatus(TSDB_RUN_STATUS_RUNING);
dInfo("TDengine is initialized successfully");
@ -151,7 +155,8 @@ int32_t dnodeInitSystem() {
void dnodeCleanUpSystem() {
if (dnodeGetRunStatus() != TSDB_RUN_STATUS_STOPPED) {
dnodeSetRunStatus(TSDB_RUN_STATUS_STOPPED);
dnodeCleanupComponents(sizeof(tsDnodeComponents) / sizeof(tsDnodeComponents[0]) - 1);
dnodeCleanupTmr();
dnodeCleanupComponents();
taos_cleanup();
taosCloseLog();
}

View File

@ -1,572 +0,0 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "cJSON.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "ttimer.h"
#include "tsdb.h"
#include "twal.h"
#include "tqueue.h"
#include "tsync.h"
#include "ttimer.h"
#include "tbalance.h"
#include "tglobal.h"
#include "dnode.h"
#include "vnode.h"
#include "mnode.h"
#include "dnodeInt.h"
#include "dnodeMgmt.h"
#include "dnodeEps.h"
#include "dnodeCfg.h"
#include "dnodeMInfos.h"
#include "dnodeVRead.h"
#include "dnodeVWrite.h"
#include "dnodeModule.h"
typedef struct {
pthread_t thread;
int32_t threadIndex;
int32_t failed;
int32_t opened;
int32_t vnodeNum;
int32_t * vnodeList;
} SOpenVnodeThread;
typedef struct {
SRpcMsg rpcMsg;
char pCont[];
} SMgmtMsg;
void * tsDnodeTmr = NULL;
static void * tsStatusTimer = NULL;
static uint32_t tsRebootTime;
static taos_qset tsMgmtQset = NULL;
static taos_queue tsMgmtQueue = NULL;
static pthread_t tsQthread;
static void dnodeProcessStatusRsp(SRpcMsg *pMsg);
static void dnodeSendStatusMsg(void *handle, void *tmrId);
static void *dnodeProcessMgmtQueue(void *param);
static int32_t dnodeOpenVnodes();
static void dnodeCloseVnodes();
static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg);
static int32_t (*dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *pMsg);
int32_t dnodeInitMgmt() {
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeProcessCreateVnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeProcessAlterVnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeProcessDropVnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeProcessAlterStreamMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeProcessConfigDnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeMsg;
dnodeAddClientRspHandle(TSDB_MSG_TYPE_DM_STATUS_RSP, dnodeProcessStatusRsp);
tsRebootTime = taosGetTimestampSec();
int32_t code = vnodeInitResources();
if (code != TSDB_CODE_SUCCESS) {
dnodeCleanupMgmt();
return -1;
}
// create the queue and thread to handle the message
tsMgmtQset = taosOpenQset();
if (tsMgmtQset == NULL) {
dError("failed to create the mgmt queue set");
dnodeCleanupMgmt();
return -1;
}
tsMgmtQueue = taosOpenQueue();
if (tsMgmtQueue == NULL) {
dError("failed to create the mgmt queue");
dnodeCleanupMgmt();
return -1;
}
taosAddIntoQset(tsMgmtQset, tsMgmtQueue, NULL);
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
code = pthread_create(&tsQthread, &thAttr, dnodeProcessMgmtQueue, NULL);
pthread_attr_destroy(&thAttr);
if (code != 0) {
dError("failed to create thread to process mgmt queue, reason:%s", strerror(errno));
dnodeCleanupMgmt();
return -1;
}
code = dnodeOpenVnodes();
if (code != TSDB_CODE_SUCCESS) {
dnodeCleanupMgmt();
return -1;
}
dInfo("dnode mgmt is initialized");
return TSDB_CODE_SUCCESS;
}
int32_t dnodeInitMgmtTimer() {
tsDnodeTmr = taosTmrInit(100, 200, 60000, "DND-DM");
if (tsDnodeTmr == NULL) {
dError("failed to init dnode timer");
dnodeCleanupMgmt();
return -1;
}
taosTmrReset(dnodeSendStatusMsg, 500, NULL, tsDnodeTmr, &tsStatusTimer);
dInfo("dnode mgmt timer is initialized");
return TSDB_CODE_SUCCESS;
}
void dnodeSendStatusMsgToMnode() {
if (tsDnodeTmr != NULL && tsStatusTimer != NULL) {
dInfo("force send status msg to mnode");
taosTmrReset(dnodeSendStatusMsg, 3, NULL, tsDnodeTmr, &tsStatusTimer);
}
}
void dnodeCleanupMgmtTimer() {
if (tsStatusTimer != NULL) {
taosTmrStopA(&tsStatusTimer);
tsStatusTimer = NULL;
}
if (tsDnodeTmr != NULL) {
taosTmrCleanUp(tsDnodeTmr);
tsDnodeTmr = NULL;
}
}
void dnodeCleanupMgmt() {
dnodeCleanupMgmtTimer();
dnodeCloseVnodes();
if (tsMgmtQset) taosQsetThreadResume(tsMgmtQset);
if (tsQthread) pthread_join(tsQthread, NULL);
if (tsMgmtQueue) taosCloseQueue(tsMgmtQueue);
if (tsMgmtQset) taosCloseQset(tsMgmtQset);
tsMgmtQset = NULL;
tsMgmtQueue = NULL;
vnodeCleanupResources();
}
static int32_t dnodeWriteToMgmtQueue(SRpcMsg *pMsg) {
int32_t size = sizeof(SMgmtMsg) + pMsg->contLen;
SMgmtMsg *pMgmt = taosAllocateQitem(size);
if (pMgmt == NULL) {
return TSDB_CODE_DND_OUT_OF_MEMORY;
}
pMgmt->rpcMsg = *pMsg;
pMgmt->rpcMsg.pCont = pMgmt->pCont;
memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen);
taosWriteQitem(tsMgmtQueue, TAOS_QTYPE_RPC, pMgmt);
return TSDB_CODE_SUCCESS;
}
void dnodeDispatchToMgmtQueue(SRpcMsg *pMsg) {
int32_t code = dnodeWriteToMgmtQueue(pMsg);
if (code != TSDB_CODE_SUCCESS) {
SRpcMsg rsp = {.handle = pMsg->handle, .code = code};
rpcSendResponse(&rsp);
}
rpcFreeCont(pMsg->pCont);
}
static void *dnodeProcessMgmtQueue(void *param) {
SMgmtMsg *pMgmt;
SRpcMsg * pMsg;
SRpcMsg rsp = {0};
int32_t qtype;
void * handle;
while (1) {
if (taosReadQitemFromQset(tsMgmtQset, &qtype, (void **)&pMgmt, &handle) == 0) {
dDebug("qset:%p, dnode mgmt got no message from qset, exit", tsMgmtQset);
break;
}
pMsg = &pMgmt->rpcMsg;
dDebug("msg:%p, ahandle:%p type:%s will be processed", pMgmt, pMsg->ahandle, taosMsg[pMsg->msgType]);
if (dnodeProcessMgmtMsgFp[pMsg->msgType]) {
rsp.code = (*dnodeProcessMgmtMsgFp[pMsg->msgType])(pMsg);
} else {
rsp.code = TSDB_CODE_DND_MSG_NOT_PROCESSED;
}
rsp.handle = pMsg->handle;
rsp.pCont = NULL;
rpcSendResponse(&rsp);
taosFreeQitem(pMsg);
}
return NULL;
}
static int32_t dnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) {
DIR *dir = opendir(tsVnodeDir);
if (dir == NULL) {
return TSDB_CODE_DND_NO_WRITE_ACCESS;
}
*numOfVnodes = 0;
struct dirent *de = NULL;
while ((de = readdir(dir)) != NULL) {
if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
if (de->d_type & DT_DIR) {
if (strncmp("vnode", de->d_name, 5) != 0) continue;
int32_t vnode = atoi(de->d_name + 5);
if (vnode == 0) continue;
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
dError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
continue;
} else {
vnodeList[*numOfVnodes - 1] = vnode;
}
}
}
closedir(dir);
return TSDB_CODE_SUCCESS;
}
static void *dnodeOpenVnode(void *param) {
SOpenVnodeThread *pThread = param;
char vnodeDir[TSDB_FILENAME_LEN * 3];
dDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
int32_t vgId = pThread->vnodeList[v];
snprintf(vnodeDir, TSDB_FILENAME_LEN * 3, "%s/vnode%d", tsVnodeDir, vgId);
if (vnodeOpen(vgId, vnodeDir) < 0) {
dError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
pThread->failed++;
} else {
dDebug("vgId:%d, is openned by thread:%d", vgId, pThread->threadIndex);
pThread->opened++;
}
}
dDebug("thread:%d, total vnodes:%d, openned:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
pThread->failed);
return NULL;
}
static int32_t dnodeOpenVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t status = dnodeGetVnodeList(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
dInfo("get dnode list failed");
return status;
}
int32_t threadNum = tsNumOfCores;
int32_t vnodesPerThread = numOfVnodes / threadNum + 1;
SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread));
for (int32_t t = 0; t < threadNum; ++t) {
threads[t].threadIndex = t;
threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t));
}
for (int32_t v = 0; v < numOfVnodes; ++v) {
int32_t t = v % threadNum;
SOpenVnodeThread *pThread = &threads[t];
pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v];
}
dDebug("start %d threads to open %d vnodes", threadNum, numOfVnodes);
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
if (pThread->vnodeNum == 0) continue;
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&pThread->thread, &thAttr, dnodeOpenVnode, pThread) != 0) {
dError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
}
pthread_attr_destroy(&thAttr);
}
int32_t openVnodes = 0;
int32_t failedVnodes = 0;
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
if (pThread->vnodeNum > 0 && pThread->thread) {
pthread_join(pThread->thread, NULL);
}
openVnodes += pThread->opened;
failedVnodes += pThread->failed;
free(pThread->vnodeList);
}
free(threads);
dInfo("there are total vnodes:%d, openned:%d failed:%d", numOfVnodes, openVnodes, failedVnodes);
return TSDB_CODE_SUCCESS;
}
static void dnodeCloseVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES]= {0};
int32_t numOfVnodes = 0;
int32_t status;
status = vnodeGetVnodeList(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
dInfo("get dnode list failed");
return;
}
for (int32_t i = 0; i < numOfVnodes; ++i) {
vnodeClose(vnodeList[i]);
}
dInfo("total vnodes:%d are all closed", numOfVnodes);
}
static void* dnodeParseVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = rpcMsg->pCont;
pCreate->cfg.vgId = htonl(pCreate->cfg.vgId);
pCreate->cfg.cfgVersion = htonl(pCreate->cfg.cfgVersion);
pCreate->cfg.maxTables = htonl(pCreate->cfg.maxTables);
pCreate->cfg.cacheBlockSize = htonl(pCreate->cfg.cacheBlockSize);
pCreate->cfg.totalBlocks = htonl(pCreate->cfg.totalBlocks);
pCreate->cfg.daysPerFile = htonl(pCreate->cfg.daysPerFile);
pCreate->cfg.daysToKeep1 = htonl(pCreate->cfg.daysToKeep1);
pCreate->cfg.daysToKeep2 = htonl(pCreate->cfg.daysToKeep2);
pCreate->cfg.daysToKeep = htonl(pCreate->cfg.daysToKeep);
pCreate->cfg.minRowsPerFileBlock = htonl(pCreate->cfg.minRowsPerFileBlock);
pCreate->cfg.maxRowsPerFileBlock = htonl(pCreate->cfg.maxRowsPerFileBlock);
pCreate->cfg.fsyncPeriod = htonl(pCreate->cfg.fsyncPeriod);
pCreate->cfg.commitTime = htonl(pCreate->cfg.commitTime);
for (int32_t j = 0; j < pCreate->cfg.replications; ++j) {
pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId);
}
return pCreate;
}
static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = dnodeParseVnodeMsg(rpcMsg);
void *pVnode = vnodeAcquire(pCreate->cfg.vgId);
if (pVnode != NULL) {
dDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId);
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
} else {
dDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId);
return vnodeCreate(pCreate);
}
}
static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) {
SAlterVnodeMsg *pAlter = dnodeParseVnodeMsg(rpcMsg);
void *pVnode = vnodeAcquire(pAlter->cfg.vgId);
if (pVnode != NULL) {
dDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId);
int32_t code = vnodeAlter(pVnode, pAlter);
vnodeRelease(pVnode);
return code;
} else {
dError("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
}
static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) {
SDropVnodeMsg *pDrop = rpcMsg->pCont;
pDrop->vgId = htonl(pDrop->vgId);
return vnodeDrop(pDrop->vgId);
}
static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg) {
// SAlterStreamMsg *pStream = pCont;
// pStream->uid = htobe64(pStream->uid);
// pStream->stime = htobe64(pStream->stime);
// pStream->vnode = htonl(pStream->vnode);
// pStream->sid = htonl(pStream->sid);
// pStream->status = htonl(pStream->status);
//
// int32_t code = dnodeCreateStream(pStream);
return 0;
}
static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg) {
SCfgDnodeMsg *pCfg = pMsg->pCont;
return taosCfgDynamicOptions(pCfg->config);
}
static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) {
SCreateMnodeMsg *pCfg = pMsg->pCont;
pCfg->dnodeId = htonl(pCfg->dnodeId);
if (pCfg->dnodeId != dnodeGetDnodeId()) {
dDebug("dnodeId:%d, in create mnode msg is not equal with saved dnodeId:%d", pCfg->dnodeId, dnodeGetDnodeId());
return TSDB_CODE_MND_DNODE_ID_NOT_CONFIGURED;
}
if (strcmp(pCfg->dnodeEp, tsLocalEp) != 0) {
dDebug("dnodeEp:%s, in create mnode msg is not equal with saved dnodeEp:%s", pCfg->dnodeEp, tsLocalEp);
return TSDB_CODE_MND_DNODE_EP_NOT_CONFIGURED;
}
dDebug("dnodeId:%d, create mnode msg is received from mnodes, numOfMnodes:%d", pCfg->dnodeId, pCfg->mnodes.mnodeNum);
for (int i = 0; i < pCfg->mnodes.mnodeNum; ++i) {
pCfg->mnodes.mnodeInfos[i].mnodeId = htonl(pCfg->mnodes.mnodeInfos[i].mnodeId);
dDebug("mnode index:%d, mnode:%d:%s", i, pCfg->mnodes.mnodeInfos[i].mnodeId, pCfg->mnodes.mnodeInfos[i].mnodeEp);
}
dnodeStartMnode(&pCfg->mnodes);
return TSDB_CODE_SUCCESS;
}
static void dnodeProcessStatusRsp(SRpcMsg *pMsg) {
if (pMsg->code != TSDB_CODE_SUCCESS) {
dError("status rsp is received, error:%s", tstrerror(pMsg->code));
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
return;
}
SStatusRsp *pStatusRsp = pMsg->pCont;
SMnodeInfos *minfos = &pStatusRsp->mnodes;
dnodeUpdateMInfos(minfos);
SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg;
pCfg->numOfVnodes = htonl(pCfg->numOfVnodes);
pCfg->moduleStatus = htonl(pCfg->moduleStatus);
pCfg->dnodeId = htonl(pCfg->dnodeId);
dnodeUpdateCfg(pCfg);
vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes);
SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess));
dnodeUpdateEps(pEps);
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
}
static void dnodeSendStatusMsg(void *handle, void *tmrId) {
if (tsDnodeTmr == NULL) {
dError("dnode timer is already released");
return;
}
if (tsStatusTimer == NULL) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
dError("failed to start status timer");
return;
}
int32_t contLen = sizeof(SStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad);
SStatusMsg *pStatus = rpcMallocCont(contLen);
if (pStatus == NULL) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
dError("failed to malloc status message");
return;
}
dnodeGetCfg(&pStatus->dnodeId, pStatus->clusterId);
pStatus->dnodeId = htonl(dnodeGetDnodeId());
pStatus->version = htonl(tsVersion);
pStatus->lastReboot = htonl(tsRebootTime);
pStatus->numOfCores = htons((uint16_t) tsNumOfCores);
pStatus->diskAvailable = tsAvailDataDirGB;
pStatus->alternativeRole = (uint8_t) tsAlternativeRole;
tstrncpy(pStatus->dnodeEp, tsLocalEp, TSDB_EP_LEN);
// fill cluster cfg parameters
pStatus->clusterCfg.numOfMnodes = htonl(tsNumOfMnodes);
pStatus->clusterCfg.enableBalance = htonl(tsEnableBalance);
pStatus->clusterCfg.mnodeEqualVnodeNum = htonl(tsMnodeEqualVnodeNum);
pStatus->clusterCfg.offlineThreshold = htonl(tsOfflineThreshold);
pStatus->clusterCfg.statusInterval = htonl(tsStatusInterval);
pStatus->clusterCfg.maxtablesPerVnode = htonl(tsMaxTablePerVnode);
pStatus->clusterCfg.maxVgroupsPerDb = htonl(tsMaxVgroupsPerDb);
tstrncpy(pStatus->clusterCfg.arbitrator, tsArbitrator, TSDB_EP_LEN);
tstrncpy(pStatus->clusterCfg.timezone, tsTimezone, 64);
pStatus->clusterCfg.checkTime = 0;
char timestr[32] = "1970-01-01 00:00:00.00";
(void)taosParseTime(timestr, &pStatus->clusterCfg.checkTime, strlen(timestr), TSDB_TIME_PRECISION_MILLI, 0);
tstrncpy(pStatus->clusterCfg.locale, tsLocale, TSDB_LOCALE_LEN);
tstrncpy(pStatus->clusterCfg.charset, tsCharset, TSDB_LOCALE_LEN);
vnodeBuildStatusMsg(pStatus);
contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad);
pStatus->openVnodes = htons(pStatus->openVnodes);
SRpcMsg rpcMsg = {
.pCont = pStatus,
.contLen = contLen,
.msgType = TSDB_MSG_TYPE_DM_STATUS
};
SRpcEpSet epSet;
dnodeGetEpSetForPeer(&epSet);
dnodeSendMsgToDnode(&epSet, &rpcMsg);
}
void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) {
SRpcConnInfo connInfo = {0};
rpcGetConnInfo(rpcMsg->handle, &connInfo);
SRpcEpSet epSet = {0};
if (forShell) {
dnodeGetEpSetForShell(&epSet);
} else {
dnodeGetEpSetForPeer(&epSet);
}
dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType],
taosIpStr(connInfo.clientIp), connInfo.user, epSet.numOfEps, epSet.inUse);
for (int i = 0; i < epSet.numOfEps; ++i) {
dDebug("mnode index:%d %s:%d", i, epSet.fqdn[i], epSet.port[i]);
epSet.port[i] = htons(epSet.port[i]);
}
rpcSendRedirectRsp(rpcMsg->handle, &epSet);
}

View File

@ -15,15 +15,10 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taosdef.h"
#include "taosmsg.h"
#include "tglobal.h"
#include "mnode.h"
#include "http.h"
#include "tmqtt.h"
#include "monitor.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeModule.h"
typedef struct {
@ -78,10 +73,10 @@ static void dnodeAllocModules() {
tsModule[TSDB_MOD_MONITOR].enable = (tsEnableMonitorModule == 1);
tsModule[TSDB_MOD_MONITOR].name = "monitor";
tsModule[TSDB_MOD_MONITOR].initFp = monitorInitSystem;
tsModule[TSDB_MOD_MONITOR].cleanUpFp = monitorCleanUpSystem;
tsModule[TSDB_MOD_MONITOR].startFp = monitorStartSystem;
tsModule[TSDB_MOD_MONITOR].stopFp = monitorStopSystem;
tsModule[TSDB_MOD_MONITOR].initFp = monInitSystem;
tsModule[TSDB_MOD_MONITOR].cleanUpFp = monCleanupSystem;
tsModule[TSDB_MOD_MONITOR].startFp = monStartSystem;
tsModule[TSDB_MOD_MONITOR].stopFp = monStopSystem;
if (tsEnableMonitorModule) {
dnodeSetModuleStatus(TSDB_MOD_MONITOR);
}
@ -102,6 +97,20 @@ void dnodeCleanupModules() {
}
}
static int32_t dnodeStartModules() {
for (EModuleType module = 1; module < TSDB_MOD_MAX; ++module) {
if (tsModule[module].enable && tsModule[module].startFp) {
int32_t code = (*tsModule[module].startFp)();
if (code != 0) {
dError("failed to start module:%s, code:%d", tsModule[module].name, code);
return code;
}
}
}
return 0;
}
int32_t dnodeInitModules() {
dnodeAllocModules();
@ -115,17 +124,7 @@ int32_t dnodeInitModules() {
}
dInfo("dnode modules is initialized");
return 0;
}
void dnodeStartModules() {
for (EModuleType module = 1; module < TSDB_MOD_MAX; ++module) {
if (tsModule[module].enable && tsModule[module].startFp) {
if ((*tsModule[module].startFp)() != 0) {
dError("failed to start module:%s", tsModule[module].name);
}
}
}
return dnodeStartModules();
}
void dnodeProcessModuleStatus(uint32_t moduleStatus) {
@ -147,8 +146,8 @@ void dnodeProcessModuleStatus(uint32_t moduleStatus) {
}
}
bool dnodeStartMnode(SMnodeInfos *minfos) {
SMnodeInfos *mnodes = minfos;
bool dnodeStartMnode(SMInfos *pMinfos) {
SMInfos *pMnodes = pMinfos;
if (tsModuleStatus & (1 << TSDB_MOD_MNODE)) {
dDebug("mnode module is already started, module status:%d", tsModuleStatus);
@ -159,7 +158,7 @@ bool dnodeStartMnode(SMnodeInfos *minfos) {
dInfo("start mnode module, module status:%d, new status:%d", tsModuleStatus, moduleStatus);
dnodeProcessModuleStatus(moduleStatus);
sdbUpdateSync(mnodes);
sdbUpdateSync(pMnodes);
return true;
}

View File

@ -21,15 +21,12 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taosmsg.h"
#include "tglobal.h"
#include "mnode.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeMgmt.h"
#include "dnodeVMgmt.h"
#include "dnodeVWrite.h"
#include "dnodeMPeer.h"
#include "dnodeMInfos.h"
#include "dnodeStep.h"
static void (*dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
static void dnodeProcessReqMsgFromDnode(SRpcMsg *pMsg, SRpcEpSet *);
@ -44,19 +41,19 @@ int32_t dnodeInitServer() {
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = dnodeDispatchToVWriteQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = dnodeDispatchToVWriteQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeDispatchToMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeDispatchToMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeDispatchToMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeDispatchToMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeDispatchToMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeDispatchToMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeDispatchToVMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeDispatchToVMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeDispatchToVMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeDispatchToVMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeDispatchToVMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeDispatchToVMgmtQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_CONFIG_TABLE] = dnodeDispatchToMPeerQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_CONFIG_VNODE] = dnodeDispatchToMPeerQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_AUTH] = dnodeDispatchToMPeerQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_GRANT] = dnodeDispatchToMPeerQueue;
dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_STATUS] = dnodeDispatchToMPeerQueue;
SRpcInit rpcInit;
memset(&rpcInit, 0, sizeof(rpcInit));
rpcInit.localPort = tsDnodeDnodePort;
@ -91,14 +88,15 @@ static void dnodeProcessReqMsgFromDnode(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
.pCont = NULL,
.contLen = 0
};
if (pMsg->pCont == NULL) return;
if (pMsg->msgType == TSDB_MSG_TYPE_NETWORK_TEST) return dnodeSendStartupStep(pMsg);
if (dnodeGetRunStatus() != TSDB_RUN_STATUS_RUNING) {
rspMsg.code = TSDB_CODE_APP_NOT_READY;
rpcSendResponse(&rspMsg);
rpcFreeCont(pMsg->pCont);
dDebug("RPC %p, msg:%s is ignored since dnode not running", pMsg->handle, taosMsg[pMsg->msgType]);
dTrace("RPC %p, msg:%s is ignored since dnode not running", pMsg->handle, taosMsg[pMsg->msgType]);
return;
}
@ -153,7 +151,7 @@ void dnodeCleanupClient() {
static void dnodeProcessRspFromDnode(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
if (dnodeGetRunStatus() == TSDB_RUN_STATUS_STOPPED) {
if (pMsg == NULL || pMsg->pCont == NULL) return;
dDebug("msg:%p is ignored since dnode is stopping", pMsg);
dTrace("msg:%p is ignored since dnode is stopping", pMsg);
rpcFreeCont(pMsg->pCont);
return;
}
@ -182,6 +180,8 @@ void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg) {
void dnodeSendMsgToMnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp) {
SRpcEpSet epSet = {0};
dnodeGetEpSetForPeer(&epSet);
assert(tsClientRpc != 0);
rpcSendRecv(tsClientRpc, &epSet, rpcMsg, rpcRsp);
}

View File

@ -15,20 +15,14 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosdef.h"
#include "taosmsg.h"
#include "tglobal.h"
#include "tutil.h"
#include "http.h"
#include "mnode.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeVRead.h"
#include "dnodeVWrite.h"
#include "dnodeMRead.h"
#include "dnodeMWrite.h"
#include "dnodeShell.h"
#include "dnodeStep.h"
static void (*dnodeProcessShellMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *);
@ -74,6 +68,8 @@ int32_t dnodeInitShell() {
dnodeProcessShellMsgFp[TSDB_MSG_TYPE_CM_SHOW] = dnodeDispatchToMReadQueue;
dnodeProcessShellMsgFp[TSDB_MSG_TYPE_CM_RETRIEVE] = dnodeDispatchToMReadQueue;
dnodeProcessShellMsgFp[TSDB_MSG_TYPE_NETWORK_TEST] = dnodeSendStartupStep;
int32_t numOfThreads = tsNumOfCores * tsNumOfThreadsPerCore;
numOfThreads = (int32_t) ((1.0 - tsRatioOfQueryThreads) * numOfThreads / 2.0);
if (numOfThreads < 1) {
@ -142,7 +138,23 @@ static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
}
}
static int32_t dnodeAuthNettestUser(char *user, char *spi, char *encrypt, char *secret, char *ckey) {
if (strcmp(user, "nettestinternal") == 0) {
char pass[32] = {0};
taosEncryptPass((uint8_t *)user, strlen(user), pass);
*spi = 0;
*encrypt = 0;
*ckey = 0;
memcpy(secret, pass, TSDB_KEY_LEN);
dTrace("nettest user is authorized");
return 0;
}
return -1;
}
static int dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char *secret, char *ckey) {
if (dnodeAuthNettestUser(user, spi, encrypt, secret, ckey) == 0) return 0;
int code = mnodeRetriveAuth(user, spi, encrypt, secret, ckey);
if (code != TSDB_CODE_APP_NOT_READY) return code;
@ -220,4 +232,4 @@ SStatisInfo dnodeGetStatisInfo() {
}
return info;
}
}

73
src/dnode/src/dnodeStep.c Normal file
View File

@ -0,0 +1,73 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "dnodeStep.h"
static SStartupStep tsStartupStep;
void dnodeReportStep(char *name, char *desc, int8_t finished) {
tstrncpy(tsStartupStep.name, name, sizeof(tsStartupStep.name));
tstrncpy(tsStartupStep.desc, desc, sizeof(tsStartupStep.desc));
tsStartupStep.finished = finished;
}
void dnodeSendStartupStep(SRpcMsg *pMsg) {
dInfo("nettest msg is received, cont:%s", (char *)pMsg->pCont);
SStartupStep *pStep = rpcMallocCont(sizeof(SStartupStep));
memcpy(pStep, &tsStartupStep, sizeof(SStartupStep));
dDebug("startup msg is sent, step:%s desc:%s finished:%d", pStep->name, pStep->desc, pStep->finished);
SRpcMsg rpcRsp = {.handle = pMsg->handle, .pCont = pStep, .contLen = sizeof(SStartupStep)};
rpcSendResponse(&rpcRsp);
rpcFreeCont(pMsg->pCont);
}
void taosStepCleanupImp(SStep *pSteps, int32_t stepId) {
for (int32_t step = stepId; step >= 0; step--) {
SStep *pStep = pSteps + step;
dDebug("step:%s will cleanup", pStep->name);
if (pStep->cleanupFp != NULL) {
(*pStep->cleanupFp)();
}
}
}
int32_t dnodeStepInit(SStep *pSteps, int32_t stepSize) {
for (int32_t step = 0; step < stepSize; step++) {
SStep *pStep = pSteps + step;
if (pStep->initFp == NULL) continue;
dnodeReportStep(pStep->name, "Start initialization", 0);
int32_t code = (*pStep->initFp)();
if (code != 0) {
dDebug("step:%s will init", pStep->name);
taosStepCleanupImp(pSteps, step);
return code;
}
dnodeReportStep(pStep->name, "Initialization complete", step + 1 >= stepSize);
}
return 0;
}
void dnodeStepCleanup(SStep *pSteps, int32_t stepSize) {
return taosStepCleanupImp(pSteps, stepSize - 1);
}

View File

@ -16,9 +16,6 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "tgrant.h"
#include "tutil.h"
#include "tglobal.h"
#include "dnodeInt.h"
#include "dnodeMain.h"
static void signal_handler(int32_t signum, siginfo_t *sigInfo, void *context);

View File

@ -15,9 +15,6 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "tglobal.h"
#include "tutil.h"
#include "osTime.h"
#include "tsocket.h"
#include "tbuffer.h"
@ -32,8 +29,6 @@
#include "mnodeTable.h"
#include "mnodeSdb.h"
#include "mnodeAcct.h"
#include "dnode.h"
#include "dnodeInt.h"
#include "dnodeTelemetry.h"
static tsem_t tsExitSem;
@ -313,4 +308,4 @@ void dnodeCleanupTelemetry() {
pthread_join(tsTelemetryThread, NULL);
tsem_destroy(&tsExitSem);
}
}
}

239
src/dnode/src/dnodeVMgmt.c Normal file
View File

@ -0,0 +1,239 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "tqueue.h"
#include "dnodeVMgmt.h"
typedef struct {
SRpcMsg rpcMsg;
char pCont[];
} SMgmtMsg;
static taos_qset tsMgmtQset = NULL;
static taos_queue tsMgmtQueue = NULL;
static pthread_t tsQthread;
static void * dnodeProcessMgmtQueue(void *param);
static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg);
static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg);
static int32_t (*dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *pMsg);
int32_t dnodeInitVMgmt() {
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeProcessCreateVnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeProcessAlterVnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeProcessDropVnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeProcessAlterStreamMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeProcessConfigDnodeMsg;
dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeMsg;
int32_t code = vnodeInitMgmt();
if (code != TSDB_CODE_SUCCESS) return -1;
tsMgmtQset = taosOpenQset();
if (tsMgmtQset == NULL) {
dError("failed to create the vmgmt queue set");
return -1;
}
tsMgmtQueue = taosOpenQueue();
if (tsMgmtQueue == NULL) {
dError("failed to create the vmgmt queue");
return -1;
}
taosAddIntoQset(tsMgmtQset, tsMgmtQueue, NULL);
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
code = pthread_create(&tsQthread, &thAttr, dnodeProcessMgmtQueue, NULL);
pthread_attr_destroy(&thAttr);
if (code != 0) {
dError("failed to create thread to process vmgmt queue, reason:%s", strerror(errno));
return -1;
}
dInfo("dnode vmgmt is initialized");
return TSDB_CODE_SUCCESS;
}
void dnodeCleanupVMgmt() {
if (tsMgmtQset) taosQsetThreadResume(tsMgmtQset);
if (tsQthread) pthread_join(tsQthread, NULL);
if (tsMgmtQueue) taosCloseQueue(tsMgmtQueue);
if (tsMgmtQset) taosCloseQset(tsMgmtQset);
tsMgmtQset = NULL;
tsMgmtQueue = NULL;
vnodeCleanupMgmt();
}
static int32_t dnodeWriteToMgmtQueue(SRpcMsg *pMsg) {
int32_t size = sizeof(SMgmtMsg) + pMsg->contLen;
SMgmtMsg *pMgmt = taosAllocateQitem(size);
if (pMgmt == NULL) return TSDB_CODE_DND_OUT_OF_MEMORY;
pMgmt->rpcMsg = *pMsg;
pMgmt->rpcMsg.pCont = pMgmt->pCont;
memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen);
taosWriteQitem(tsMgmtQueue, TAOS_QTYPE_RPC, pMgmt);
return TSDB_CODE_SUCCESS;
}
void dnodeDispatchToVMgmtQueue(SRpcMsg *pMsg) {
int32_t code = dnodeWriteToMgmtQueue(pMsg);
if (code != TSDB_CODE_SUCCESS) {
SRpcMsg rsp = {.handle = pMsg->handle, .code = code};
rpcSendResponse(&rsp);
}
rpcFreeCont(pMsg->pCont);
}
static void *dnodeProcessMgmtQueue(void *param) {
SMgmtMsg *pMgmt;
SRpcMsg * pMsg;
SRpcMsg rsp = {0};
int32_t qtype;
void * handle;
while (1) {
if (taosReadQitemFromQset(tsMgmtQset, &qtype, (void **)&pMgmt, &handle) == 0) {
dDebug("qset:%p, dnode mgmt got no message from qset, exit", tsMgmtQset);
break;
}
pMsg = &pMgmt->rpcMsg;
dTrace("msg:%p, ahandle:%p type:%s will be processed", pMgmt, pMsg->ahandle, taosMsg[pMsg->msgType]);
if (dnodeProcessMgmtMsgFp[pMsg->msgType]) {
rsp.code = (*dnodeProcessMgmtMsgFp[pMsg->msgType])(pMsg);
} else {
rsp.code = TSDB_CODE_DND_MSG_NOT_PROCESSED;
}
dTrace("msg:%p, is processed, code:0x%x", pMgmt, rsp.code);
if (rsp.code != TSDB_CODE_DND_ACTION_IN_PROGRESS) {
rsp.handle = pMsg->handle;
rsp.pCont = NULL;
rpcSendResponse(&rsp);
}
taosFreeQitem(pMsg);
}
return NULL;
}
static SCreateVnodeMsg* dnodeParseVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = rpcMsg->pCont;
pCreate->cfg.vgId = htonl(pCreate->cfg.vgId);
pCreate->cfg.cfgVersion = htonl(pCreate->cfg.cfgVersion);
pCreate->cfg.maxTables = htonl(pCreate->cfg.maxTables);
pCreate->cfg.cacheBlockSize = htonl(pCreate->cfg.cacheBlockSize);
pCreate->cfg.totalBlocks = htonl(pCreate->cfg.totalBlocks);
pCreate->cfg.daysPerFile = htonl(pCreate->cfg.daysPerFile);
pCreate->cfg.daysToKeep1 = htonl(pCreate->cfg.daysToKeep1);
pCreate->cfg.daysToKeep2 = htonl(pCreate->cfg.daysToKeep2);
pCreate->cfg.daysToKeep = htonl(pCreate->cfg.daysToKeep);
pCreate->cfg.minRowsPerFileBlock = htonl(pCreate->cfg.minRowsPerFileBlock);
pCreate->cfg.maxRowsPerFileBlock = htonl(pCreate->cfg.maxRowsPerFileBlock);
pCreate->cfg.fsyncPeriod = htonl(pCreate->cfg.fsyncPeriod);
pCreate->cfg.commitTime = htonl(pCreate->cfg.commitTime);
for (int32_t j = 0; j < pCreate->cfg.replications; ++j) {
pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId);
}
return pCreate;
}
static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = dnodeParseVnodeMsg(rpcMsg);
void *pVnode = vnodeAcquire(pCreate->cfg.vgId);
if (pVnode != NULL) {
dDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId);
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
} else {
dDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId);
return vnodeCreate(pCreate);
}
}
static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) {
SAlterVnodeMsg *pAlter = dnodeParseVnodeMsg(rpcMsg);
void *pVnode = vnodeAcquire(pAlter->cfg.vgId);
if (pVnode != NULL) {
dDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId);
int32_t code = vnodeAlter(pVnode, pAlter);
vnodeRelease(pVnode);
return code;
} else {
dError("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
}
static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) {
SDropVnodeMsg *pDrop = rpcMsg->pCont;
pDrop->vgId = htonl(pDrop->vgId);
return vnodeDrop(pDrop->vgId);
}
static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg) {
return 0;
}
static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg) {
SCfgDnodeMsg *pCfg = pMsg->pCont;
return taosCfgDynamicOptions(pCfg->config);
}
static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) {
SCreateMnodeMsg *pCfg = pMsg->pCont;
pCfg->dnodeId = htonl(pCfg->dnodeId);
if (pCfg->dnodeId != dnodeGetDnodeId()) {
dDebug("dnodeId:%d, in create mnode msg is not equal with saved dnodeId:%d", pCfg->dnodeId, dnodeGetDnodeId());
return TSDB_CODE_MND_DNODE_ID_NOT_CONFIGURED;
}
if (strcmp(pCfg->dnodeEp, tsLocalEp) != 0) {
dDebug("dnodeEp:%s, in create mnode msg is not equal with saved dnodeEp:%s", pCfg->dnodeEp, tsLocalEp);
return TSDB_CODE_MND_DNODE_EP_NOT_CONFIGURED;
}
dDebug("dnodeId:%d, create mnode msg is received from mnodes, numOfMnodes:%d", pCfg->dnodeId, pCfg->mnodes.mnodeNum);
for (int i = 0; i < pCfg->mnodes.mnodeNum; ++i) {
pCfg->mnodes.mnodeInfos[i].mnodeId = htonl(pCfg->mnodes.mnodeInfos[i].mnodeId);
dDebug("mnode index:%d, mnode:%d:%s", i, pCfg->mnodes.mnodeInfos[i].mnodeId, pCfg->mnodes.mnodeInfos[i].mnodeEp);
}
dnodeStartMnode(&pCfg->mnodes);
return TSDB_CODE_SUCCESS;
}

View File

@ -15,12 +15,8 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tglobal.h"
#include "tqueue.h"
#include "vnode.h"
#include "dnodeInt.h"
#include "dnodeVRead.h"
typedef struct {
pthread_t thread; // thread

View File

@ -15,13 +15,8 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tglobal.h"
#include "tqueue.h"
#include "twal.h"
#include "vnode.h"
#include "dnodeInt.h"
#include "dnodeVWrite.h"
typedef struct {
taos_qall qall;

278
src/dnode/src/dnodeVnodes.c Normal file
View File

@ -0,0 +1,278 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "ttimer.h"
#include "dnodeEps.h"
#include "dnodeCfg.h"
#include "dnodeMInfos.h"
#include "dnodeVnodes.h"
typedef struct {
pthread_t thread;
int32_t threadIndex;
int32_t failed;
int32_t opened;
int32_t vnodeNum;
int32_t * vnodeList;
} SOpenVnodeThread;
extern void * tsDnodeTmr;
static void * tsStatusTimer = NULL;
static uint32_t tsRebootTime = 0;
static void dnodeSendStatusMsg(void *handle, void *tmrId);
static void dnodeProcessStatusRsp(SRpcMsg *pMsg);
int32_t dnodeInitStatusTimer() {
dnodeAddClientRspHandle(TSDB_MSG_TYPE_DM_STATUS_RSP, dnodeProcessStatusRsp);
tsRebootTime = taosGetTimestampSec();
taosTmrReset(dnodeSendStatusMsg, 500, NULL, tsDnodeTmr, &tsStatusTimer);
dInfo("dnode status timer is initialized");
return TSDB_CODE_SUCCESS;
}
void dnodeCleanupStatusTimer() {
if (tsStatusTimer != NULL) {
taosTmrStopA(&tsStatusTimer);
tsStatusTimer = NULL;
}
}
static int32_t dnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) {
DIR *dir = opendir(tsVnodeDir);
if (dir == NULL) return TSDB_CODE_DND_NO_WRITE_ACCESS;
*numOfVnodes = 0;
struct dirent *de = NULL;
while ((de = readdir(dir)) != NULL) {
if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
if (de->d_type & DT_DIR) {
if (strncmp("vnode", de->d_name, 5) != 0) continue;
int32_t vnode = atoi(de->d_name + 5);
if (vnode == 0) continue;
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
dError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
continue;
} else {
vnodeList[*numOfVnodes - 1] = vnode;
}
}
}
closedir(dir);
return TSDB_CODE_SUCCESS;
}
static void *dnodeOpenVnode(void *param) {
SOpenVnodeThread *pThread = param;
dDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
int32_t vgId = pThread->vnodeList[v];
if (vnodeOpen(vgId) < 0) {
dError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
pThread->failed++;
} else {
dDebug("vgId:%d, is openned by thread:%d", vgId, pThread->threadIndex);
pThread->opened++;
}
}
dDebug("thread:%d, total vnodes:%d, openned:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
pThread->failed);
return NULL;
}
int32_t dnodeInitVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t status = dnodeGetVnodeList(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
dInfo("get dnode list failed");
return status;
}
int32_t threadNum = tsNumOfCores;
int32_t vnodesPerThread = numOfVnodes / threadNum + 1;
SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread));
for (int32_t t = 0; t < threadNum; ++t) {
threads[t].threadIndex = t;
threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t));
}
for (int32_t v = 0; v < numOfVnodes; ++v) {
int32_t t = v % threadNum;
SOpenVnodeThread *pThread = &threads[t];
pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v];
}
dDebug("start %d threads to open %d vnodes", threadNum, numOfVnodes);
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
if (pThread->vnodeNum == 0) continue;
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&pThread->thread, &thAttr, dnodeOpenVnode, pThread) != 0) {
dError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
}
pthread_attr_destroy(&thAttr);
}
int32_t openVnodes = 0;
int32_t failedVnodes = 0;
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
if (pThread->vnodeNum > 0 && pThread->thread) {
pthread_join(pThread->thread, NULL);
}
openVnodes += pThread->opened;
failedVnodes += pThread->failed;
free(pThread->vnodeList);
}
free(threads);
dInfo("there are total vnodes:%d, openned:%d", numOfVnodes, openVnodes);
if (failedVnodes != 0) {
dError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes);
return -1;
}
return TSDB_CODE_SUCCESS;
}
void dnodeCleanupVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES]= {0};
int32_t numOfVnodes = 0;
int32_t status;
status = vnodeGetVnodeList(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
dInfo("get dnode list failed");
return;
}
for (int32_t i = 0; i < numOfVnodes; ++i) {
vnodeClose(vnodeList[i]);
}
dInfo("total vnodes:%d are all closed", numOfVnodes);
}
static void dnodeProcessStatusRsp(SRpcMsg *pMsg) {
if (pMsg->code != TSDB_CODE_SUCCESS) {
dError("status rsp is received, error:%s", tstrerror(pMsg->code));
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
return;
}
SStatusRsp *pStatusRsp = pMsg->pCont;
SMInfos *minfos = &pStatusRsp->mnodes;
dnodeUpdateMInfos(minfos);
SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg;
pCfg->numOfVnodes = htonl(pCfg->numOfVnodes);
pCfg->moduleStatus = htonl(pCfg->moduleStatus);
pCfg->dnodeId = htonl(pCfg->dnodeId);
dnodeUpdateCfg(pCfg);
vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes);
SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess));
dnodeUpdateEps(pEps);
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
}
static void dnodeSendStatusMsg(void *handle, void *tmrId) {
if (tsDnodeTmr == NULL) {
dError("dnode timer is already released");
return;
}
if (tsStatusTimer == NULL) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
dError("failed to start status timer");
return;
}
int32_t contLen = sizeof(SStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad);
SStatusMsg *pStatus = rpcMallocCont(contLen);
if (pStatus == NULL) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
dError("failed to malloc status message");
return;
}
dnodeGetCfg(&pStatus->dnodeId, pStatus->clusterId);
pStatus->dnodeId = htonl(dnodeGetDnodeId());
pStatus->version = htonl(tsVersion);
pStatus->lastReboot = htonl(tsRebootTime);
pStatus->numOfCores = htons((uint16_t) tsNumOfCores);
pStatus->diskAvailable = tsAvailDataDirGB;
pStatus->alternativeRole = (uint8_t) tsAlternativeRole;
tstrncpy(pStatus->dnodeEp, tsLocalEp, TSDB_EP_LEN);
// fill cluster cfg parameters
pStatus->clusterCfg.numOfMnodes = htonl(tsNumOfMnodes);
pStatus->clusterCfg.enableBalance = htonl(tsEnableBalance);
pStatus->clusterCfg.mnodeEqualVnodeNum = htonl(tsMnodeEqualVnodeNum);
pStatus->clusterCfg.offlineThreshold = htonl(tsOfflineThreshold);
pStatus->clusterCfg.statusInterval = htonl(tsStatusInterval);
pStatus->clusterCfg.maxtablesPerVnode = htonl(tsMaxTablePerVnode);
pStatus->clusterCfg.maxVgroupsPerDb = htonl(tsMaxVgroupsPerDb);
tstrncpy(pStatus->clusterCfg.arbitrator, tsArbitrator, TSDB_EP_LEN);
tstrncpy(pStatus->clusterCfg.timezone, tsTimezone, 64);
pStatus->clusterCfg.checkTime = 0;
char timestr[32] = "1970-01-01 00:00:00.00";
(void)taosParseTime(timestr, &pStatus->clusterCfg.checkTime, strlen(timestr), TSDB_TIME_PRECISION_MILLI, 0);
tstrncpy(pStatus->clusterCfg.locale, tsLocale, TSDB_LOCALE_LEN);
tstrncpy(pStatus->clusterCfg.charset, tsCharset, TSDB_LOCALE_LEN);
vnodeBuildStatusMsg(pStatus);
contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad);
pStatus->openVnodes = htons(pStatus->openVnodes);
SRpcMsg rpcMsg = {
.pCont = pStatus,
.contLen = contLen,
.msgType = TSDB_MSG_TYPE_DM_STATUS
};
SRpcEpSet epSet;
dnodeGetEpSetForPeer(&epSet);
dnodeSendMsgToDnode(&epSet, &rpcMsg);
}
void dnodeSendStatusMsgToMnode() {
if (tsDnodeTmr != NULL && tsStatusTimer != NULL) {
dInfo("force send status msg to mnode");
taosTmrReset(dnodeSendStatusMsg, 3, NULL, tsDnodeTmr, &tsStatusTimer);
}
}

Some files were not shown because too many files have changed in this diff Show More