xxq250
/
homework-jianmu
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Merge pull request #4418 from taosdata/feature/wal 

[TD-2270]<fix>: race condition in balance module while scheduled by timer
This commit is contained in:
Shengliang Guan 2020-12-03 19:15:05 +08:00 committed by GitHub
parent 659e0d2104 5ab5299316
commit 0c971b9a91
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
27 changed files with 817 additions and 604 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
packaging/cfg/taos.cfg
View File

@ -236,7 +236,7 @@
# httpDebugFlag 131
# debug flag for monitor
# monitorDebugFlag 131
# monDebugFlag 131
# debug flag for query
# qDebugflag 131

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

33
src/balance/inc/bnThread.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_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

566
src/balance/src/balance.c → src/balance/src/bnMain.c
View File

@ -15,17 +15,12 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "tutil.h"
#include "tbalance.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 +28,18 @@
#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;
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 +63,7 @@ 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 +88,26 @@ 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 +124,8 @@ 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 +155,33 @@ 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 +208,7 @@ 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 +250,17 @@ 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 +275,13 @@ 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 +309,25 @@ 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 +338,19 @@ 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 +368,7 @@ 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 +389,7 @@ void balanceReset() {
tsAccessSquence = 0;
}
static int32_t balanceMonitorVgroups() {
static int32_t bnMonitorVgroups() {
void * pIter = NULL;
SVgObj *pVgroup = NULL;
bool hasUpdatingVgroup = false;
@ -429,11 +405,11 @@ 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 +422,7 @@ 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 +432,7 @@ 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 +450,7 @@ static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) {
return false;
}
static bool balanceMontiorDropping() {
static bool bnMontiorDropping() {
void *pIter = NULL;
SDnodeObj *pDnode = NULL;
@ -499,7 +475,7 @@ 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 +485,31 @@ 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;
@ -551,7 +525,7 @@ static void balanceSetVgroupOffline(SDnodeObj* pDnode) {
}
}
static void balanceCheckDnodeAccess() {
void bnCheckStatus() {
void * pIter = NULL;
SDnodeObj *pDnode = NULL;
@ -564,84 +538,39 @@ 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);
}
}
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 +579,7 @@ 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 +594,17 @@ 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 +628,7 @@ 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 +652,29 @@ 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;
}

129
src/balance/src/bnThread.c Normal file
View File

@ -0,0 +1,129 @@
/*
* 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));
}
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_DETACHED);
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() {
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);
}

2
src/client/src/tscSQLParser.c
View File

@ -5103,7 +5103,7 @@ 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},

2
src/common/inc/tglobal.h
View File

@ -179,7 +179,7 @@ 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;

26
src/common/src/tglobal.c
View File

@ -209,7 +209,7 @@ 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;
@ -219,9 +219,9 @@ int32_t wDebugFlag = 135;
int32_t tsdbDebugFlag = 131;
int32_t cqDebugFlag = 135;
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"};
@ -238,7 +238,7 @@ void taosSetAllDebugFlag() {
odbcDebugFlag = debugFlag;
httpDebugFlag = debugFlag;
mqttDebugFlag = debugFlag;
monitorDebugFlag = debugFlag;
monDebugFlag = debugFlag;
qDebugFlag = debugFlag;
rpcDebugFlag = debugFlag;
uDebugFlag = debugFlag;
@ -279,15 +279,15 @@ 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");
@ -310,8 +310,8 @@ 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");
@ -1240,8 +1240,8 @@ 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;

2
src/dnode/src/dnodeMgmt.c
View File

@ -24,7 +24,7 @@
#include "tqueue.h"
#include "tsync.h"
#include "ttimer.h"
#include "tbalance.h"
#include "tbn.h"
#include "tglobal.h"
#include "dnode.h"
#include "vnode.h"

8
src/dnode/src/dnodeModule.c
View File

@ -78,10 +78,10 @@ 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);
}

14
src/inc/monitor.h
View File

@ -47,13 +47,13 @@ typedef struct {
int8_t accessState;
} SAcctMonitorObj;
int32_t monitorInitSystem();
int32_t monitorStartSystem();
void monitorStopSystem();
void monitorCleanUpSystem();
void monitorSaveAcctLog(SAcctMonitorObj *pMonObj);
void monitorSaveLog(int32_t level, const char *const format, ...);
void monitorExecuteSQL(char *sql);
int32_t monInitSystem();
int32_t monStartSystem();
void monStopSystem();
void monCleanupSystem();
void monSaveAcctLog(SAcctMonitorObj *pMonObj);
void monSaveLog(int32_t level, const char *const format, ...);
void monExecuteSQL(char *sql);
#ifdef __cplusplus
}

16
src/inc/tbalance.h → src/inc/tbn.h
View File

@ -23,14 +23,14 @@ extern "C" {
struct SVgObj;
struct SDnodeObj;
int32_t balanceInit();
void balanceCleanUp();
void balanceAsyncNotify();
void balanceSyncNotify();
void balanceReset();
int32_t balanceAllocVnodes(struct SVgObj *pVgroup);
int32_t balanceAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId);
int32_t balanceDropDnode(struct SDnodeObj *pDnode);
int32_t bnInit();
void bnCleanUp();
void bnNotify();
void bnCheckModules();
void bnReset();
int32_t bnAllocVnodes(struct SVgObj *pVgroup);
int32_t bnAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId);
int32_t bnDropDnode(struct SDnodeObj *pDnode);
#ifdef __cplusplus
}

6
src/mnode/inc/mnodeInt.h
View File

@ -41,9 +41,9 @@ extern int32_t sdbDebugFlag;
#define sdbDebug(...) { if (sdbDebugFlag & DEBUG_DEBUG) { taosPrintLog("SDB ", sdbDebugFlag, __VA_ARGS__); }}
#define sdbTrace(...) { if (sdbDebugFlag & DEBUG_TRACE) { taosPrintLog("SDB ", sdbDebugFlag, __VA_ARGS__); }}
#define mLError(...) { monitorSaveLog(2, __VA_ARGS__); mError(__VA_ARGS__) }
#define mLWarn(...) { monitorSaveLog(1, __VA_ARGS__); mWarn(__VA_ARGS__) }
#define mLInfo(...) { monitorSaveLog(0, __VA_ARGS__); mInfo(__VA_ARGS__) }
#define mLError(...) { monSaveLog(2, __VA_ARGS__); mError(__VA_ARGS__) }
#define mLWarn(...) { monSaveLog(1, __VA_ARGS__); mWarn(__VA_ARGS__) }
#define mLInfo(...) { monSaveLog(0, __VA_ARGS__); mInfo(__VA_ARGS__) }
#ifdef __cplusplus
}

4
src/mnode/src/mnodeDb.c
View File

@ -20,7 +20,7 @@
#include "tgrant.h"
#include "tglobal.h"
#include "tname.h"
#include "tbalance.h"
#include "tbn.h"
#include "tdataformat.h"
#include "mnode.h"
#include "mnodeDef.h"
@ -1004,7 +1004,7 @@ static int32_t mnodeAlterDbCb(SMnodeMsg *pMsg, int32_t code) {
mDebug("db:%s, all vgroups is altered", pDb->name);
mLInfo("db:%s, is alterd by %s", pDb->name, mnodeGetUserFromMsg(pMsg));
balanceAsyncNotify();
bnNotify();
return TSDB_CODE_SUCCESS;
}

28
src/mnode/src/mnodeDnode.c
View File

@ -16,12 +16,12 @@
#define _DEFAULT_SOURCE
#include "os.h"
#include "tgrant.h"
#include "tbalance.h"
#include "tbn.h"
#include "tglobal.h"
#include "tconfig.h"
#include "tutil.h"
#include "tsocket.h"
#include "tbalance.h"
#include "tbn.h"
#include "tsync.h"
#include "tdataformat.h"
#include "mnode.h"
@ -115,7 +115,7 @@ static int32_t mnodeDnodeActionDelete(SSdbRow *pRow) {
mnodeDropAllDnodeVgroups(pDnode);
#endif
mnodeDropMnodeLocal(pDnode->dnodeId);
balanceAsyncNotify();
bnNotify();
mnodeUpdateDnodeEps();
mDebug("dnode:%d, all vgroups is dropped from sdb", pDnode->dnodeId);
@ -347,7 +347,7 @@ static int32_t mnodeProcessCfgDnodeMsg(SMnodeMsg *pMsg) {
return TSDB_CODE_MND_INVALID_DNODE_CFG_OPTION;
}
int32_t code = balanceAlterDnode(pDnode, vnodeId, dnodeId);
int32_t code = bnAlterDnode(pDnode, vnodeId, dnodeId);
mnodeDecDnodeRef(pDnode);
return code;
} else {
@ -591,8 +591,8 @@ static int32_t mnodeProcessDnodeStatusMsg(SMnodeMsg *pMsg) {
mInfo("dnode:%d, from offline to online", pDnode->dnodeId);
pDnode->status = TAOS_DN_STATUS_READY;
pDnode->offlineReason = TAOS_DN_OFF_ONLINE;
balanceSyncNotify();
balanceAsyncNotify();
bnCheckModules();
bnNotify();
}
if (openVnodes != pDnode->openVnodes) {
@ -708,7 +708,7 @@ static int32_t mnodeDropDnodeByEp(char *ep, SMnodeMsg *pMsg) {
#ifndef _SYNC
int32_t code = mnodeDropDnode(pDnode, pMsg);
#else
int32_t code = balanceDropDnode(pDnode);
int32_t code = bnDropDnode(pDnode);
#endif
mnodeDecDnodeRef(pDnode);
return code;
@ -1182,12 +1182,12 @@ static char* mnodeGetDnodeAlternativeRoleStr(int32_t alternativeRole) {
#ifndef _SYNC
int32_t balanceInit() { return TSDB_CODE_SUCCESS; }
void balanceCleanUp() {}
void balanceAsyncNotify() {}
void balanceSyncNotify() {}
void balanceReset() {}
int32_t balanceAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId) { return TSDB_CODE_SYN_NOT_ENABLED; }
int32_t bnInit() { return TSDB_CODE_SUCCESS; }
void bnCleanUp() {}
void bnNotify() {}
void bnCheckModules() {}
void bnReset() {}
int32_t bnAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId) { return TSDB_CODE_SYN_NOT_ENABLED; }
char* syncRole[] = {
"offline",
@ -1197,7 +1197,7 @@ char* syncRole[] = {
"master"
};
int32_t balanceAllocVnodes(SVgObj *pVgroup) {
int32_t bnAllocVnodes(SVgObj *pVgroup) {
void * pIter = NULL;
SDnodeObj *pDnode = NULL;
SDnodeObj *pSelDnode = NULL;

4
src/mnode/src/mnodeMain.c
View File

@ -17,7 +17,7 @@
#include "os.h"
#include "taosdef.h"
#include "tsched.h"
#include "tbalance.h"
#include "tbn.h"
#include "tgrant.h"
#include "ttimer.h"
#include "tglobal.h"
@ -58,7 +58,7 @@ static const SMnodeComponent tsMnodeComponents[] = {
{"tables", mnodeInitTables, mnodeCleanupTables},
{"mnodes", mnodeInitMnodes, mnodeCleanupMnodes},
{"sdb", sdbInit, sdbCleanUp},
{"balance", balanceInit, balanceCleanUp},
{"balance", bnInit, bnCleanUp},
{"grant", grantInit, grantCleanUp},
{"show", mnodeInitShow, mnodeCleanUpShow}
};

2
src/mnode/src/mnodeMnode.c
View File

@ -19,7 +19,7 @@
#include "tglobal.h"
#include "trpc.h"
#include "tsync.h"
#include "tbalance.h"
#include "tbn.h"
#include "tutil.h"
#include "tsocket.h"
#include "tdataformat.h"

2
src/mnode/src/mnodePeer.c
View File

@ -20,7 +20,7 @@
#include "tsystem.h"
#include "tutil.h"
#include "tgrant.h"
#include "tbalance.h"
#include "tbn.h"
#include "tglobal.h"
#include "mnode.h"
#include "dnode.h"

2
src/mnode/src/mnodeRead.c
View File

@ -17,7 +17,7 @@
#include "os.h"
#include "taosdef.h"
#include "tsched.h"
#include "tbalance.h"
#include "tbn.h"
#include "tgrant.h"
#include "ttimer.h"
#include "tglobal.h"

4
src/mnode/src/mnodeSdb.c
View File

@ -19,7 +19,7 @@
#include "hash.h"
#include "tutil.h"
#include "tref.h"
#include "tbalance.h"
#include "tbn.h"
#include "tqueue.h"
#include "twal.h"
#include "tsync.h"
@ -244,7 +244,7 @@ static void sdbNotifyRole(int32_t vgId, int8_t role) {
sdbInfo("vgId:1, mnode role changed from %s to %s", syncRole[tsSdbMgmt.role], syncRole[role]);
if (role == TAOS_SYNC_ROLE_MASTER && tsSdbMgmt.role != TAOS_SYNC_ROLE_MASTER) {
balanceReset();
bnReset();
}
tsSdbMgmt.role = role;

4
src/mnode/src/mnodeVgroup.c
View File

@ -20,7 +20,7 @@
#include "tsocket.h"
#include "tidpool.h"
#include "tsync.h"
#include "tbalance.h"
#include "tbn.h"
#include "tglobal.h"
#include "tdataformat.h"
#include "dnode.h"
@ -563,7 +563,7 @@ int32_t mnodeCreateVgroup(SMnodeMsg *pMsg) {
pVgroup->numOfVnodes = pDb->cfg.replications;
pVgroup->createdTime = taosGetTimestampMs();
pVgroup->accessState = TSDB_VN_ALL_ACCCESS;
int32_t code = balanceAllocVnodes(pVgroup);
int32_t code = bnAllocVnodes(pVgroup);
if (code != TSDB_CODE_SUCCESS) {
mError("db:%s, no enough dnode to alloc %d vnodes to vgroup, reason:%s", pDb->name, pVgroup->numOfVnodes,
tstrerror(code));

2
src/mnode/src/mnodeWrite.c
View File

@ -17,7 +17,7 @@
#include "os.h"
#include "taosdef.h"
#include "tsched.h"
#include "tbalance.h"
#include "tbn.h"
#include "tgrant.h"
#include "tglobal.h"
#include "trpc.h"

150
src/plugins/monitor/src/monitorMain.c → src/plugins/monitor/src/monMain.c
View File

@ -27,12 +27,12 @@
#include "monitor.h"
#include "taoserror.h"
#define mnFatal(...) { if (monitorDebugFlag & DEBUG_FATAL) { taosPrintLog("MON FATAL ", 255, __VA_ARGS__); }}
#define mnError(...) { if (monitorDebugFlag & DEBUG_ERROR) { taosPrintLog("MON ERROR ", 255, __VA_ARGS__); }}
#define mnWarn(...) { if (monitorDebugFlag & DEBUG_WARN) { taosPrintLog("MON WARN ", 255, __VA_ARGS__); }}
#define mnInfo(...) { if (monitorDebugFlag & DEBUG_INFO) { taosPrintLog("MON ", 255, __VA_ARGS__); }}
#define mnDebug(...) { if (monitorDebugFlag & DEBUG_DEBUG) { taosPrintLog("MON ", monitorDebugFlag, __VA_ARGS__); }}
#define mnTrace(...) { if (monitorDebugFlag & DEBUG_TRACE) { taosPrintLog("MON ", monitorDebugFlag, __VA_ARGS__); }}
#define monFatal(...) { if (monDebugFlag & DEBUG_FATAL) { taosPrintLog("MON FATAL ", 255, __VA_ARGS__); }}
#define monError(...) { if (monDebugFlag & DEBUG_ERROR) { taosPrintLog("MON ERROR ", 255, __VA_ARGS__); }}
#define monWarn(...) { if (monDebugFlag & DEBUG_WARN) { taosPrintLog("MON WARN ", 255, __VA_ARGS__); }}
#define monInfo(...) { if (monDebugFlag & DEBUG_INFO) { taosPrintLog("MON ", 255, __VA_ARGS__); }}
#define monDebug(...) { if (monDebugFlag & DEBUG_DEBUG) { taosPrintLog("MON ", monDebugFlag, __VA_ARGS__); }}
#define monTrace(...) { if (monDebugFlag & DEBUG_TRACE) { taosPrintLog("MON ", monDebugFlag, __VA_ARGS__); }}
#define SQL_LENGTH 1030
#define LOG_LEN_STR 100
@ -48,12 +48,12 @@ typedef enum {
MON_CMD_CREATE_TB_ACCT_ROOT,
MON_CMD_CREATE_TB_SLOWQUERY,
MON_CMD_MAX
} EMonitorCommand;
} EMonCmd;
typedef enum {
MON_STATE_NOT_INIT,
MON_STATE_INITED
} EMonitorState;
} EMonState;
typedef struct {
pthread_t thread;
@ -64,17 +64,17 @@ typedef struct {
int8_t start; // enable/disable by mnode
int8_t quiting; // taosd is quiting
char sql[SQL_LENGTH + 1];
} SMonitorConn;
} SMonConn;
static SMonitorConn tsMonitor = {0};
static void monitorSaveSystemInfo();
static void *monitorThreadFunc(void *param);
static void monitorBuildMonitorSql(char *sql, int32_t cmd);
extern int32_t (*monitorStartSystemFp)();
extern void (*monitorStopSystemFp)();
extern void (*monitorExecuteSQLFp)(char *sql);
static SMonConn tsMonitor = {0};
static void monSaveSystemInfo();
static void *monThreadFunc(void *param);
static void monBuildMonitorSql(char *sql, int32_t cmd);
extern int32_t (*monStartSystemFp)();
extern void (*monStopSystemFp)();
extern void (*monExecuteSQLFp)(char *sql);
int32_t monitorInitSystem() {
int32_t monInitSystem() {
if (tsMonitor.ep[0] == 0) {
strcpy(tsMonitor.ep, tsLocalEp);
}
@ -90,29 +90,29 @@ int32_t monitorInitSystem() {
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&tsMonitor.thread, &thAttr, monitorThreadFunc, NULL)) {
mnError("failed to create thread to for monitor module, reason:%s", strerror(errno));
if (pthread_create(&tsMonitor.thread, &thAttr, monThreadFunc, NULL)) {
monError("failed to create thread to for monitor module, reason:%s", strerror(errno));
return -1;
}
pthread_attr_destroy(&thAttr);
mnDebug("monitor thread is launched");
monDebug("monitor thread is launched");
monitorStartSystemFp = monitorStartSystem;
monitorStopSystemFp = monitorStopSystem;
monStartSystemFp = monStartSystem;
monStopSystemFp = monStopSystem;
return 0;
}
int32_t monitorStartSystem() {
int32_t monStartSystem() {
taos_init();
tsMonitor.start = 1;
monitorExecuteSQLFp = monitorExecuteSQL;
mnInfo("monitor module start");
monExecuteSQLFp = monExecuteSQL;
monInfo("monitor module start");
return 0;
}
static void *monitorThreadFunc(void *param) {
mnDebug("starting to initialize monitor module ...");
static void *monThreadFunc(void *param) {
monDebug("starting to initialize monitor module ...");
while (1) {
static int32_t accessTimes = 0;
@ -121,7 +121,7 @@ static void *monitorThreadFunc(void *param) {
if (tsMonitor.quiting) {
tsMonitor.state = MON_STATE_NOT_INIT;
mnInfo("monitor thread will quit, for taosd is quiting");
monInfo("monitor thread will quit, for taosd is quiting");
break;
} else {
taosGetDisk();
@ -132,7 +132,7 @@ static void *monitorThreadFunc(void *param) {
}
if (dnodeGetDnodeId() <= 0) {
mnDebug("dnode not initialized, waiting for 3000 ms to start monitor module");
monDebug("dnode not initialized, waiting for 3000 ms to start monitor module");
continue;
}
@ -140,10 +140,10 @@ static void *monitorThreadFunc(void *param) {
tsMonitor.state = MON_STATE_NOT_INIT;
tsMonitor.conn = taos_connect(NULL, "monitor", tsInternalPass, "", 0);
if (tsMonitor.conn == NULL) {
mnError("failed to connect to database, reason:%s", tstrerror(terrno));
monError("failed to connect to database, reason:%s", tstrerror(terrno));
continue;
} else {
mnDebug("connect to database success");
monDebug("connect to database success");
}
}
@ -151,16 +151,16 @@ static void *monitorThreadFunc(void *param) {
int code = 0;
for (; tsMonitor.cmdIndex < MON_CMD_MAX; ++tsMonitor.cmdIndex) {
monitorBuildMonitorSql(tsMonitor.sql, tsMonitor.cmdIndex);
monBuildMonitorSql(tsMonitor.sql, tsMonitor.cmdIndex);
void *res = taos_query(tsMonitor.conn, tsMonitor.sql);
code = taos_errno(res);
taos_free_result(res);
if (code != 0) {
mnError("failed to exec sql:%s, reason:%s", tsMonitor.sql, tstrerror(code));
monError("failed to exec sql:%s, reason:%s", tsMonitor.sql, tstrerror(code));
break;
} else {
mnDebug("successfully to exec sql:%s", tsMonitor.sql);
monDebug("successfully to exec sql:%s", tsMonitor.sql);
}
}
@ -171,16 +171,16 @@ static void *monitorThreadFunc(void *param) {
if (tsMonitor.state == MON_STATE_INITED) {
if (accessTimes % tsMonitorInterval == 0) {
monitorSaveSystemInfo();
monSaveSystemInfo();
}
}
}
mnInfo("monitor thread is stopped");
monInfo("monitor thread is stopped");
return NULL;
}
static void monitorBuildMonitorSql(char *sql, int32_t cmd) {
static void monBuildMonitorSql(char *sql, int32_t cmd) {
memset(sql, 0, SQL_LENGTH);
if (cmd == MON_CMD_CREATE_DB) {
@ -236,47 +236,47 @@ static void monitorBuildMonitorSql(char *sql, int32_t cmd) {
sql[SQL_LENGTH] = 0;
}
void monitorStopSystem() {
void monStopSystem() {
tsMonitor.start = 0;
tsMonitor.state = MON_STATE_NOT_INIT;
monitorExecuteSQLFp = NULL;
mnInfo("monitor module stopped");
monExecuteSQLFp = NULL;
monInfo("monitor module stopped");
}
void monitorCleanUpSystem() {
void monCleanupSystem() {
tsMonitor.quiting = 1;
monitorStopSystem();
monStopSystem();
pthread_join(tsMonitor.thread, NULL);
if (tsMonitor.conn != NULL) {
taos_close(tsMonitor.conn);
tsMonitor.conn = NULL;
}
mnInfo("monitor module is cleaned up");
monInfo("monitor module is cleaned up");
}
// unit is MB
static int32_t monitorBuildMemorySql(char *sql) {
static int32_t monBuildMemorySql(char *sql) {
float sysMemoryUsedMB = 0;
bool suc = taosGetSysMemory(&sysMemoryUsedMB);
if (!suc) {
mnDebug("failed to get sys memory info");
monDebug("failed to get sys memory info");
}
float procMemoryUsedMB = 0;
suc = taosGetProcMemory(&procMemoryUsedMB);
if (!suc) {
mnDebug("failed to get proc memory info");
monDebug("failed to get proc memory info");
}
return sprintf(sql, ", %f, %f, %d", procMemoryUsedMB, sysMemoryUsedMB, tsTotalMemoryMB);
}
// unit is %
static int32_t monitorBuildCpuSql(char *sql) {
static int32_t monBuildCpuSql(char *sql) {
float sysCpuUsage = 0, procCpuUsage = 0;
bool suc = taosGetCpuUsage(&sysCpuUsage, &procCpuUsage);
if (!suc) {
mnDebug("failed to get cpu usage");
monDebug("failed to get cpu usage");
}
if (sysCpuUsage <= procCpuUsage) {
@ -287,72 +287,72 @@ static int32_t monitorBuildCpuSql(char *sql) {
}
// unit is GB
static int32_t monitorBuildDiskSql(char *sql) {
static int32_t monBuildDiskSql(char *sql) {
return sprintf(sql, ", %f, %d", (tsTotalDataDirGB - tsAvailDataDirGB), (int32_t)tsTotalDataDirGB);
}
// unit is Kb
static int32_t monitorBuildBandSql(char *sql) {
static int32_t monBuildBandSql(char *sql) {
float bandSpeedKb = 0;
bool suc = taosGetBandSpeed(&bandSpeedKb);
if (!suc) {
mnDebug("failed to get bandwidth speed");
monDebug("failed to get bandwidth speed");
}
return sprintf(sql, ", %f", bandSpeedKb);
}
static int32_t monitorBuildReqSql(char *sql) {
static int32_t monBuildReqSql(char *sql) {
SStatisInfo info = dnodeGetStatisInfo();
return sprintf(sql, ", %d, %d, %d)", info.httpReqNum, info.queryReqNum, info.submitReqNum);
}
static int32_t monitorBuildIoSql(char *sql) {
static int32_t monBuildIoSql(char *sql) {
float readKB = 0, writeKB = 0;
bool suc = taosGetProcIO(&readKB, &writeKB);
if (!suc) {
mnDebug("failed to get io info");
monDebug("failed to get io info");
}
return sprintf(sql, ", %f, %f", readKB, writeKB);
}
static void monitorSaveSystemInfo() {
static void monSaveSystemInfo() {
int64_t ts = taosGetTimestampUs();
char * sql = tsMonitor.sql;
int32_t pos = snprintf(sql, SQL_LENGTH, "insert into %s.dn%d values(%" PRId64, tsMonitorDbName, dnodeGetDnodeId(), ts);
pos += monitorBuildCpuSql(sql + pos);
pos += monitorBuildMemorySql(sql + pos);
pos += monitorBuildDiskSql(sql + pos);
pos += monitorBuildBandSql(sql + pos);
pos += monitorBuildIoSql(sql + pos);
pos += monitorBuildReqSql(sql + pos);
pos += monBuildCpuSql(sql + pos);
pos += monBuildMemorySql(sql + pos);
pos += monBuildDiskSql(sql + pos);
pos += monBuildBandSql(sql + pos);
pos += monBuildIoSql(sql + pos);
pos += monBuildReqSql(sql + pos);
void *res = taos_query(tsMonitor.conn, tsMonitor.sql);
int code = taos_errno(res);
taos_free_result(res);
if (code != 0) {
mnError("failed to save system info, reason:%s, sql:%s", tstrerror(code), tsMonitor.sql);
monError("failed to save system info, reason:%s, sql:%s", tstrerror(code), tsMonitor.sql);
} else {
mnDebug("successfully to save system info, sql:%s", tsMonitor.sql);
monDebug("successfully to save system info, sql:%s", tsMonitor.sql);
}
}
static void montiorExecSqlCb(void *param, TAOS_RES *result, int32_t code) {
static void monExecSqlCb(void *param, TAOS_RES *result, int32_t code) {
int32_t c = taos_errno(result);
if (c != TSDB_CODE_SUCCESS) {
mnError("save %s failed, reason:%s", (char *)param, tstrerror(c));
monError("save %s failed, reason:%s", (char *)param, tstrerror(c));
} else {
int32_t rows = taos_affected_rows(result);
mnDebug("save %s succ, rows:%d", (char *)param, rows);
monDebug("save %s succ, rows:%d", (char *)param, rows);
}
taos_free_result(result);
}
void monitorSaveAcctLog(SAcctMonitorObj *pMon) {
void monSaveAcctLog(SAcctMonitorObj *pMon) {
if (tsMonitor.state != MON_STATE_INITED) return;
char sql[1024] = {0};
@ -382,11 +382,11 @@ void monitorSaveAcctLog(SAcctMonitorObj *pMon) {
pMon->totalConns, pMon->maxConns,
pMon->accessState);
mnDebug("save account info, sql:%s", sql);
taos_query_a(tsMonitor.conn, sql, montiorExecSqlCb, "account info");
monDebug("save account info, sql:%s", sql);
taos_query_a(tsMonitor.conn, sql, monExecSqlCb, "account info");
}
void monitorSaveLog(int32_t level, const char *const format, ...) {
void monSaveLog(int32_t level, const char *const format, ...) {
if (tsMonitor.state != MON_STATE_INITED) return;
va_list argpointer;
@ -403,13 +403,13 @@ void monitorSaveLog(int32_t level, const char *const format, ...) {
len += sprintf(sql + len, "', '%s')", tsLocalEp);
sql[len++] = 0;
mnDebug("save log, sql: %s", sql);
taos_query_a(tsMonitor.conn, sql, montiorExecSqlCb, "log");
monDebug("save log, sql: %s", sql);
taos_query_a(tsMonitor.conn, sql, monExecSqlCb, "log");
}
void monitorExecuteSQL(char *sql) {
void monExecuteSQL(char *sql) {
if (tsMonitor.state != MON_STATE_INITED) return;
mnDebug("execute sql:%s", sql);
taos_query_a(tsMonitor.conn, sql, montiorExecSqlCb, "sql");
monDebug("execute sql:%s", sql);
taos_query_a(tsMonitor.conn, sql, monExecSqlCb, "sql");
}

8
tests/script/general/alter/dnode.sim
View File

@ -24,7 +24,7 @@ sql alter dnode 1 debugFlag 135
sql alter dnode 1 debugFlag 131
sql alter dnode 1 monitor 0
sql alter dnode 1 debugFlag 135
sql alter dnode 1 monitorDebugFlag 135
sql alter dnode 1 monDebugFlag 135
sql alter dnode 1 vDebugFlag 135
sql alter dnode 1 mDebugFlag 135
sql alter dnode 1 cDebugFlag 135
@ -44,15 +44,15 @@ sql_error alter dnode 2 tmrDebugFlag 135
print ======== step3
sql_error alter $hostname1 debugFlag 135
sql_error alter $hostname1 monitorDebugFlag 135
sql_error alter $hostname1 monDebugFlag 135
sql_error alter $hostname1 vDebugFlag 135
sql_error alter $hostname1 mDebugFlag 135
sql_error alter dnode $hostname2 debugFlag 135
sql_error alter dnode $hostname2 monitorDebugFlag 135
sql_error alter dnode $hostname2 monDebugFlag 135
sql_error alter dnode $hostname2 vDebugFlag 135
sql_error alter dnode $hostname2 mDebugFlag 135
sql alter dnode $hostname1 debugFlag 135
sql alter dnode $hostname1 monitorDebugFlag 135
sql alter dnode $hostname1 monDebugFlag 135
sql alter dnode $hostname1 vDebugFlag 135
sql alter dnode $hostname1 tmrDebugFlag 131

2
tests/script/sh/deploy.sh
View File

@ -120,7 +120,7 @@ echo "cDebugFlag 143" >> $TAOS_CFG
echo "jnidebugFlag 143" >> $TAOS_CFG
echo "odbcdebugFlag 143" >> $TAOS_CFG
echo "httpDebugFlag 143" >> $TAOS_CFG
echo "monitorDebugFlag 143" >> $TAOS_CFG
echo "monDebugFlag 143" >> $TAOS_CFG
echo "mqttDebugFlag 143" >> $TAOS_CFG
echo "qdebugFlag 143" >> $TAOS_CFG
echo "rpcDebugFlag 143" >> $TAOS_CFG