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

Merge remote-tracking branch 'origin/feature/raft-interface' into feature/dnode3

This commit is contained in:
Shengliang Guan 2021-10-26 10:24:10 +08:00
parent 8f462f7a8d 18cd0c861a
commit e8c5ed863d
24 changed files with 1407 additions and 2007 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
include/common/taosmsg.h
View File

@ -707,41 +707,30 @@ typedef struct {
} SStatusRsp; } SStatusRsp;
typedef struct { typedef struct {
uint32_t vgId; uint16_t port;
int32_t dbCfgVersion; char fqdn[TSDB_FQDN_LEN];
int32_t maxTables;
int32_t cacheBlockSize;
int32_t totalBlocks;
int32_t daysPerFile;
int32_t daysToKeep;
int32_t daysToKeep1;
int32_t daysToKeep2;
int32_t minRowsPerFileBlock;
int32_t maxRowsPerFileBlock;
int32_t commitTime;
int32_t fsyncPeriod;
int8_t precision;
int8_t compression;
int8_t walLevel;
int8_t vgReplica;
int8_t wals;
int8_t quorum;
int8_t update;
int8_t cacheLastRow;
int32_t vgCfgVersion;
int8_t dbReplica;
int8_t dbType;
int8_t reserved[8];
} SVnodeCfg;
typedef struct {
int32_t nodeId;
char nodeEp[TSDB_EP_LEN];
} SVnodeDesc; } SVnodeDesc;
typedef struct { typedef struct {
char db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN]; char db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
SVnodeCfg cfg; uint32_t vgId;
int32_t cacheBlockSize;
int32_t totalBlocks;
int32_t daysPerFile;
int32_t daysToKeep0;
int32_t daysToKeep1;
int32_t daysToKeep2;
int32_t minRowsPerFileBlock;
int32_t maxRowsPerFileBlock;
int8_t precision;
int8_t compression;
int8_t cacheLastRow;
int8_t update;
int8_t walLevel;
int8_t replica;
int8_t quorum;
int8_t reserved[9];
int32_t fsyncPeriod;
SVnodeDesc nodes[TSDB_MAX_REPLICA]; SVnodeDesc nodes[TSDB_MAX_REPLICA];
} SCreateVnodeMsg, SAlterVnodeMsg; } SCreateVnodeMsg, SAlterVnodeMsg;

33
include/libs/sync/sync.h
View File

@ -24,10 +24,10 @@ extern "C" {
#include "taosdef.h" #include "taosdef.h"
#include "wal.h" #include "wal.h"
typedef uint32_t SyncNodeId; typedef int64_t SyncNodeId;
typedef int32_t SyncGroupId; typedef int32_t SyncGroupId;
typedef int64_t SyncIndex; typedef int64_t SyncIndex;
typedef uint64_t SyncTerm; typedef uint64_t SSyncTerm;
typedef enum { typedef enum {
TAOS_SYNC_ROLE_FOLLOWER = 0, TAOS_SYNC_ROLE_FOLLOWER = 0,
@ -41,9 +41,8 @@ typedef struct {
} SSyncBuffer; } SSyncBuffer;
typedef struct { typedef struct {
SyncNodeId nodeId; // node ID assigned by TDengine uint16_t nodePort; // node sync Port
uint16_t nodePort; // node sync Port char nodeFqdn[TSDB_FQDN_LEN]; // node FQDN
char nodeFqdn[TSDB_FQDN_LEN]; // node FQDN
} SNodeInfo; } SNodeInfo;
typedef struct { typedef struct {
@ -53,9 +52,9 @@ typedef struct {
} SSyncCluster; } SSyncCluster;
typedef struct { typedef struct {
int32_t selfIndex; int32_t selfIndex;
int nNode; int nNode;
SyncNodeId* nodeId; SNodeInfo* node;
ESyncRole* role; ESyncRole* role;
} SNodesRole; } SNodesRole;
@ -85,8 +84,8 @@ typedef struct SSyncFSM {
} SSyncFSM; } SSyncFSM;
typedef struct SSyncServerState { typedef struct SSyncServerState {
SyncNodeId voteFor; SNodeInfo voteFor;
SyncTerm term; SSyncTerm term;
} SSyncServerState; } SSyncServerState;
typedef struct SStateManager { typedef struct SStateManager {
@ -106,8 +105,8 @@ typedef struct {
twalh walHandle; twalh walHandle;
SyncIndex snapshotIndex; // initial version SyncIndex snapshotIndex;
SSyncCluster syncCfg; // configuration from mgmt SSyncCluster syncCfg;
SSyncFSM fsm; SSyncFSM fsm;
@ -122,7 +121,11 @@ void syncStop(SyncNodeId);
int32_t syncPropose(SyncNodeId nodeId, SSyncBuffer buffer, void* pData, bool isWeak); int32_t syncPropose(SyncNodeId nodeId, SSyncBuffer buffer, void* pData, bool isWeak);
extern int32_t syncDebugFlag; int32_t syncAddNode(SyncNodeId nodeId, const SNodeInfo *pNode);
int32_t syncRemoveNode(SyncNodeId nodeId, const SNodeInfo *pNode);
extern int32_t syncDebugFlag;
#ifdef __cplusplus #ifdef __cplusplus
} }

2
include/util/tdef.h
View File

@ -423,6 +423,8 @@ do { \
//#define varDataNetLen(v) (htons(((VarDataLenT *)(v))[0])) //#define varDataNetLen(v) (htons(((VarDataLenT *)(v))[0]))
//#define varDataNetTLen(v) (sizeof(VarDataLenT) + varDataNetLen(v)) //#define varDataNetTLen(v) (sizeof(VarDataLenT) + varDataNetLen(v))
enum { TRANS_STAT_INIT = 0, TRANS_STAT_EXECUTING, TRANS_STAT_EXECUTED, TRANS_STAT_ROLLBACKING, TRANS_STAT_ROLLBACKED };
enum { TRANS_OPER_INIT = 0, TRANS_OPER_EXECUTE, TRANS_OPER_ROLLBACK };
#ifdef __cplusplus #ifdef __cplusplus
} }

7
source/libs/wal/src/wal.c
View File

@ -15,5 +15,10 @@
#include "wal.h" #include "wal.h"
int32_t walInit() {return 0;} int32_t walInit() { return 0; }
void walCleanUp() {} void walCleanUp() {}
twalh walOpen(char *path, SWalCfg *pCfg) { return NULL; }
int32_t walAlter(twalh pWal, SWalCfg *pCfg) { return 0; }

9
source/server/dnode/src/dnodeTrans.c
View File

@ -96,8 +96,12 @@ static int32_t dnodeInitServer() {
tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_GRANT] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_GRANT] = mnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_STATUS] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_STATUS] = mnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_CONNECT] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_CONNECT] = vnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_CONSUME] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_DISCONNECT] = vnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_ACK] = vnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_RESET] = vnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_CONSUME] = vnodeProcessMsg;
tsTrans.peerMsgFp[TSDB_MSG_TYPE_MQ_QUERY] = vnodeProcessMsg;
SRpcInit rpcInit; SRpcInit rpcInit;
memset(&rpcInit, 0, sizeof(rpcInit)); memset(&rpcInit, 0, sizeof(rpcInit));
@ -143,6 +147,7 @@ static void dnodeProcessPeerRsp(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
RpcMsgFp fp = tsTrans.peerMsgFp[msgType]; RpcMsgFp fp = tsTrans.peerMsgFp[msgType];
if (fp != NULL) { if (fp != NULL) {
dTrace("RPC %p, peer rsp:%s will be processed", pMsg->handle, taosMsg[msgType]);
(*fp)(pMsg); (*fp)(pMsg);
} else { } else {
dDebug("RPC %p, peer rsp:%s not processed", pMsg->handle, taosMsg[msgType]); dDebug("RPC %p, peer rsp:%s not processed", pMsg->handle, taosMsg[msgType]);

6
source/server/vnode/inc/vnodeCfg.h → source/server/vnode/inc/vnodeFile.h
View File

@ -21,8 +21,10 @@ extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeReadCfg(SVnode *pVnode); int32_t vnodeReadCfg(int32_t vgId, SVnodeCfg *pCfg);
int32_t vnodeWriteCfg(SCreateVnodeMsg *pVnodeCfg); int32_t vnodeWriteCfg(int32_t vgId, SVnodeCfg *pCfg);
int32_t vnodeReadTerm(int32_t vgId, SSyncServerState *pState);
int32_t vnodeWriteTerm(int32_t vgid, SSyncServerState *pState);
#ifdef __cplusplus #ifdef __cplusplus
} }

104
source/server/vnode/inc/vnodeInt.h
View File

@ -16,11 +16,12 @@
#ifndef _TD_VNODE_INT_H_ #ifndef _TD_VNODE_INT_H_
#define _TD_VNODE_INT_H_ #define _TD_VNODE_INT_H_
#include "os.h"
#include "amalloc.h" #include "amalloc.h"
#include "meta.h" #include "meta.h"
#include "os.h"
#include "sync.h" #include "sync.h"
#include "taosmsg.h" #include "taosmsg.h"
#include "tglobal.h"
#include "tlog.h" #include "tlog.h"
#include "tq.h" #include "tq.h"
#include "tqueue.h" #include "tqueue.h"
@ -43,58 +44,77 @@ extern int32_t vDebugFlag;
#define vDebug(...) { if (vDebugFlag & DEBUG_DEBUG) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }} #define vDebug(...) { if (vDebugFlag & DEBUG_DEBUG) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
#define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }} #define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
typedef struct STsdbCfg {
int32_t cacheBlockSize; // MB
int32_t totalBlocks;
int32_t daysPerFile;
int32_t daysToKeep0;
int32_t daysToKeep1;
int32_t daysToKeep2;
int32_t minRowsPerFileBlock;
int32_t maxRowsPerFileBlock;
uint8_t precision; // time resolution
int8_t compression;
int8_t cacheLastRow;
int8_t update;
} STsdbCfg;
typedef struct SMetaCfg {
} SMetaCfg;
typedef struct SSyncCluster {
int8_t replica;
int8_t quorum;
SNodeInfo nodes[TSDB_MAX_REPLICA];
} SSyncCfg;
typedef struct SVnodeCfg {
char db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
int8_t dropped;
SWalCfg wal;
STsdbCfg tsdb;
SMetaCfg meta;
SSyncCfg sync;
} SVnodeCfg;
typedef struct { typedef struct {
int32_t vgId; // global vnode group ID int32_t vgId; // global vnode group ID
int32_t refCount; // reference count int32_t refCount; // reference count
SMemAllocator *allocator; SMemAllocator *allocator;
SMeta *pMeta; SMeta *pMeta;
STsdb *pTsdb; STsdb *pTsdb;
STQ *pTQ; STQ *pTQ;
twalh pWal; twalh pWal;
SyncNodeId syncNode; void *pQuery;
taos_queue pWriteQ; // write queue SyncNodeId syncNode;
taos_queue pQueryQ; // read query queue taos_queue pWriteQ; // write queue
taos_queue pFetchQ; // read fetch/cancel queue taos_queue pQueryQ; // read query queue
SWalCfg walCfg; taos_queue pFetchQ; // read fetch/cancel queue
SSyncCluster syncCfg; SVnodeCfg cfg;
char db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN]; SSyncServerState term;
int64_t queuedWMsgSize; int64_t queuedWMsgSize;
int32_t queuedWMsg; int32_t queuedWMsg;
int32_t queuedRMsg; int32_t queuedRMsg;
int32_t numOfQHandle; // current initialized and existed query handle in current dnode int32_t numOfQHandle; // current initialized and existed query handle in current dnode
int8_t status; int8_t role;
int8_t role; int8_t accessState;
int8_t accessState; int8_t dropped;
int8_t dropped; int8_t status;
pthread_mutex_t statusMutex; pthread_mutex_t statusMutex;
} SVnode; } SVnode;
typedef struct { typedef struct {
int32_t len; int32_t len;
void * rsp; void *rsp;
void * qhandle; // used by query and retrieve msg void *qhandle; // used by query and retrieve msg
} SVnRsp; } SVnRsp;
void vnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg); void vnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg);
void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg); void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg);
void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port); void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port);
int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg);
int32_t vnodeDrop(int32_t vgId);
int32_t vnodeOpen(int32_t vgId);
int32_t vnodeAlter(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg);
int32_t vnodeSync(int32_t vgId);
int32_t vnodeClose(int32_t vgId);
void vnodeCleanUp(SVnode *pVnode);
void vnodeDestroy(SVnode *pVnode);
int32_t vnodeCompact(int32_t vgId);
void vnodeBackup(int32_t vgId);
void vnodeGetStatus(struct SStatusMsg *status);
SVnode *vnodeAcquire(int32_t vgId);
SVnode *vnodeAcquireNotClose(int32_t vgId);
void vnodeRelease(SVnode *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif

27
source/server/vnode/inc/vnodeMgmtMsg.h → source/server/vnode/inc/vnodeMain.h
View File

@ -13,23 +13,30 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef _TD_VNODE_MGMT_MSG_H_ #ifndef _TD_VNODE_MAIN_H_
#define _TD_VNODE_MGMT_MSG_H_ #define _TD_VNODE_MAIN_H_
#include "vnodeInt.h"
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h"
int32_t vnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg); int32_t vnodeInitMain();
int32_t vnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg); void vnodeCleanupMain();
int32_t vnodeProcessSyncVnodeMsg(SRpcMsg *rpcMsg);
int32_t vnodeProcessCompactVnodeMsg(SRpcMsg *rpcMsg); SVnode *vnodeAcquireInAllState(int32_t vgId);
int32_t vnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg); SVnode *vnodeAcquire(int32_t vgId);
int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg); void vnodeRelease(SVnode *pVnode);
int32_t vnodeCreateVnode(int32_t vgId, SVnodeCfg *pCfg);
int32_t vnodeAlterVnode(SVnode *pVnode, SVnodeCfg *pCfg);
int32_t vnodeDropVnode(SVnode *pVnode);
int32_t vnodeSyncVnode(SVnode *pVnode);
int32_t vnodeCompactVnode(SVnode *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif /*_TD_VNODE_MGMT_H_*/ #endif /*_TD_VNODE_MAIN_H_*/

8
source/server/vnode/inc/vnodeMgmt.h
View File

@ -21,6 +21,14 @@ extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
typedef struct {
SVnode *pVnode;
SRpcMsg rpcMsg;
char pCont[];
} SVnMgmtMsg;
int32_t vnodeInitMgmt(); int32_t vnodeInitMgmt();
void vnodeCleanupMgmt(); void vnodeCleanupMgmt();
void vnodeProcessMgmtMsg(SRpcMsg *pMsg); void vnodeProcessMgmtMsg(SRpcMsg *pMsg);

47
source/server/vnode/inc/vnodeStatus.h
View File

@ -1,47 +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/>.
*/
#ifndef _TD_VNODE_STATUS_H_
#define _TD_VNODE_STATUS_H_
#ifdef __cplusplus
extern "C" {
#endif
#include "vnodeInt.h"
typedef enum _VN_STATUS {
TAOS_VN_STATUS_INIT = 0,
TAOS_VN_STATUS_READY = 1,
TAOS_VN_STATUS_CLOSING = 2,
TAOS_VN_STATUS_UPDATING = 3
} EVnodeStatus;
// vnodeStatus
extern char* vnodeStatus[];
bool vnodeSetInitStatus(SVnode* pVnode);
bool vnodeSetReadyStatus(SVnode* pVnode);
bool vnodeSetClosingStatus(SVnode* pVnode);
bool vnodeSetUpdatingStatus(SVnode* pVnode);
bool vnodeInInitStatus(SVnode* pVnode);
bool vnodeInReadyStatus(SVnode* pVnode);
bool vnodeInClosingStatus(SVnode* pVnode);
#ifdef __cplusplus
}
#endif
#endif /*_TD_VNODE_STATUS_H_*/

31
source/server/vnode/inc/vnodeVersion.h
View File

@ -1,31 +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/>.
*/
#ifndef _TD_VNODE_VERSION_H_
#define _TD_VNODE_VERSION_H_
#ifdef __cplusplus
extern "C" {
#endif
#include "vnodeInt.h"
int32_t vnodeReadVersion(SVnode *pVnode);
int32_t vnodeSaveVersion(SVnode *pVnode);
#ifdef __cplusplus
}
#endif
#endif /*_TD_VNODE_VERSION_H_*/

34
source/server/vnode/inc/vnodeWorker.h
View File

@ -1,34 +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/>.
*/
#ifndef _TD_VNODE_WORKER_H_
#define _TD_VNODE_WORKER_H_
#ifdef __cplusplus
extern "C" {
#endif
#include "vnodeInt.h"
int32_t vnodeInitWorker();
void vnodeCleanupWorker();
void vnodeProcessCleanupTask(SVnode *pVnode);
void vnodeProcessDestroyTask(SVnode *pVnode);
void vnodeProcessBackupTask(SVnode *pVnode);
#ifdef __cplusplus
}
#endif
#endif /*_TD_VNODE_WORKER_H_*/

381
source/server/vnode/src/vnodeCfg.c
View File

@ -1,381 +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 "tglobal.h"
#include "vnodeCfg.h"
static void vnodeLoadCfg(SVnode *pVnode, SCreateVnodeMsg *vnodeMsg) {
#if 0
tstrncpy(pVnode->db, vnodeMsg->db, sizeof(pVnode->db));
pVnode->dbCfgVersion = vnodeMsg->cfg.dbCfgVersion;
pVnode->vgCfgVersion = vnodeMsg->cfg.vgCfgVersion;
pVnode->tsdbCfg.cacheBlockSize = vnodeMsg->cfg.cacheBlockSize;
pVnode->tsdbCfg.totalBlocks = vnodeMsg->cfg.totalBlocks;
pVnode->tsdbCfg.daysPerFile = vnodeMsg->cfg.daysPerFile;
pVnode->tsdbCfg.keep = vnodeMsg->cfg.daysToKeep;
pVnode->tsdbCfg.keep1 = vnodeMsg->cfg.daysToKeep1;
pVnode->tsdbCfg.keep2 = vnodeMsg->cfg.daysToKeep2;
pVnode->tsdbCfg.minRowsPerFileBlock = vnodeMsg->cfg.minRowsPerFileBlock;
pVnode->tsdbCfg.maxRowsPerFileBlock = vnodeMsg->cfg.maxRowsPerFileBlock;
pVnode->tsdbCfg.precision = vnodeMsg->cfg.precision;
pVnode->tsdbCfg.compression = vnodeMsg->cfg.compression;
pVnode->tsdbCfg.update = vnodeMsg->cfg.update;
pVnode->tsdbCfg.cacheLastRow = vnodeMsg->cfg.cacheLastRow;
pVnode->walCfg.walLevel = vnodeMsg->cfg.walLevel;
pVnode->walCfg.fsyncPeriod = vnodeMsg->cfg.fsyncPeriod;
pVnode->walCfg.keep = TAOS_WAL_NOT_KEEP;
pVnode->syncCfg.replica = vnodeMsg->cfg.vgReplica;
pVnode->syncCfg.quorum = vnodeMsg->cfg.quorum;
pVnode->dbReplica = vnodeMsg->cfg.dbReplica;
pVnode->dbType = vnodeMsg->cfg.dbType;
for (int i = 0; i < pVnode->syncCfg.replica; ++i) {
SVnodeDesc *node = &vnodeMsg->nodes[i];
pVnode->syncCfg.nodeInfo[i].nodeId = node->nodeId;
taosGetFqdnPortFromEp(node->nodeEp, pVnode->syncCfg.nodeInfo[i].nodeFqdn, &pVnode->syncCfg.nodeInfo[i].nodePort);
pVnode->syncCfg.nodeInfo[i].nodePort += TSDB_PORT_SYNC;
}
vInfo("vgId:%d, load vnode cfg successfully, replcia:%d", pVnode->vgId, pVnode->syncCfg.replica);
for (int32_t i = 0; i < pVnode->syncCfg.replica; i++) {
SNodeInfo *node = &pVnode->syncCfg.nodeInfo[i];
vInfo("vgId:%d, dnode:%d, %s:%u", pVnode->vgId, node->nodeId, node->nodeFqdn, node->nodePort);
}
#endif
}
int32_t vnodeReadCfg(SVnode *pVnode) {
#if 0
int32_t ret = TSDB_CODE_VND_APP_ERROR;
int32_t len = 0;
int maxLen = 1000;
char * content = calloc(1, maxLen + 1);
cJSON * root = NULL;
FILE * fp = NULL;
bool nodeChanged = false;
SCreateVnodeMsg vnodeMsg;
char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, pVnode->vgId);
vnodeMsg.cfg.vgId = pVnode->vgId;
fp = fopen(file, "r");
if (!fp) {
vError("vgId:%d, failed to open vnode cfg file:%s to read, error:%s", pVnode->vgId, file, strerror(errno));
ret = TAOS_SYSTEM_ERROR(errno);
goto PARSE_VCFG_ERROR;
}
len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) {
vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
content[len] = 0;
root = cJSON_Parse(content);
if (root == NULL) {
vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
cJSON *db = cJSON_GetObjectItem(root, "db");
if (!db || db->type != cJSON_String || db->valuestring == NULL) {
vError("vgId:%d, failed to read %s, db not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
tstrncpy(vnodeMsg.db, db->valuestring, sizeof(vnodeMsg.db));
cJSON *dbCfgVersion = cJSON_GetObjectItem(root, "cfgVersion");
if (!dbCfgVersion || dbCfgVersion->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, cfgVersion not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.dbCfgVersion = (int32_t)dbCfgVersion->valueint;
cJSON *vgCfgVersion = cJSON_GetObjectItem(root, "vgCfgVersion");
if (!vgCfgVersion || vgCfgVersion->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, vgCfgVersion not found", pVnode->vgId, file);
vnodeMsg.cfg.vgCfgVersion = 0;
} else {
vnodeMsg.cfg.vgCfgVersion = (int32_t)vgCfgVersion->valueint;
}
cJSON *cacheBlockSize = cJSON_GetObjectItem(root, "cacheBlockSize");
if (!cacheBlockSize || cacheBlockSize->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, cacheBlockSize not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.cacheBlockSize = (int32_t)cacheBlockSize->valueint;
cJSON *totalBlocks = cJSON_GetObjectItem(root, "totalBlocks");
if (!totalBlocks || totalBlocks->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, totalBlocks not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.totalBlocks = (int32_t)totalBlocks->valueint;
cJSON *daysPerFile = cJSON_GetObjectItem(root, "daysPerFile");
if (!daysPerFile || daysPerFile->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysPerFile not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.daysPerFile = (int32_t)daysPerFile->valueint;
cJSON *daysToKeep = cJSON_GetObjectItem(root, "daysToKeep");
if (!daysToKeep || daysToKeep->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysToKeep not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.daysToKeep = (int32_t)daysToKeep->valueint;
cJSON *daysToKeep1 = cJSON_GetObjectItem(root, "daysToKeep1");
if (!daysToKeep1 || daysToKeep1->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysToKeep1 not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.daysToKeep1 = (int32_t)daysToKeep1->valueint;
cJSON *daysToKeep2 = cJSON_GetObjectItem(root, "daysToKeep2");
if (!daysToKeep2 || daysToKeep2->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysToKeep2 not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.daysToKeep2 = (int32_t)daysToKeep2->valueint;
cJSON *minRowsPerFileBlock = cJSON_GetObjectItem(root, "minRowsPerFileBlock");
if (!minRowsPerFileBlock || minRowsPerFileBlock->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, minRowsPerFileBlock not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.minRowsPerFileBlock = (int32_t)minRowsPerFileBlock->valueint;
cJSON *maxRowsPerFileBlock = cJSON_GetObjectItem(root, "maxRowsPerFileBlock");
if (!maxRowsPerFileBlock || maxRowsPerFileBlock->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, maxRowsPerFileBlock not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.maxRowsPerFileBlock = (int32_t)maxRowsPerFileBlock->valueint;
cJSON *precision = cJSON_GetObjectItem(root, "precision");
if (!precision || precision->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, precision not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.precision = (int8_t)precision->valueint;
cJSON *compression = cJSON_GetObjectItem(root, "compression");
if (!compression || compression->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, compression not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.compression = (int8_t)compression->valueint;
cJSON *walLevel = cJSON_GetObjectItem(root, "walLevel");
if (!walLevel || walLevel->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, walLevel not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.walLevel = (int8_t)walLevel->valueint;
cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsync");
if (!walLevel || walLevel->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, fsyncPeriod not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.fsyncPeriod = (int32_t)fsyncPeriod->valueint;
cJSON *wals = cJSON_GetObjectItem(root, "wals");
if (!wals || wals->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, wals not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.wals = (int8_t)wals->valueint;
cJSON *vgReplica = cJSON_GetObjectItem(root, "replica");
if (!vgReplica || vgReplica->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, replica not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.vgReplica = (int8_t)vgReplica->valueint;
cJSON *dbReplica = cJSON_GetObjectItem(root, "dbReplica");
if (!dbReplica || dbReplica->type != cJSON_Number) {
vWarn("vgId:%d, failed to read %s, dbReplica not found", pVnode->vgId, file);
vnodeMsg.cfg.dbReplica = vnodeMsg.cfg.vgReplica;
vnodeMsg.cfg.vgCfgVersion = 0;
} else {
vnodeMsg.cfg.dbReplica = (int8_t)dbReplica->valueint;
}
cJSON *quorum = cJSON_GetObjectItem(root, "quorum");
if (!quorum || quorum->type != cJSON_Number) {
vError("vgId: %d, failed to read %s, quorum not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
vnodeMsg.cfg.quorum = (int8_t)quorum->valueint;
cJSON *update = cJSON_GetObjectItem(root, "update");
if (!update || update->type != cJSON_Number) {
vWarn("vgId: %d, failed to read %s, update not found", pVnode->vgId, file);
vnodeMsg.cfg.update = 0;
vnodeMsg.cfg.vgCfgVersion = 0;
} else {
vnodeMsg.cfg.update = (int8_t)update->valueint;
}
cJSON *cacheLastRow = cJSON_GetObjectItem(root, "cacheLastRow");
if (!cacheLastRow || cacheLastRow->type != cJSON_Number) {
vWarn("vgId: %d, failed to read %s, cacheLastRow not found", pVnode->vgId, file);
vnodeMsg.cfg.cacheLastRow = 0;
vnodeMsg.cfg.vgCfgVersion = 0;
} else {
vnodeMsg.cfg.cacheLastRow = (int8_t)cacheLastRow->valueint;
}
cJSON *dbType = cJSON_GetObjectItem(root, "dbType");
if (!dbType || dbType->type != cJSON_Number) {
vWarn("vgId: %d, failed to read %s, dbType not found", pVnode->vgId, file);
vnodeMsg.cfg.dbType = 0;
} else {
vnodeMsg.cfg.dbType = (int8_t)dbType->valueint;
}
cJSON *nodeInfos = cJSON_GetObjectItem(root, "nodeInfos");
if (!nodeInfos || nodeInfos->type != cJSON_Array) {
vError("vgId:%d, failed to read %s, nodeInfos not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
int size = cJSON_GetArraySize(nodeInfos);
if (size != vnodeMsg.cfg.vgReplica) {
vError("vgId:%d, failed to read %s, nodeInfos size not matched", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
for (int i = 0; i < size; ++i) {
cJSON *nodeInfo = cJSON_GetArrayItem(nodeInfos, i);
if (nodeInfo == NULL) continue;
SVnodeDesc *node = &vnodeMsg.nodes[i];
cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "nodeId");
if (!nodeId || nodeId->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, nodeId not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
node->nodeId = (int32_t)nodeId->valueint;
cJSON *nodeEp = cJSON_GetObjectItem(nodeInfo, "nodeEp");
if (!nodeEp || nodeEp->type != cJSON_String || nodeEp->valuestring == NULL) {
vError("vgId:%d, failed to read %s, nodeFqdn not found", pVnode->vgId, file);
goto PARSE_VCFG_ERROR;
}
tstrncpy(node->nodeEp, nodeEp->valuestring, TSDB_EP_LEN);
char nodeEpStr[TSDB_EP_LEN];
vnodeGetDnodeEp(node->nodeId, nodeEpStr, NULL, NULL);
bool changed = (strcmp(node->nodeEp, nodeEpStr) != 0);
if (changed) {
tstrncpy(node->nodeEp, nodeEpStr, TSDB_EP_LEN);
nodeChanged = changed;
}
}
ret = TSDB_CODE_SUCCESS;
PARSE_VCFG_ERROR:
if (content != NULL) free(content);
if (root != NULL) cJSON_Delete(root);
if (fp != NULL) fclose(fp);
if (nodeChanged) {
vnodeWriteCfg(&vnodeMsg);
}
if (ret == TSDB_CODE_SUCCESS) {
vnodeLoadCfg(pVnode, &vnodeMsg);
}
terrno = 0;
return ret;
}
int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) {
char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, pMsg->cfg.vgId);
FILE *fp = fopen(file, "w");
if (!fp) {
vError("vgId:%d, failed to write %s error:%s", pMsg->cfg.vgId, file, strerror(errno));
terrno = TAOS_SYSTEM_ERROR(errno);
return terrno;
}
int32_t len = 0;
int32_t maxLen = 1000;
char * content = calloc(1, maxLen + 1);
len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"db\": \"%s\",\n", pMsg->db);
len += snprintf(content + len, maxLen - len, " \"cfgVersion\": %d,\n", pMsg->cfg.dbCfgVersion);
len += snprintf(content + len, maxLen - len, " \"vgCfgVersion\": %d,\n", pMsg->cfg.vgCfgVersion);
len += snprintf(content + len, maxLen - len, " \"cacheBlockSize\": %d,\n", pMsg->cfg.cacheBlockSize);
len += snprintf(content + len, maxLen - len, " \"totalBlocks\": %d,\n", pMsg->cfg.totalBlocks);
len += snprintf(content + len, maxLen - len, " \"daysPerFile\": %d,\n", pMsg->cfg.daysPerFile);
len += snprintf(content + len, maxLen - len, " \"daysToKeep\": %d,\n", pMsg->cfg.daysToKeep);
len += snprintf(content + len, maxLen - len, " \"daysToKeep1\": %d,\n", pMsg->cfg.daysToKeep1);
len += snprintf(content + len, maxLen - len, " \"daysToKeep2\": %d,\n", pMsg->cfg.daysToKeep2);
len += snprintf(content + len, maxLen - len, " \"minRowsPerFileBlock\": %d,\n", pMsg->cfg.minRowsPerFileBlock);
len += snprintf(content + len, maxLen - len, " \"maxRowsPerFileBlock\": %d,\n", pMsg->cfg.maxRowsPerFileBlock);
len += snprintf(content + len, maxLen - len, " \"precision\": %d,\n", pMsg->cfg.precision);
len += snprintf(content + len, maxLen - len, " \"compression\": %d,\n", pMsg->cfg.compression);
len += snprintf(content + len, maxLen - len, " \"walLevel\": %d,\n", pMsg->cfg.walLevel);
len += snprintf(content + len, maxLen - len, " \"fsync\": %d,\n", pMsg->cfg.fsyncPeriod);
len += snprintf(content + len, maxLen - len, " \"replica\": %d,\n", pMsg->cfg.vgReplica);
len += snprintf(content + len, maxLen - len, " \"dbReplica\": %d,\n", pMsg->cfg.dbReplica);
len += snprintf(content + len, maxLen - len, " \"wals\": %d,\n", pMsg->cfg.wals);
len += snprintf(content + len, maxLen - len, " \"quorum\": %d,\n", pMsg->cfg.quorum);
len += snprintf(content + len, maxLen - len, " \"update\": %d,\n", pMsg->cfg.update);
len += snprintf(content + len, maxLen - len, " \"cacheLastRow\": %d,\n", pMsg->cfg.cacheLastRow);
len += snprintf(content + len, maxLen - len, " \"dbType\": %d,\n", pMsg->cfg.dbType);
len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n");
for (int32_t i = 0; i < pMsg->cfg.vgReplica; i++) {
SVnodeDesc *node = &pMsg->nodes[i];
vnodeGetDnodeEp(node->nodeId, node->nodeEp, NULL, NULL);
len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", node->nodeId);
len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", node->nodeEp);
if (i < pMsg->cfg.vgReplica - 1) {
len += snprintf(content + len, maxLen - len, " },{\n");
} else {
len += snprintf(content + len, maxLen - len, " }]\n");
}
}
len += snprintf(content + len, maxLen - len, "}\n");
fwrite(content, 1, len, fp);
taosFsync(fileno(fp));
fclose(fp);
free(content);
terrno = 0;
vInfo("vgId:%d, successed to write %s", pMsg->cfg.vgId, file);
#endif
return TSDB_CODE_SUCCESS;
}

369
source/server/vnode/src/vnodeFile.c Normal file
View File

@ -0,0 +1,369 @@
/*
* 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 "cJSON.h"
#include "vnodeFile.h"
int32_t vnodeReadCfg(int32_t vgId, SVnodeCfg *pCfg) {
int32_t ret = TSDB_CODE_VND_APP_ERROR;
int32_t len = 0;
int maxLen = 1000;
char *content = calloc(1, maxLen + 1);
cJSON *root = NULL;
FILE *fp = NULL;
char file[PATH_MAX + 30] = {0};
sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, vgId);
fp = fopen(file, "r");
if (!fp) {
vError("vgId:%d, failed to open vnode cfg file:%s to read, error:%s", vgId, file, strerror(errno));
ret = TAOS_SYSTEM_ERROR(errno);
goto PARSE_VCFG_ERROR;
}
len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) {
vError("vgId:%d, failed to read %s, content is null", vgId, file);
goto PARSE_VCFG_ERROR;
}
content[len] = 0;
root = cJSON_Parse(content);
if (root == NULL) {
vError("vgId:%d, failed to read %s, invalid json format", vgId, file);
goto PARSE_VCFG_ERROR;
}
cJSON *db = cJSON_GetObjectItem(root, "db");
if (!db || db->type != cJSON_String || db->valuestring == NULL) {
vError("vgId:%d, failed to read %s, db not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
tstrncpy(pCfg->db, db->valuestring, sizeof(pCfg->db));
cJSON *dropped = cJSON_GetObjectItem(root, "dropped");
if (!dropped || dropped->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, dropped not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->dropped = (int32_t)dropped->valueint;
cJSON *cacheBlockSize = cJSON_GetObjectItem(root, "cacheBlockSize");
if (!cacheBlockSize || cacheBlockSize->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, cacheBlockSize not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.cacheBlockSize = (int32_t)cacheBlockSize->valueint;
cJSON *totalBlocks = cJSON_GetObjectItem(root, "totalBlocks");
if (!totalBlocks || totalBlocks->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, totalBlocks not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.totalBlocks = (int32_t)totalBlocks->valueint;
cJSON *daysPerFile = cJSON_GetObjectItem(root, "daysPerFile");
if (!daysPerFile || daysPerFile->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysPerFile not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.daysPerFile = (int32_t)daysPerFile->valueint;
cJSON *daysToKeep0 = cJSON_GetObjectItem(root, "daysToKeep0");
if (!daysToKeep0 || daysToKeep0->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysToKeep0 not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.daysToKeep0 = (int32_t)daysToKeep0->valueint;
cJSON *daysToKeep1 = cJSON_GetObjectItem(root, "daysToKeep1");
if (!daysToKeep1 || daysToKeep1->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysToKeep1 not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.daysToKeep1 = (int32_t)daysToKeep1->valueint;
cJSON *daysToKeep2 = cJSON_GetObjectItem(root, "daysToKeep2");
if (!daysToKeep2 || daysToKeep2->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, daysToKeep2 not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.daysToKeep2 = (int32_t)daysToKeep2->valueint;
cJSON *minRowsPerFileBlock = cJSON_GetObjectItem(root, "minRowsPerFileBlock");
if (!minRowsPerFileBlock || minRowsPerFileBlock->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, minRowsPerFileBlock not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.minRowsPerFileBlock = (int32_t)minRowsPerFileBlock->valueint;
cJSON *maxRowsPerFileBlock = cJSON_GetObjectItem(root, "maxRowsPerFileBlock");
if (!maxRowsPerFileBlock || maxRowsPerFileBlock->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, maxRowsPerFileBlock not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.maxRowsPerFileBlock = (int32_t)maxRowsPerFileBlock->valueint;
cJSON *precision = cJSON_GetObjectItem(root, "precision");
if (!precision || precision->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, precision not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.precision = (int8_t)precision->valueint;
cJSON *compression = cJSON_GetObjectItem(root, "compression");
if (!compression || compression->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, compression not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.compression = (int8_t)compression->valueint;
cJSON *update = cJSON_GetObjectItem(root, "update");
if (!update || update->type != cJSON_Number) {
vError("vgId: %d, failed to read %s, update not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.update = (int8_t)update->valueint;
cJSON *cacheLastRow = cJSON_GetObjectItem(root, "cacheLastRow");
if (!cacheLastRow || cacheLastRow->type != cJSON_Number) {
vError("vgId: %d, failed to read %s, cacheLastRow not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->tsdb.cacheLastRow = (int8_t)cacheLastRow->valueint;
cJSON *walLevel = cJSON_GetObjectItem(root, "walLevel");
if (!walLevel || walLevel->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, walLevel not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->wal.walLevel = (int8_t)walLevel->valueint;
cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsyncPeriod");
if (!walLevel || walLevel->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, fsyncPeriod not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->wal.fsyncPeriod = (int32_t)fsyncPeriod->valueint;
cJSON *replica = cJSON_GetObjectItem(root, "replica");
if (!replica || replica->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, replica not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->sync.replica = (int8_t)replica->valueint;
cJSON *quorum = cJSON_GetObjectItem(root, "quorum");
if (!quorum || quorum->type != cJSON_Number) {
vError("vgId: %d, failed to read %s, quorum not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
pCfg->sync.quorum = (int8_t)quorum->valueint;
cJSON *nodes = cJSON_GetObjectItem(root, "nodes");
if (!nodes || nodes->type != cJSON_Array) {
vError("vgId:%d, failed to read %s, nodes not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
int size = cJSON_GetArraySize(nodes);
if (size != pCfg->sync.replica) {
vError("vgId:%d, failed to read %s, nodes size not matched", vgId, file);
goto PARSE_VCFG_ERROR;
}
for (int i = 0; i < size; ++i) {
cJSON *nodeInfo = cJSON_GetArrayItem(nodes, i);
if (nodeInfo == NULL) continue;
SNodeInfo *node = &pCfg->sync.nodes[i];
cJSON *port = cJSON_GetObjectItem(nodeInfo, "port");
if (!port || port->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, port not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
node->nodePort = (uint16_t)port->valueint;
cJSON *fqdn = cJSON_GetObjectItem(nodeInfo, "fqdn");
if (!fqdn || fqdn->type != cJSON_String || fqdn->valuestring == NULL) {
vError("vgId:%d, failed to read %s, fqdn not found", vgId, file);
goto PARSE_VCFG_ERROR;
}
tstrncpy(node->nodeFqdn, fqdn->valuestring, TSDB_FQDN_LEN);
}
ret = TSDB_CODE_SUCCESS;
PARSE_VCFG_ERROR:
if (content != NULL) free(content);
if (root != NULL) cJSON_Delete(root);
if (fp != NULL) fclose(fp);
terrno = 0;
return ret;
}
int32_t vnodeWriteCfg(int32_t vgId, SVnodeCfg *pCfg) {
int32_t code = 0;
char file[PATH_MAX + 30] = {0};
sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, vgId);
FILE *fp = fopen(file, "w");
if (!fp) {
vError("vgId:%d, failed to write %s error:%s", vgId, file, strerror(errno));
terrno = TAOS_SYSTEM_ERROR(errno);
return terrno;
}
int32_t len = 0;
int32_t maxLen = 1000;
char *content = calloc(1, maxLen + 1);
len += snprintf(content + len, maxLen - len, "{\n");
// vnode
len += snprintf(content + len, maxLen - len, " \"vgId\": %d,\n", vgId);
len += snprintf(content + len, maxLen - len, " \"db\": \"%s\",\n", pCfg->db);
len += snprintf(content + len, maxLen - len, " \"dropped\": %d,\n", pCfg->dropped);
// tsdb
len += snprintf(content + len, maxLen - len, " \"cacheBlockSize\": %d,\n", pCfg->tsdb.cacheBlockSize);
len += snprintf(content + len, maxLen - len, " \"totalBlocks\": %d,\n", pCfg->tsdb.totalBlocks);
len += snprintf(content + len, maxLen - len, " \"daysPerFile\": %d,\n", pCfg->tsdb.daysPerFile);
len += snprintf(content + len, maxLen - len, " \"daysToKeep0\": %d,\n", pCfg->tsdb.daysToKeep0);
len += snprintf(content + len, maxLen - len, " \"daysToKeep1\": %d,\n", pCfg->tsdb.daysToKeep1);
len += snprintf(content + len, maxLen - len, " \"daysToKeep2\": %d,\n", pCfg->tsdb.daysToKeep2);
len += snprintf(content + len, maxLen - len, " \"minRowsPerFileBlock\": %d,\n", pCfg->tsdb.minRowsPerFileBlock);
len += snprintf(content + len, maxLen - len, " \"maxRowsPerFileBlock\": %d,\n", pCfg->tsdb.maxRowsPerFileBlock);
len += snprintf(content + len, maxLen - len, " \"precision\": %d,\n", pCfg->tsdb.precision);
len += snprintf(content + len, maxLen - len, " \"compression\": %d,\n", pCfg->tsdb.compression);
len += snprintf(content + len, maxLen - len, " \"cacheLastRow\": %d,\n", pCfg->tsdb.cacheLastRow);
len += snprintf(content + len, maxLen - len, " \"update\": %d,\n", pCfg->tsdb.update);
// wal
len += snprintf(content + len, maxLen - len, " \"walLevel\": %d,\n", pCfg->wal.walLevel);
len += snprintf(content + len, maxLen - len, " \"fsyncPeriod\": %d,\n", pCfg->wal.fsyncPeriod);
// sync
len += snprintf(content + len, maxLen - len, " \"quorum\": %d,\n", pCfg->sync.quorum);
len += snprintf(content + len, maxLen - len, " \"replica\": %d,\n", pCfg->sync.replica);
len += snprintf(content + len, maxLen - len, " \"nodes\": [{\n");
for (int32_t i = 0; i < pCfg->sync.replica; i++) {
SNodeInfo *node = &pCfg->sync.nodes[i];
len += snprintf(content + len, maxLen - len, " \"port\": %u,\n", node->nodePort);
len += snprintf(content + len, maxLen - len, " \"fqdn\": \"%s\"\n", node->nodeFqdn);
if (i < pCfg->sync.replica - 1) {
len += snprintf(content + len, maxLen - len, " },{\n");
} else {
len += snprintf(content + len, maxLen - len, " }]\n");
}
}
len += snprintf(content + len, maxLen - len, "}\n");
fwrite(content, 1, len, fp);
taosFsyncFile(fileno(fp));
fclose(fp);
free(content);
terrno = 0;
vInfo("vgId:%d, successed to write %s", vgId, file);
return TSDB_CODE_SUCCESS;
}
int32_t vnodeReadTerm(int32_t vgId, SSyncServerState *pState){
#if 0
int32_t len = 0;
int32_t maxLen = 100;
char * content = calloc(1, maxLen + 1);
cJSON * root = NULL;
FILE * fp = NULL;
terrno = TSDB_CODE_VND_INVALID_VRESION_FILE;
char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
fp = fopen(file, "r");
if (!fp) {
if (errno != ENOENT) {
vError("vgId:%d, failed to read %s, error:%s", pVnode->vgId, file, strerror(errno));
terrno = TAOS_SYSTEM_ERROR(errno);
} else {
terrno = TSDB_CODE_SUCCESS;
}
goto PARSE_VER_ERROR;
}
len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) {
vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file);
goto PARSE_VER_ERROR;
}
root = cJSON_Parse(content);
if (root == NULL) {
vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file);
goto PARSE_VER_ERROR;
}
cJSON *ver = cJSON_GetObjectItem(root, "version");
if (!ver || ver->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, version not found", pVnode->vgId, file);
goto PARSE_VER_ERROR;
}
#if 0
pVnode->version = (uint64_t)ver->valueint;
terrno = TSDB_CODE_SUCCESS;
vInfo("vgId:%d, read %s successfully, fver:%" PRIu64, pVnode->vgId, file, pVnode->version);
#endif
PARSE_VER_ERROR:
if (content != NULL) free(content);
if (root != NULL) cJSON_Delete(root);
if (fp != NULL) fclose(fp);
return terrno;
#endif
return 0;
}
int32_t vnodeWriteTerm(int32_t vgid, SSyncServerState *pState) {
#if 0
char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
FILE *fp = fopen(file, "w");
if (!fp) {
vError("vgId:%d, failed to write %s, reason:%s", pVnode->vgId, file, strerror(errno));
return -1;
}
int32_t len = 0;
int32_t maxLen = 100;
char * content = calloc(1, maxLen + 1);
#if 0
len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"version\": %" PRIu64 "\n", pVnode->fversion);
len += snprintf(content + len, maxLen - len, "}\n");
#endif
fwrite(content, 1, len, fp);
taosFsyncFile(fileno(fp));
fclose(fp);
free(content);
terrno = 0;
// vInfo("vgId:%d, successed to write %s, fver:%" PRIu64, pVnode->vgId, file, pVnode->fversion);
#endif
return TSDB_CODE_SUCCESS;
}

916
source/server/vnode/src/vnodeInt.c
View File

@ -14,937 +14,41 @@
*/ */
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "tglobal.h" #include "os.h"
#include "ttimer.h"
#include "thash.h"
// #include "query.h"
#include "vnodeCfg.h"
#include "vnodeMgmt.h"
#include "vnodeRead.h"
#include "vnodeStatus.h"
#include "vnodeVersion.h"
#include "vnodeWorker.h"
#include "vnodeWrite.h"
#include "tstep.h" #include "tstep.h"
#include "vnodeMain.h"
#include "vnodeMgmt.h" #include "vnodeMgmt.h"
#include "vnodeRead.h" #include "vnodeRead.h"
#include "vnodeWorker.h"
#include "vnodeWrite.h" #include "vnodeWrite.h"
typedef struct {
pthread_t thread;
int32_t threadIndex;
int32_t failed;
int32_t opened;
int32_t vnodeNum;
int32_t * vnodeList;
} SOpenVnodeThread;
static struct { static struct {
SSteps *steps; SSteps *steps;
SVnodeFp fp; SVnodeFp fp;
void * timer; } tsVint;
SHashObj *hash;
int32_t openVnodes;
int32_t totalVnodes;
void (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
} tsVnode;
void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port) { void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port) {
return (*tsVnode.fp.GetDnodeEp)(dnodeId, ep, fqdn, port); return (*tsVint.fp.GetDnodeEp)(dnodeId, ep, fqdn, port);
} }
void vnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg) { void vnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg) {
(*tsVnode.fp.SendMsgToDnode)(epSet, rpcMsg); (*tsVint.fp.SendMsgToDnode)(epSet, rpcMsg);
} }
void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg) { return (*tsVnode.fp.SendMsgToMnode)(rpcMsg); } void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg) { return (*tsVint.fp.SendMsgToMnode)(rpcMsg); }
static void vnodeIncRef(void *ptNode) {
assert(ptNode != NULL);
SVnode **ppVnode = (SVnode **)ptNode;
assert(ppVnode);
assert(*ppVnode);
SVnode *pVnode = *ppVnode;
atomic_add_fetch_32(&pVnode->refCount, 1);
vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
}
SVnode *vnodeAcquire(int32_t vgId) {
SVnode *pVnode = NULL;
// taosHashGetClone(tsVnode.hash, &vgId, sizeof(int32_t), vnodeIncRef, (void*)&pVnode);
if (pVnode == NULL) {
terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
vDebug("vgId:%d, not exist", vgId);
return NULL;
}
return pVnode;
}
SVnode *vnodeAcquireNotClose(int32_t vgId) {
// SVnode *pVnode = vnodeAcquire(vgId);
// if (pVnode != NULL && pVnode->preClose == 1) {
// vnodeRelease(pVnode);
// terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
// vDebug("vgId:%d, not exist, pre closing", vgId);
// return NULL;
// }
// return pVnode;
return NULL;
}
void vnodeRelease(SVnode *pVnode) {
if (pVnode == NULL) return;
int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
int32_t vgId = pVnode->vgId;
vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode);
assert(refCount >= 0);
if (refCount <= 0) {
vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode);
vnodeProcessDestroyTask(pVnode);
int32_t count = taosHashGetSize(tsVnode.hash);
vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count);
}
}
static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno);
int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) {
int32_t code;
SVnode *pVnode = vnodeAcquire(pVnodeCfg->cfg.vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, vnode already exist, refCount:%d pVnode:%p", pVnodeCfg->cfg.vgId, pVnode->refCount, pVnode);
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
}
#if 0
if (tfsMkdir("vnode") < 0) {
vError("vgId:%d, failed to create vnode dir, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno));
return terrno;
}
char vnodeDir[TSDB_FILENAME_LEN] = "\0";
snprintf(vnodeDir, TSDB_FILENAME_LEN, "/vnode/vnode%d", pVnodeCfg->cfg.vgId);
if (tfsMkdir(vnodeDir) < 0) {
vError("vgId:%d, failed to create vnode dir %s, reason:%s", pVnodeCfg->cfg.vgId, vnodeDir, strerror(errno));
return terrno;
}
code = vnodeWriteCfg(pVnodeCfg);
if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to save vnode cfg, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(code));
return code;
}
if (tsdbCreateRepo(pVnodeCfg->cfg.vgId) < 0) {
vError("vgId:%d, failed to create tsdb in vnode, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno));
return TSDB_CODE_VND_INIT_FAILED;
}
#endif
vInfo("vgId:%d, vnode dir is created, walLevel:%d fsyncPeriod:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel,
pVnodeCfg->cfg.fsyncPeriod);
code = vnodeOpen(pVnodeCfg->cfg.vgId);
return code;
}
int32_t vnodeSync(int32_t vgId) {
#if 0
SVnode *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) {
vDebug("vgId:%d, failed to sync, vnode not find", vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
if (pVnode->role == TAOS_SYNC_ROLE_SLAVE) {
vInfo("vgId:%d, vnode will sync, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
pVnode->version = 0;
pVnode->fversion = 0;
walResetVersion(pVnode->wal, pVnode->fversion);
syncRecover(pVnode->sync);
}
vnodeRelease(pVnode);
#endif
return TSDB_CODE_SUCCESS;
}
int32_t vnodeDrop(int32_t vgId) {
SVnode *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) {
vDebug("vgId:%d, failed to drop, vnode not find", vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
if (pVnode->dropped) {
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
}
vInfo("vgId:%d, vnode will be dropped, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
pVnode->dropped = 1;
vnodeRelease(pVnode);
vnodeProcessCleanupTask(pVnode);
return TSDB_CODE_SUCCESS;
}
int32_t vnodeCompact(int32_t vgId) {
#if 0
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, compact vnode msg is received", vgId);
// not care success or not
tsdbCompact(((SVnode *)pVnode)->tsdb);
vnodeRelease(pVnode);
} else {
vInfo("vgId:%d, vnode not exist, can't compact it", vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
#endif
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeAlterImp(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg) {
#if 0
STsdbCfg tsdbCfg = pVnode->tsdbCfg;
SSyncCfg syncCfg = pVnode->syncCfg;
int32_t dbCfgVersion = pVnode->dbCfgVersion;
int32_t vgCfgVersion = pVnode->vgCfgVersion;
int32_t code = vnodeWriteCfg(pVnodeCfg);
if (code != TSDB_CODE_SUCCESS) {
pVnode->dbCfgVersion = dbCfgVersion;
pVnode->vgCfgVersion = vgCfgVersion;
pVnode->syncCfg = syncCfg;
pVnode->tsdbCfg = tsdbCfg;
return code;
}
code = vnodeReadCfg(pVnode);
if (code != TSDB_CODE_SUCCESS) {
pVnode->dbCfgVersion = dbCfgVersion;
pVnode->vgCfgVersion = vgCfgVersion;
pVnode->syncCfg = syncCfg;
pVnode->tsdbCfg = tsdbCfg;
return code;
}
code = walAlter(pVnode->wal, &pVnode->walCfg);
if (code != TSDB_CODE_SUCCESS) {
pVnode->dbCfgVersion = dbCfgVersion;
pVnode->vgCfgVersion = vgCfgVersion;
pVnode->syncCfg = syncCfg;
pVnode->tsdbCfg = tsdbCfg;
return code;
}
bool tsdbCfgChanged = (memcmp(&tsdbCfg, &pVnode->tsdbCfg, sizeof(STsdbCfg)) != 0);
bool syncCfgChanged = (memcmp(&syncCfg, &pVnode->syncCfg, sizeof(SSyncCfg)) != 0);
vDebug("vgId:%d, tsdbchanged:%d syncchanged:%d while alter vnode", pVnode->vgId, tsdbCfgChanged, syncCfgChanged);
if (tsdbCfgChanged || syncCfgChanged) {
// vnode in non-ready state and still needs to return success instead of TSDB_CODE_VND_INVALID_STATUS
// dbCfgVersion can be corrected by status msg
if (syncCfgChanged) {
if (!vnodeSetUpdatingStatus(pVnode)) {
vDebug("vgId:%d, vnode is not ready, do alter operation later", pVnode->vgId);
pVnode->dbCfgVersion = dbCfgVersion;
pVnode->vgCfgVersion = vgCfgVersion;
pVnode->syncCfg = syncCfg;
pVnode->tsdbCfg = tsdbCfg;
return TSDB_CODE_SUCCESS;
}
code = syncReconfig(pVnode->sync, &pVnode->syncCfg);
if (code != TSDB_CODE_SUCCESS) {
pVnode->dbCfgVersion = dbCfgVersion;
pVnode->vgCfgVersion = vgCfgVersion;
pVnode->syncCfg = syncCfg;
pVnode->tsdbCfg = tsdbCfg;
vnodeSetReadyStatus(pVnode);
return code;
}
}
if (tsdbCfgChanged && pVnode->tsdb) {
code = tsdbConfigRepo(pVnode->tsdb, &pVnode->tsdbCfg);
if (code != TSDB_CODE_SUCCESS) {
pVnode->dbCfgVersion = dbCfgVersion;
pVnode->vgCfgVersion = vgCfgVersion;
pVnode->syncCfg = syncCfg;
pVnode->tsdbCfg = tsdbCfg;
vnodeSetReadyStatus(pVnode);
return code;
}
}
vnodeSetReadyStatus(pVnode);
}
#endif
return 0;
}
int32_t vnodeAlter(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg) {
#if 0
vDebug("vgId:%d, current dbCfgVersion:%d vgCfgVersion:%d, input dbCfgVersion:%d vgCfgVersion:%d", pVnode->vgId,
pVnode->dbCfgVersion, pVnode->vgCfgVersion, pVnodeCfg->cfg.dbCfgVersion, pVnodeCfg->cfg.vgCfgVersion);
if (pVnode->dbCfgVersion == pVnodeCfg->cfg.dbCfgVersion && pVnode->vgCfgVersion == pVnodeCfg->cfg.vgCfgVersion) {
vDebug("vgId:%d, cfg not change", pVnode->vgId);
return TSDB_CODE_SUCCESS;
}
int32_t code = vnodeAlterImp(pVnode, pVnodeCfg);
if (code != 0) {
vError("vgId:%d, failed to alter vnode, code:0x%x", pVnode->vgId, code);
} else {
vDebug("vgId:%d, vnode is altered", pVnode->vgId);
}
return code;
#endif
return 0;
}
static void vnodeFindWalRootDir(int32_t vgId, char *walRootDir) {
#if 0
char vnodeDir[TSDB_FILENAME_LEN] = "\0";
snprintf(vnodeDir, TSDB_FILENAME_LEN, "/vnode/vnode%d/wal", vgId);
TDIR *tdir = tfsOpendir(vnodeDir);
if (!tdir) return;
const TFILE *tfile = tfsReaddir(tdir);
if (!tfile) {
tfsClosedir(tdir);
return;
}
sprintf(walRootDir, "%s/vnode/vnode%d", TFS_DISK_PATH(tfile->level, tfile->id), vgId);
tfsClosedir(tdir);
#endif
}
int32_t vnodeOpen(int32_t vgId) {
#if 0
char temp[TSDB_FILENAME_LEN * 3];
char rootDir[TSDB_FILENAME_LEN * 2];
char walRootDir[TSDB_FILENAME_LEN * 2] = {0};
snprintf(rootDir, TSDB_FILENAME_LEN * 2, "%s/vnode%d", tsVnodeDir, vgId);
SVnode *pVnode = calloc(sizeof(SVnode), 1);
if (pVnode == NULL) {
vError("vgId:%d, failed to open vnode since no enough memory", vgId);
return TAOS_SYSTEM_ERROR(errno);
}
atomic_add_fetch_32(&pVnode->refCount, 1);
pVnode->vgId = vgId;
pVnode->fversion = 0;
pVnode->version = 0;
pVnode->tsdbCfg.tsdbId = pVnode->vgId;
pVnode->rootDir = strdup(rootDir);
pVnode->accessState = TSDB_VN_ALL_ACCCESS;
tsem_init(&pVnode->sem, 0, 0);
pthread_mutex_init(&pVnode->statusMutex, NULL);
vnodeSetInitStatus(pVnode);
tsdbIncCommitRef(pVnode->vgId);
int32_t code = vnodeReadCfg(pVnode);
if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to read config file, set cfgVersion to 0", pVnode->vgId);
vnodeCleanUp(pVnode);
return 0;
}
code = vnodeReadVersion(pVnode);
if (code != TSDB_CODE_SUCCESS) {
pVnode->version = 0;
vError("vgId:%d, failed to read file version, generate it from data file", pVnode->vgId);
// Allow vnode start even when read file version fails, set file version as wal version or zero
// vnodeCleanUp(pVnode);
// return code;
}
pVnode->fversion = pVnode->version;
pVnode->pWriteQ = vnodeAllocWriteQueue(pVnode);
pVnode->pQueryQ = vnodeAllocQueryQueue(pVnode);
pVnode->pFetchQ = vnodeAllocFetchQueue(pVnode);
if (pVnode->pWriteQ == NULL || pVnode->pQueryQ == NULL || pVnode->pFetchQ == NULL) {
vnodeCleanUp(pVnode);
return terrno;
}
STsdbAppH appH = {0};
appH.appH = (void *)pVnode;
appH.notifyStatus = vnodeProcessTsdbStatus;
appH.cqH = pVnode->cq;
appH.cqCreateFunc = cqCreate;
appH.cqDropFunc = cqDrop;
terrno = 0;
pVnode->tsdb = tsdbOpenRepo(&(pVnode->tsdbCfg), &appH);
if (pVnode->tsdb == NULL) {
vnodeCleanUp(pVnode);
return terrno;
} else if (tsdbGetState(pVnode->tsdb) != TSDB_STATE_OK) {
vError("vgId:%d, failed to open tsdb(state: %d), replica:%d reason:%s", pVnode->vgId, tsdbGetState(pVnode->tsdb),
pVnode->syncCfg.replica, tstrerror(terrno));
if (pVnode->syncCfg.replica <= 1) {
vnodeCleanUp(pVnode);
return TSDB_CODE_VND_INVALID_TSDB_STATE;
} else {
pVnode->fversion = 0;
pVnode->version = 0;
}
}
// walRootDir for wal & syncInfo.path (not empty dir of /vnode/vnode{pVnode->vgId}/wal)
vnodeFindWalRootDir(pVnode->vgId, walRootDir);
if (walRootDir[0] == 0) {
int level = -1, id = -1;
tfsAllocDisk(TFS_PRIMARY_LEVEL, &level, &id);
if (level < 0 || id < 0) {
vnodeCleanUp(pVnode);
return terrno;
}
sprintf(walRootDir, "%s/vnode/vnode%d", TFS_DISK_PATH(level, id), vgId);
}
sprintf(temp, "%s/wal", walRootDir);
pVnode->walCfg.vgId = pVnode->vgId;
pVnode->wal = walOpen(temp, &pVnode->walCfg);
if (pVnode->wal == NULL) {
vnodeCleanUp(pVnode);
return terrno;
}
walRestore(pVnode->wal, pVnode, (FWalWrite)vnodeProcessWalMsg);
if (pVnode->version == 0) {
pVnode->fversion = 0;
pVnode->version = walGetVersion(pVnode->wal);
}
code = tsdbSyncCommit(pVnode->tsdb);
if (code != 0) {
vError("vgId:%d, failed to commit after restore from wal since %s", pVnode->vgId, tstrerror(code));
vnodeCleanUp(pVnode);
return code;
}
walRemoveAllOldFiles(pVnode->wal);
walRenew(pVnode->wal);
pVnode->qMgmt = qOpenQueryMgmt(pVnode->vgId);
if (pVnode->qMgmt == NULL) {
vnodeCleanUp(pVnode);
return terrno;
}
pVnode->events = NULL;
vDebug("vgId:%d, vnode is opened in %s - %s, pVnode:%p", pVnode->vgId, rootDir, walRootDir, pVnode);
taosHashPut(tsVnode.hash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnode *));
vnodeSetReadyStatus(pVnode);
pVnode->role = TAOS_SYNC_ROLE_MASTER;
#endif
return TSDB_CODE_SUCCESS;
}
int32_t vnodeClose(int32_t vgId) {
SVnode *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) return 0;
if (pVnode->dropped) {
vnodeRelease(pVnode);
return 0;
}
// pVnode->preClose = 1;
vDebug("vgId:%d, vnode will be closed, pVnode:%p", pVnode->vgId, pVnode);
vnodeRelease(pVnode);
vnodeCleanUp(pVnode);
return 0;
}
void vnodeDestroy(SVnode *pVnode) {
#if 0
int32_t code = 0;
int32_t vgId = pVnode->vgId;
if (pVnode->qMgmt) {
qCleanupQueryMgmt(pVnode->qMgmt);
pVnode->qMgmt = NULL;
}
if (pVnode->wal) {
walStop(pVnode->wal);
}
if (pVnode->tsdb) {
// the deleted vnode does not need to commit, so as to speed up the deletion
int toCommit = 1;
if (pVnode->dropped) toCommit = 0;
code = tsdbCloseRepo(pVnode->tsdb, toCommit);
pVnode->tsdb = NULL;
}
// stop continuous query
if (pVnode->cq) {
void *cq = pVnode->cq;
pVnode->cq = NULL;
cqClose(cq);
}
if (pVnode->wal) {
if (code != 0) {
vError("vgId:%d, failed to commit while close tsdb repo, keep wal", pVnode->vgId);
} else {
walRemoveAllOldFiles(pVnode->wal);
}
walClose(pVnode->wal);
pVnode->wal = NULL;
}
if (pVnode->pWriteQ) {
vnodeFreeWriteQueue(pVnode->pWriteQ);
pVnode->pWriteQ = NULL;
}
if (pVnode->pQueryQ) {
vnodeFreeQueryQueue(pVnode->pQueryQ);
pVnode->pQueryQ = NULL;
}
if (pVnode->pFetchQ) {
vnodeFreeFetchQueue(pVnode->pFetchQ);
pVnode->pFetchQ = NULL;
}
tfree(pVnode->rootDir);
if (pVnode->dropped) {
char rootDir[TSDB_FILENAME_LEN] = {0};
char stagingDir[TSDB_FILENAME_LEN] = {0};
sprintf(rootDir, "%s/vnode%d", "vnode", vgId);
sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
tfsRename(rootDir, stagingDir);
vnodeProcessBackupTask(pVnode);
// dnodeSendStatusMsgToMnode();
}
tsem_destroy(&pVnode->sem);
pthread_mutex_destroy(&pVnode->statusMutex);
free(pVnode);
tsdbDecCommitRef(vgId);
#endif
}
void vnodeCleanUp(SVnode *pVnode) {
#if 0
vDebug("vgId:%d, vnode will cleanup, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
vnodeSetClosingStatus(pVnode);
taosHashRemove(tsVnode.hash, &pVnode->vgId, sizeof(int32_t));
// stop replication module
if (pVnode->sync > 0) {
int64_t sync = pVnode->sync;
pVnode->sync = -1;
syncStop(sync);
}
vDebug("vgId:%d, vnode is cleaned, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
vnodeRelease(pVnode);
#endif
}
#if 0
static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) {
SVnode *pVnode = arg;
if (eno != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to commit since %s, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, tstrerror(eno),
pVnode->fversion, pVnode->version);
pVnode->isCommiting = 0;
pVnode->isFull = 1;
return 0;
}
if (status == TSDB_STATUS_COMMIT_START) {
pVnode->isCommiting = 1;
pVnode->cversion = pVnode->version;
vInfo("vgId:%d, start commit, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version);
if (!vnodeInInitStatus(pVnode)) {
return walRenew(pVnode->wal);
}
return 0;
}
if (status == TSDB_STATUS_COMMIT_OVER) {
pVnode->isCommiting = 0;
pVnode->isFull = 0;
pVnode->fversion = pVnode->cversion;
vInfo("vgId:%d, commit over, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version);
if (!vnodeInInitStatus(pVnode)) {
walRemoveOneOldFile(pVnode->wal);
}
return vnodeSaveVersion(pVnode);
}
// timer thread callback
if (status == TSDB_STATUS_COMMIT_NOBLOCK) {
qSolveCommitNoBlock(pVnode->tsdb, pVnode->qMgmt);
}
return 0;
}
#endif
static void *vnodeOpenVnode(void *param) {
SOpenVnodeThread *pThread = param;
vDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
setThreadName("vnodeOpenVnode");
for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
int32_t vgId = pThread->vnodeList[v];
char stepDesc[TSDB_STEP_DESC_LEN] = {0};
snprintf(stepDesc, TSDB_STEP_DESC_LEN, "vgId:%d, start to restore, %d of %d have been opened", vgId,
tsVnode.openVnodes, tsVnode.totalVnodes);
// (*vnodeInst()->fp.ReportStartup)("open-vnodes", stepDesc);
if (vnodeOpen(vgId) < 0) {
vError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
pThread->failed++;
} else {
vDebug("vgId:%d, is opened by thread:%d", vgId, pThread->threadIndex);
pThread->opened++;
}
atomic_add_fetch_32(&tsVnode.openVnodes, 1);
}
vDebug("thread:%d, total vnodes:%d, opened:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
pThread->failed);
return NULL;
}
static int32_t vnodeGetVnodeListFromDisk(int32_t vnodeList[], int32_t *numOfVnodes) {
#if 0
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) {
vError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
closedir(dir);
return TSDB_CODE_DND_TOO_MANY_VNODES;
} else {
vnodeList[*numOfVnodes - 1] = vnode;
}
}
}
closedir(dir);
#endif
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeOpenVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t status = vnodeGetVnodeListFromDisk(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
vInfo("failed to get vnode list from disk since code:%d", status);
return status;
}
tsVnode.totalVnodes = numOfVnodes;
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];
}
vInfo("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, vnodeOpenVnode, pThread) != 0) {
vError("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 && taosCheckPthreadValid(pThread->thread)) {
pthread_join(pThread->thread, NULL);
}
openVnodes += pThread->opened;
failedVnodes += pThread->failed;
free(pThread->vnodeList);
}
free(threads);
vInfo("there are total vnodes:%d, opened:%d", numOfVnodes, openVnodes);
if (failedVnodes != 0) {
vError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes);
return -1;
}
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) {
void *pIter = taosHashIterate(tsVnode.hash, NULL);
while (pIter) {
SVnode **pVnode = pIter;
if (*pVnode) {
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES);
continue;
} else {
vnodeList[*numOfVnodes - 1] = (*pVnode)->vgId;
}
}
pIter = taosHashIterate(tsVnode.hash, pIter);
}
return TSDB_CODE_SUCCESS;
}
static void vnodeCleanupVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t code = vnodeGetVnodeList(vnodeList, &numOfVnodes);
if (code != TSDB_CODE_SUCCESS) {
vInfo("failed to get dnode list since code %d", code);
return;
}
for (int32_t i = 0; i < numOfVnodes; ++i) {
vnodeClose(vnodeList[i]);
}
vInfo("total vnodes:%d are all closed", numOfVnodes);
}
static void vnodeInitMsgFp() {
tsVnode.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessWriteMsg;
//mq related
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_CONNECT] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_DISCONNECT] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_ACK] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_RESET] = vnodeProcessWriteMsg;
//mq related end
tsVnode.msgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessReadMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessReadMsg;
//mq related
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_QUERY] = vnodeProcessReadMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_CONSUME] = vnodeProcessReadMsg;
//mq related end
}
void vnodeProcessMsg(SRpcMsg *pMsg) {
if (tsVnode.msgFp[pMsg->msgType]) {
(*tsVnode.msgFp[pMsg->msgType])(pMsg);
} else {
assert(0);
}
}
int32_t vnodeInitMain() {
vnodeInitMsgFp();
tsVnode.timer = taosTmrInit(100, 200, 60000, "VND-TIMER");
if (tsVnode.timer == NULL) {
vError("failed to init vnode timer");
return -1;
}
tsVnode.hash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
if (tsVnode.hash == NULL) {
taosTmrCleanUp(tsVnode.timer);
vError("failed to init vnode mgmt");
return -1;
}
vInfo("vnode main is initialized");
return vnodeOpenVnodes();
}
void vnodeCleanupMain() {
taosTmrCleanUp(tsVnode.timer);
tsVnode.timer = NULL;
vnodeCleanupVnodes();
taosHashCleanup(tsVnode.hash);
tsVnode.hash = NULL;
}
static void vnodeBuildVloadMsg(SVnode *pVnode, SStatusMsg *pStatus) {
#if 0
int64_t totalStorage = 0;
int64_t compStorage = 0;
int64_t pointsWritten = 0;
if (vnodeInClosingStatus(pVnode)) return;
if (pStatus->openVnodes >= TSDB_MAX_VNODES) return;
if (pVnode->tsdb) {
tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage);
}
SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
pLoad->vgId = htonl(pVnode->vgId);
pLoad->dbCfgVersion = htonl(pVnode->dbCfgVersion);
pLoad->vgCfgVersion = htonl(pVnode->vgCfgVersion);
pLoad->totalStorage = htobe64(totalStorage);
pLoad->compStorage = htobe64(compStorage);
pLoad->pointsWritten = htobe64(pointsWritten);
pLoad->vnodeVersion = htobe64(pVnode->version);
pLoad->status = pVnode->status;
pLoad->role = pVnode->role;
pLoad->replica = pVnode->syncCfg.replica;
pLoad->compact = (pVnode->tsdb != NULL) ? tsdbGetCompactState(pVnode->tsdb) : 0;
#endif
}
void vnodeGetStatus(struct SStatusMsg *pStatus) {
void *pIter = taosHashIterate(tsVnode.hash, NULL);
while (pIter) {
SVnode **pVnode = pIter;
if (*pVnode) {
vnodeBuildVloadMsg(*pVnode, pStatus);
}
pIter = taosHashIterate(tsVnode.hash, pIter);
}
}
void vnodeSetAccess(struct SVgroupAccess *pAccess, int32_t numOfVnodes) {
for (int32_t i = 0; i < numOfVnodes; ++i) {
pAccess[i].vgId = htonl(pAccess[i].vgId);
SVnode *pVnode = vnodeAcquireNotClose(pAccess[i].vgId);
if (pVnode != NULL) {
pVnode->accessState = pAccess[i].accessState;
if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) {
vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState);
}
vnodeRelease(pVnode);
}
}
}
void vnodeBackup(int32_t vgId) {
char newDir[TSDB_FILENAME_LEN] = {0};
char stagingDir[TSDB_FILENAME_LEN] = {0};
sprintf(newDir, "%s/vnode%d", "vnode_bak", vgId);
sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
#if 0
if (tsEnableVnodeBak) {
tfsRmdir(newDir);
tfsRename(stagingDir, newDir);
} else {
vInfo("vgId:%d, vnode backup not enabled", vgId);
tfsRmdir(stagingDir);
}
#endif
}
int32_t vnodeInit(SVnodePara para) { int32_t vnodeInit(SVnodePara para) {
tsVnode.fp = para.fp; tsVint.fp = para.fp;
struct SSteps *steps = taosStepInit(8, NULL); struct SSteps *steps = taosStepInit(8, NULL);
if (steps == NULL) return -1; if (steps == NULL) return -1;
taosStepAdd(steps, "vnode-main", vnodeInitMain, vnodeCleanupMain); taosStepAdd(steps, "vnode-main", vnodeInitMain, vnodeCleanupMain);
taosStepAdd(steps, "vnode-worker", vnodeInitWorker, vnodeCleanupWorker);
taosStepAdd(steps, "vnode-read", vnodeInitRead, vnodeCleanupRead); taosStepAdd(steps, "vnode-read", vnodeInitRead, vnodeCleanupRead);
taosStepAdd(steps, "vnode-mgmt", vnodeInitMgmt, vnodeCleanupMgmt); taosStepAdd(steps, "vnode-mgmt", vnodeInitMgmt, vnodeCleanupMgmt);
taosStepAdd(steps, "vnode-write", vnodeInitWrite, vnodeCleanupWrite); taosStepAdd(steps, "vnode-write", vnodeInitWrite, vnodeCleanupWrite);
tsVnode.steps = steps; tsVint.steps = steps;
return taosStepExec(tsVnode.steps); return taosStepExec(tsVint.steps);
} }
void vnodeCleanup() { taosStepCleanup(tsVnode.steps); } void vnodeCleanup() { taosStepCleanup(tsVint.steps); }

664
source/server/vnode/src/vnodeMain.c Normal file
View File

@ -0,0 +1,664 @@
/*
* 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 "thash.h"
#include "tthread.h"
#include "vnodeFile.h"
#include "vnodeMain.h"
#include "vnodeMgmt.h"
#include "vnodeRead.h"
#include "vnodeWrite.h"
typedef enum _VN_STATUS {
TAOS_VN_STATUS_INIT = 0,
TAOS_VN_STATUS_READY = 1,
TAOS_VN_STATUS_CLOSING = 2,
TAOS_VN_STATUS_UPDATING = 3
} EVnodeStatus;
char *vnodeStatus[] = {"init", "ready", "closing", "updating"};
typedef struct {
pthread_t *threadId;
int32_t threadIndex;
int32_t failed;
int32_t opened;
int32_t vnodeNum;
int32_t *vnodeList;
} SOpenVnodeThread;
static struct {
SHashObj *hash;
int32_t openVnodes;
int32_t totalVnodes;
void (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
} tsVnode;
static bool vnodeSetInitStatus(SVnode *pVnode) {
pthread_mutex_lock(&pVnode->statusMutex);
pVnode->status = TAOS_VN_STATUS_INIT;
pthread_mutex_unlock(&pVnode->statusMutex);
return true;
}
static bool vnodeSetReadyStatus(SVnode *pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_UPDATING) {
pVnode->status = TAOS_VN_STATUS_READY;
set = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
static bool vnodeSetUpdatingStatus(SVnode *pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_READY) {
pVnode->status = TAOS_VN_STATUS_UPDATING;
set = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
static bool vnodeSetClosingStatus(SVnode *pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY) {
pVnode->status = TAOS_VN_STATUS_CLOSING;
set = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
static bool vnodeInStatus(SVnode *pVnode, EVnodeStatus status) {
bool in = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == status) {
in = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return in;
}
static void vnodeDestroyVnode(SVnode *pVnode) {
int32_t code = 0;
int32_t vgId = pVnode->vgId;
if (pVnode->pQuery) {
// todo
}
if (pVnode->pMeta) {
// todo
}
if (pVnode->pTsdb) {
// todo
}
if (pVnode->pTQ) {
// todo
}
if (pVnode->pWal) {
// todo
}
if (pVnode->allocator) {
// todo
}
if (pVnode->pWriteQ) {
vnodeFreeWriteQueue(pVnode->pWriteQ);
pVnode->pWriteQ = NULL;
}
if (pVnode->pQueryQ) {
vnodeFreeQueryQueue(pVnode->pQueryQ);
pVnode->pQueryQ = NULL;
}
if (pVnode->pFetchQ) {
vnodeFreeFetchQueue(pVnode->pFetchQ);
pVnode->pFetchQ = NULL;
}
if (pVnode->dropped) {
// todo
}
pthread_mutex_destroy(&pVnode->statusMutex);
free(pVnode);
}
static void vnodeCleanupVnode(SVnode *pVnode) {
vnodeSetClosingStatus(pVnode);
taosHashRemove(tsVnode.hash, &pVnode->vgId, sizeof(int32_t));
vnodeRelease(pVnode);
}
static int32_t vnodeOpenVnode(int32_t vgId) {
int32_t code = 0;
SVnode *pVnode = calloc(sizeof(SVnode), 1);
if (pVnode == NULL) {
vError("vgId:%d, failed to open vnode since no enough memory", vgId);
return TAOS_SYSTEM_ERROR(errno);
}
pVnode->vgId = vgId;
pVnode->accessState = TAOS_VN_STATUS_INIT;
pVnode->status = TSDB_VN_ALL_ACCCESS;
pVnode->refCount = 1;
pVnode->role = TAOS_SYNC_ROLE_CANDIDATE;
pthread_mutex_init(&pVnode->statusMutex, NULL);
code = vnodeReadCfg(vgId, &pVnode->cfg);
if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to read config file, set cfgVersion to 0", pVnode->vgId);
pVnode->cfg.dropped = 1;
vnodeCleanupVnode(pVnode);
return 0;
}
code = vnodeReadTerm(vgId, &pVnode->term);
if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to read term file since %s", pVnode->vgId, tstrerror(code));
pVnode->cfg.dropped = 1;
vnodeCleanupVnode(pVnode);
return code;
}
pVnode->pWriteQ = vnodeAllocWriteQueue(pVnode);
pVnode->pQueryQ = vnodeAllocQueryQueue(pVnode);
pVnode->pFetchQ = vnodeAllocFetchQueue(pVnode);
if (pVnode->pWriteQ == NULL || pVnode->pQueryQ == NULL || pVnode->pFetchQ == NULL) {
vnodeCleanupVnode(pVnode);
return terrno;
}
char path[PATH_MAX + 20];
snprintf(path, sizeof(path), "%s/vnode%d/wal", tsVnodeDir, vgId);
pVnode->pWal = walOpen(path, &pVnode->cfg.wal);
if (pVnode->pWal == NULL) {
vnodeCleanupVnode(pVnode);
return terrno;
}
vDebug("vgId:%d, vnode is opened", pVnode->vgId);
taosHashPut(tsVnode.hash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnode *));
vnodeSetReadyStatus(pVnode);
return TSDB_CODE_SUCCESS;
}
int32_t vnodeCreateVnode(int32_t vgId, SVnodeCfg *pCfg) {
int32_t code = 0;
char path[PATH_MAX + 20] = {0};
snprintf(path, sizeof(path), "%s/vnode%d", tsVnodeDir, vgId);
if (taosMkDir(path) < 0) {
code = TAOS_SYSTEM_ERROR(errno);
vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
return code;
}
snprintf(path, sizeof(path), "%s/vnode%d/cfg", tsVnodeDir, vgId);
if (taosMkDir(path) < 0) {
code = TAOS_SYSTEM_ERROR(errno);
vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
return code;
}
snprintf(path, sizeof(path), "%s/vnode%d/wal", tsVnodeDir, vgId);
if (taosMkDir(path) < 0) {
code = TAOS_SYSTEM_ERROR(errno);
vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
return code;
}
snprintf(path, sizeof(path), "%s/vnode%d/tq", tsVnodeDir, vgId);
if (taosMkDir(path) < 0) {
code = TAOS_SYSTEM_ERROR(errno);
vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
return code;
}
snprintf(path, sizeof(path), "%s/vnode%d/tsdb", tsVnodeDir, vgId);
if (taosMkDir(path) < 0) {
code = TAOS_SYSTEM_ERROR(errno);
vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
return code;
}
snprintf(path, sizeof(path), "%s/vnode%d/meta", tsVnodeDir, vgId);
if (taosMkDir(path) < 0) {
code = TAOS_SYSTEM_ERROR(errno);
vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
return code;
}
code = vnodeWriteCfg(vgId, pCfg);
if (code != 0) {
vError("vgId:%d, failed to save vnode cfg since %s", vgId, tstrerror(code));
return code;
}
return vnodeOpenVnode(vgId);
}
int32_t vnodeAlterVnode(SVnode * pVnode, SVnodeCfg *pCfg) {
int32_t code = 0;
int32_t vgId = pVnode->vgId;
bool walChanged = (memcmp(&pCfg->wal, &pVnode->cfg.wal, sizeof(SWalCfg)) != 0);
bool tsdbChanged = (memcmp(&pCfg->tsdb, &pVnode->cfg.tsdb, sizeof(STsdbCfg)) != 0);
bool metaChanged = (memcmp(&pCfg->meta, &pVnode->cfg.meta, sizeof(SMetaCfg)) != 0);
bool syncChanged = (memcmp(&pCfg->sync, &pVnode->cfg.sync, sizeof(SSyncCluster)) != 0);
if (!walChanged && !tsdbChanged && !metaChanged && !syncChanged) {
vDebug("vgId:%d, nothing changed", vgId);
vnodeRelease(pVnode);
return code;
}
code = vnodeWriteCfg(pVnode->vgId, pCfg);
if (code != 0) {
vError("vgId:%d, failed to write alter msg to file since %s", vgId, tstrerror(code));
vnodeRelease(pVnode);
return code;
}
pVnode->cfg = *pCfg;
if (walChanged) {
code = walAlter(pVnode->pWal, &pVnode->cfg.wal);
if (code != 0) {
vDebug("vgId:%d, failed to alter wal since %s", vgId, tstrerror(code));
vnodeRelease(pVnode);
return code;
}
}
if (tsdbChanged) {
// todo
}
if (metaChanged) {
// todo
}
if (syncChanged) {
// todo
}
vnodeRelease(pVnode);
return code;
}
int32_t vnodeDropVnode(SVnode *pVnode) {
if (pVnode->cfg.dropped) {
vInfo("vgId:%d, already set drop flag, ref:%d", pVnode->vgId, pVnode->refCount);
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
}
pVnode->cfg.dropped = 1;
int32_t code = vnodeWriteCfg(pVnode->vgId, &pVnode->cfg);
if (code == 0) {
vInfo("vgId:%d, set drop flag, ref:%d", pVnode->vgId, pVnode->refCount);
vnodeCleanupVnode(pVnode);
} else {
vError("vgId:%d, failed to set drop flag since %s", pVnode->vgId, tstrerror(code));
pVnode->cfg.dropped = 0;
}
vnodeRelease(pVnode);
return code;
}
int32_t vnodeSyncVnode(SVnode *pVnode) {
return TSDB_CODE_SUCCESS;
}
int32_t vnodeCompactVnode(SVnode *pVnode) {
return TSDB_CODE_SUCCESS;
}
static void *vnodeOpenVnodeFunc(void *param) {
SOpenVnodeThread *pThread = param;
vDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
setThreadName("vnodeOpenVnode");
for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
int32_t vgId = pThread->vnodeList[v];
char stepDesc[TSDB_STEP_DESC_LEN] = {0};
snprintf(stepDesc, TSDB_STEP_DESC_LEN, "vgId:%d, start to restore, %d of %d have been opened", vgId,
tsVnode.openVnodes, tsVnode.totalVnodes);
// (*vnodeInst()->fp.ReportStartup)("open-vnodes", stepDesc);
if (vnodeOpenVnode(vgId) < 0) {
vError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
pThread->failed++;
} else {
vDebug("vgId:%d, is opened by thread:%d", vgId, pThread->threadIndex);
pThread->opened++;
}
atomic_add_fetch_32(&tsVnode.openVnodes, 1);
}
vDebug("thread:%d, total vnodes:%d, opened:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
pThread->failed);
return NULL;
}
static int32_t vnodeGetVnodeListFromDisk(int32_t vnodeList[], int32_t *numOfVnodes) {
#if 0
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) {
vError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
closedir(dir);
return TSDB_CODE_DND_TOO_MANY_VNODES;
} else {
vnodeList[*numOfVnodes - 1] = vnode;
}
}
}
closedir(dir);
#endif
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeOpenVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t status = vnodeGetVnodeListFromDisk(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
vInfo("failed to get vnode list from disk since code:%d", status);
return status;
}
tsVnode.totalVnodes = numOfVnodes;
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];
}
vInfo("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->threadId = taosCreateThread(vnodeOpenVnodeFunc, pThread);
if (pThread->threadId == NULL) {
vError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
}
}
int32_t openVnodes = 0;
int32_t failedVnodes = 0;
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
taosDestoryThread(pThread->threadId);
pThread->threadId = NULL;
openVnodes += pThread->opened;
failedVnodes += pThread->failed;
free(pThread->vnodeList);
}
free(threads);
vInfo("there are total vnodes:%d, opened:%d", numOfVnodes, openVnodes);
if (failedVnodes != 0) {
vError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes);
return -1;
}
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeGetVnodeList(SVnode *vnodeList[], int32_t *numOfVnodes) {
void *pIter = taosHashIterate(tsVnode.hash, NULL);
while (pIter) {
SVnode **pVnode = pIter;
if (*pVnode) {
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES);
continue;
} else {
vnodeList[*numOfVnodes - 1] = (*pVnode);
}
}
pIter = taosHashIterate(tsVnode.hash, pIter);
}
return TSDB_CODE_SUCCESS;
}
static void vnodeCleanupVnodes() {
SVnode* vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t code = vnodeGetVnodeList(vnodeList, &numOfVnodes);
if (code != TSDB_CODE_SUCCESS) {
vInfo("failed to get dnode list since code %d", code);
return;
}
for (int32_t i = 0; i < numOfVnodes; ++i) {
vnodeCleanupVnode(vnodeList[i]);
}
vInfo("total vnodes:%d are all closed", numOfVnodes);
}
static void vnodeIncRef(void *ptNode) {
assert(ptNode != NULL);
SVnode **ppVnode = (SVnode **)ptNode;
assert(ppVnode);
assert(*ppVnode);
SVnode *pVnode = *ppVnode;
atomic_add_fetch_32(&pVnode->refCount, 1);
vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
}
SVnode *vnodeAcquireInAllState(int32_t vgId) {
SVnode *pVnode = NULL;
// taosHashGetClone(tsVnode.hash, &vgId, sizeof(int32_t), vnodeIncRef, (void*)&pVnode);
if (pVnode == NULL) {
vDebug("vgId:%d, can't accquire since not exist", vgId);
terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
return NULL;
}
return pVnode;
}
SVnode *vnodeAcquire(int32_t vgId) {
SVnode *pVnode = vnodeAcquireInAllState(vgId);
if (pVnode == NULL) return NULL;
if (vnodeInStatus(pVnode, TAOS_VN_STATUS_READY)) {
return pVnode;
} else {
vDebug("vgId:%d, can't accquire since not in ready status", vgId);
vnodeRelease(pVnode);
terrno = TSDB_CODE_VND_INVALID_TSDB_STATE;
return NULL;
}
}
void vnodeRelease(SVnode *pVnode) {
if (pVnode == NULL) return;
int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
int32_t vgId = pVnode->vgId;
vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode);
assert(refCount >= 0);
if (refCount <= 0) {
vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode);
vnodeDestroyVnode(pVnode);
int32_t count = taosHashGetSize(tsVnode.hash);
vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count);
}
}
static void vnodeInitMsgFp() {
tsVnode.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessMgmtMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessWriteMsg;
// mq related
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_CONNECT] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_DISCONNECT] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_ACK] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_RESET] = vnodeProcessWriteMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_QUERY] = vnodeProcessReadMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_MQ_CONSUME] = vnodeProcessReadMsg;
// mq related end
tsVnode.msgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessReadMsg;
tsVnode.msgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessReadMsg;
}
int32_t vnodeInitMain() {
vnodeInitMsgFp();
tsVnode.hash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
if (tsVnode.hash == NULL) {
vError("failed to init vnode mgmt");
return -1;
}
vInfo("vnode main is initialized");
return vnodeOpenVnodes();
}
void vnodeCleanupMain() {
vnodeCleanupVnodes();
taosHashCleanup(tsVnode.hash);
tsVnode.hash = NULL;
}
static void vnodeBuildVloadMsg(SVnode *pVnode, SStatusMsg *pStatus) {
int64_t totalStorage = 0;
int64_t compStorage = 0;
int64_t pointsWritten = 0;
if (pStatus->openVnodes >= TSDB_MAX_VNODES) return;
// if (pVnode->tsdb) {
// tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage);
// }
SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
pLoad->vgId = htonl(pVnode->vgId);
pLoad->totalStorage = htobe64(totalStorage);
pLoad->compStorage = htobe64(compStorage);
pLoad->pointsWritten = htobe64(pointsWritten);
pLoad->status = pVnode->status;
pLoad->role = pVnode->role;
}
void vnodeGetStatus(SStatusMsg *pStatus) {
void *pIter = taosHashIterate(tsVnode.hash, NULL);
while (pIter) {
SVnode **pVnode = pIter;
if (*pVnode) {
vnodeBuildVloadMsg(*pVnode, pStatus);
}
pIter = taosHashIterate(tsVnode.hash, pIter);
}
}
void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes) {
for (int32_t i = 0; i < numOfVnodes; ++i) {
pAccess[i].vgId = htonl(pAccess[i].vgId);
SVnode *pVnode = vnodeAcquire(pAccess[i].vgId);
if (pVnode != NULL) {
pVnode->accessState = pAccess[i].accessState;
if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) {
vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState);
}
vnodeRelease(pVnode);
}
}
}
void vnodeProcessMsg(SRpcMsg *pMsg) {
if (tsVnode.msgFp[pMsg->msgType]) {
(*tsVnode.msgFp[pMsg->msgType])(pMsg);
} else {
assert(0);
}
}

248
source/server/vnode/src/vnodeMgmt.c
View File

@ -15,21 +15,185 @@
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "os.h"
#include "vnodeMain.h"
#include "vnodeMgmt.h" #include "vnodeMgmt.h"
#include "vnodeMgmtMsg.h"
typedef struct {
SRpcMsg rpcMsg;
char pCont[];
} SVnMgmtMsg;
static struct { static struct {
SWorkerPool pool; SWorkerPool createPool;
taos_queue pQueue; taos_queue createQueue;
SWorkerPool workerPool;
taos_queue workerQueue;
int32_t (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *); int32_t (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
} tsVmgmt = {0}; } tsVmgmt = {0};
static int32_t vnodeParseCreateVnodeReq(SRpcMsg *rpcMsg, int32_t *vgId, SVnodeCfg *pCfg) {
SCreateVnodeMsg *pCreate = rpcMsg->pCont;
*vgId = htonl(pCreate->vgId);
pCfg->dropped = 0;
tstrncpy(pCfg->db, pCreate->db, sizeof(pCfg->db));
pCfg->tsdb.cacheBlockSize = htonl(pCreate->cacheBlockSize);
pCfg->tsdb.totalBlocks = htonl(pCreate->totalBlocks);
pCfg->tsdb.daysPerFile = htonl(pCreate->daysPerFile);
pCfg->tsdb.daysToKeep1 = htonl(pCreate->daysToKeep1);
pCfg->tsdb.daysToKeep2 = htonl(pCreate->daysToKeep2);
pCfg->tsdb.daysToKeep0 = htonl(pCreate->daysToKeep0);
pCfg->tsdb.minRowsPerFileBlock = htonl(pCreate->minRowsPerFileBlock);
pCfg->tsdb.maxRowsPerFileBlock = htonl(pCreate->maxRowsPerFileBlock);
pCfg->tsdb.precision = pCreate->precision;
pCfg->tsdb.compression = pCreate->compression;
pCfg->tsdb.cacheLastRow = pCreate->cacheLastRow;
pCfg->tsdb.update = pCreate->update;
pCfg->wal.fsyncPeriod = htonl(pCreate->fsyncPeriod);
pCfg->wal.walLevel = pCreate->walLevel;
pCfg->sync.replica = pCreate->replica;
pCfg->sync.quorum = pCreate->quorum;
for (int32_t j = 0; j < pCreate->replica; ++j) {
pCfg->sync.nodes[j].nodePort = htons(pCreate->nodes[j].port);
tstrncpy(pCfg->sync.nodes[j].nodeFqdn, pCreate->nodes[j].fqdn, TSDB_FQDN_LEN);
}
return 0;
}
static int32_t vnodeProcessCreateVnodeReq(SRpcMsg *rpcMsg) {
SVnodeCfg vnodeCfg = {0};
int32_t vgId = 0;
int32_t code = vnodeParseCreateVnodeReq(rpcMsg, &vgId, &vnodeCfg);
if (code != 0) {
vError("failed to parse create vnode msg since %s", tstrerror(code));
}
vDebug("vgId:%d, create vnode req is received", vgId);
SVnode *pVnode = vnodeAcquireInAllState(vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, already exist, return success", vgId);
vnodeRelease(pVnode);
return code;
}
code = vnodeCreateVnode(vgId, &vnodeCfg);
if (code != 0) {
vError("vgId:%d, failed to create vnode since %s", vgId, tstrerror(code));
}
return code;
}
static int32_t vnodeProcessAlterVnodeReq(SRpcMsg *rpcMsg) {
SVnodeCfg vnodeCfg = {0};
int32_t vgId = 0;
int32_t code = vnodeParseCreateVnodeReq(rpcMsg, &vgId, &vnodeCfg);
if (code != 0) {
vError("failed to parse create vnode msg since %s", tstrerror(code));
}
vDebug("vgId:%d, alter vnode req is received", vgId);
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
code = terrno;
vDebug("vgId:%d, failed to alter vnode since %s", vgId, tstrerror(code));
return code;
}
code = vnodeAlterVnode(pVnode, &vnodeCfg);
if (code != 0) {
vError("vgId:%d, failed to alter vnode since %s", vgId, tstrerror(code));
}
vnodeRelease(pVnode);
return code;
}
static SDropVnodeMsg *vnodeParseDropVnodeReq(SRpcMsg *rpcMsg) {
SDropVnodeMsg *pDrop = rpcMsg->pCont;
pDrop->vgId = htonl(pDrop->vgId);
return pDrop;
}
static int32_t vnodeProcessSyncVnodeReq(SRpcMsg *rpcMsg) {
SSyncVnodeMsg *pSync = (SSyncVnodeMsg *)vnodeParseDropVnodeReq(rpcMsg);
int32_t code = 0;
int32_t vgId = pSync->vgId;
vDebug("vgId:%d, sync vnode req is received", vgId);
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
code = terrno;
vDebug("vgId:%d, failed to sync since %s", vgId, tstrerror(code));
return code;
}
code = vnodeSyncVnode(pVnode);
if (code != 0) {
vError("vgId:%d, failed to compact vnode since %s", vgId, tstrerror(code));
}
vnodeRelease(pVnode);
return code;
}
static int32_t vnodeProcessCompactVnodeReq(SRpcMsg *rpcMsg) {
SCompactVnodeMsg *pCompact = (SCompactVnodeMsg *)vnodeParseDropVnodeReq(rpcMsg);
int32_t code = 0;
int32_t vgId = pCompact->vgId;
vDebug("vgId:%d, compact vnode req is received", vgId);
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
code = terrno;
vDebug("vgId:%d, failed to compact since %s", vgId, tstrerror(code));
return code;
}
code = vnodeCompactVnode(pVnode);
if (code != 0) {
vError("vgId:%d, failed to compact vnode since %s", vgId, tstrerror(code));
}
vnodeRelease(pVnode);
return code;
}
static int32_t vnodeProcessDropVnodeReq(SRpcMsg *rpcMsg) {
SDropVnodeMsg *pDrop = vnodeParseDropVnodeReq(rpcMsg);
int32_t code = 0;
int32_t vgId = pDrop->vgId;
vDebug("vgId:%d, drop vnode req is received", vgId);
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
code = terrno;
vDebug("vgId:%d, failed to drop since %s", vgId, tstrerror(code));
return code;
}
code = vnodeDropVnode(pVnode);
if (code != 0) {
vError("vgId:%d, failed to drop vnode since %s", vgId, tstrerror(code));
}
vnodeRelease(pVnode);
return code;
}
static int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg) {
vError("alter stream msg not processed");
return TSDB_CODE_VND_MSG_NOT_PROCESSED;
}
static int32_t vnodeProcessMgmtStart(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype) { static int32_t vnodeProcessMgmtStart(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype) {
SRpcMsg *pMsg = &pMgmt->rpcMsg; SRpcMsg *pMsg = &pMgmt->rpcMsg;
int32_t msgType = pMsg->msgType; int32_t msgType = pMsg->msgType;
@ -43,27 +207,21 @@ static int32_t vnodeProcessMgmtStart(void *unused, SVnMgmtMsg *pMgmt, int32_t qt
} }
} }
static void vnodeSendMgmtEnd(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype, int32_t code) { static void vnodeProcessMgmtEnd(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype, int32_t code) {
SRpcMsg *pMsg = &pMgmt->rpcMsg; SRpcMsg *pMsg = &pMgmt->rpcMsg;
SRpcMsg rsp = {0}; vTrace("msg:%p, is processed, result:%s", pMgmt, tstrerror(code));
rsp.code = code; SRpcMsg rsp = {.code = code, .handle = pMsg->handle};
vTrace("msg:%p, is processed, code:0x%x", pMgmt, rsp.code); rpcSendResponse(&rsp);
if (rsp.code != TSDB_CODE_DND_ACTION_IN_PROGRESS) { taosFreeQitem(pMgmt);
rsp.handle = pMsg->handle;
rsp.pCont = NULL;
rpcSendResponse(&rsp);
}
taosFreeQitem(pMsg);
} }
static void vnodeInitMgmtReqFp() { static void vnodeInitMgmtReqFp() {
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessCreateVnodeMsg; tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessCreateVnodeReq;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessAlterVnodeMsg; tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessAlterVnodeReq;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessSyncVnodeMsg; tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessSyncVnodeReq;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE]= vnodeProcessCompactVnodeMsg; tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessCompactVnodeReq;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessDropVnodeMsg; tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessDropVnodeReq;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessAlterStreamReq; tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessAlterStreamReq;
} }
@ -75,14 +233,18 @@ static int32_t vnodeWriteToMgmtQueue(SRpcMsg *pMsg) {
pMgmt->rpcMsg = *pMsg; pMgmt->rpcMsg = *pMsg;
pMgmt->rpcMsg.pCont = pMgmt->pCont; pMgmt->rpcMsg.pCont = pMgmt->pCont;
memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen); memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen);
taosWriteQitem(tsVmgmt.pQueue, TAOS_QTYPE_RPC, pMgmt);
return TSDB_CODE_SUCCESS; if (pMsg->msgType == TSDB_MSG_TYPE_MD_CREATE_VNODE) {
return taosWriteQitem(tsVmgmt.createQueue, TAOS_QTYPE_RPC, pMgmt);
} else {
return taosWriteQitem(tsVmgmt.workerQueue, TAOS_QTYPE_RPC, pMgmt);
}
} }
void vnodeProcessMgmtMsg(SRpcMsg *pMsg) { void vnodeProcessMgmtMsg(SRpcMsg *pMsg) {
int32_t code = vnodeWriteToMgmtQueue(pMsg); int32_t code = vnodeWriteToMgmtQueue(pMsg);
if (code != TSDB_CODE_SUCCESS) { if (code != TSDB_CODE_SUCCESS) {
vError("msg, ahandle:%p type:%s not processed since %s", pMsg->ahandle, taosMsg[pMsg->msgType], tstrerror(code));
SRpcMsg rsp = {.handle = pMsg->handle, .code = code}; SRpcMsg rsp = {.handle = pMsg->handle, .code = code};
rpcSendResponse(&rsp); rpcSendResponse(&rsp);
} }
@ -93,25 +255,41 @@ void vnodeProcessMgmtMsg(SRpcMsg *pMsg) {
int32_t vnodeInitMgmt() { int32_t vnodeInitMgmt() {
vnodeInitMgmtReqFp(); vnodeInitMgmtReqFp();
SWorkerPool *pPool = &tsVmgmt.pool; SWorkerPool *pPool = &tsVmgmt.createPool;
pPool->name = "vmgmt"; pPool->name = "vnode-mgmt-create";
pPool->startFp = (ProcessStartFp)vnodeProcessMgmtStart; pPool->startFp = (ProcessStartFp)vnodeProcessMgmtStart;
pPool->endFp = (ProcessEndFp)vnodeSendMgmtEnd; pPool->endFp = (ProcessEndFp)vnodeProcessMgmtEnd;
pPool->min = 1; pPool->min = 1;
pPool->max = 1; pPool->max = 1;
if (tWorkerInit(pPool) != 0) { if (tWorkerInit(pPool) != 0) {
return TSDB_CODE_VND_OUT_OF_MEMORY; return TSDB_CODE_VND_OUT_OF_MEMORY;
} }
tsVmgmt.pQueue = tWorkerAllocQueue(pPool, NULL); tsVmgmt.createQueue = tWorkerAllocQueue(pPool, NULL);
vInfo("vmgmt is initialized, max worker %d", pPool->max); pPool = &tsVmgmt.workerPool;
pPool->name = "vnode-mgmt-worker";
pPool->startFp = (ProcessStartFp)vnodeProcessMgmtStart;
pPool->endFp = (ProcessEndFp)vnodeProcessMgmtEnd;
pPool->min = 1;
pPool->max = 1;
if (tWorkerInit(pPool) != 0) {
return TSDB_CODE_VND_OUT_OF_MEMORY;
}
tsVmgmt.workerQueue = tWorkerAllocQueue(pPool, NULL);
vInfo("vmgmt is initialized");
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
void vnodeCleanupMgmt() { void vnodeCleanupMgmt() {
tWorkerFreeQueue(&tsVmgmt.pool, tsVmgmt.pQueue); tWorkerFreeQueue(&tsVmgmt.createPool, tsVmgmt.createQueue);
tWorkerCleanup(&tsVmgmt.pool); tWorkerCleanup(&tsVmgmt.createPool);
tsVmgmt.pQueue = NULL; tsVmgmt.createQueue = NULL;
tWorkerFreeQueue(&tsVmgmt.workerPool, tsVmgmt.workerQueue);
tWorkerCleanup(&tsVmgmt.workerPool);
tsVmgmt.createQueue = NULL;
vInfo("vmgmt is closed"); vInfo("vmgmt is closed");
} }

93
source/server/vnode/src/vnodeMgmtMsg.c
View File

@ -1,93 +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 "vnodeMgmtMsg.h"
static SCreateVnodeMsg* vnodeParseVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = rpcMsg->pCont;
pCreate->cfg.vgId = htonl(pCreate->cfg.vgId);
pCreate->cfg.dbCfgVersion = htonl(pCreate->cfg.dbCfgVersion);
pCreate->cfg.vgCfgVersion = htonl(pCreate->cfg.vgCfgVersion);
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.vgReplica; ++j) {
pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId);
}
return pCreate;
}
int32_t vnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = vnodeParseVnodeMsg(rpcMsg);
SVnode *pVnode = vnodeAcquire(pCreate->cfg.vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId);
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
} else {
vDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId);
return vnodeCreate(pCreate);
}
}
int32_t vnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) {
SAlterVnodeMsg *pAlter = vnodeParseVnodeMsg(rpcMsg);
void *pVnode = vnodeAcquireNotClose(pAlter->cfg.vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId);
int32_t code = vnodeAlter(pVnode, pAlter);
vnodeRelease(pVnode);
return code;
} else {
vInfo("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
}
int32_t vnodeProcessSyncVnodeMsg(SRpcMsg *rpcMsg) {
SSyncVnodeMsg *pSyncVnode = rpcMsg->pCont;
pSyncVnode->vgId = htonl(pSyncVnode->vgId);
return vnodeSync(pSyncVnode->vgId);
}
int32_t vnodeProcessCompactVnodeMsg(SRpcMsg *rpcMsg) {
SCompactVnodeMsg *pCompactVnode = rpcMsg->pCont;
pCompactVnode->vgId = htonl(pCompactVnode->vgId);
return vnodeCompact(pCompactVnode->vgId);
}
int32_t vnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) {
SDropVnodeMsg *pDrop = rpcMsg->pCont;
pDrop->vgId = htonl(pDrop->vgId);
return vnodeDrop(pDrop->vgId);
}
int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg) { return 0; }

14
source/server/vnode/src/vnodeRead.c
View File

@ -14,14 +14,9 @@
*/ */
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "vnodeMain.h"
#include "taosmsg.h"
#include "tglobal.h"
// #include "query.h"
#include "vnodeRead.h" #include "vnodeRead.h"
#include "vnodeReadMsg.h" #include "vnodeReadMsg.h"
#include "vnodeStatus.h"
static struct { static struct {
SWorkerPool query; SWorkerPool query;
@ -50,11 +45,6 @@ static int32_t vnodeWriteToRQueue(SVnode *pVnode, void *pCont, int32_t contLen,
} }
#endif #endif
if (!vnodeInReadyStatus(pVnode)) {
vDebug("vgId:%d, failed to write into vread queue, vnode status is %s", pVnode->vgId, vnodeStatus[pVnode->status]);
return TSDB_CODE_APP_NOT_READY;
}
int32_t size = sizeof(SReadMsg) + contLen; int32_t size = sizeof(SReadMsg) + contLen;
SReadMsg *pRead = taosAllocateQitem(size); SReadMsg *pRead = taosAllocateQitem(size);
if (pRead == NULL) { if (pRead == NULL) {
@ -119,7 +109,7 @@ void vnodeProcessReadMsg(SRpcMsg *pMsg) {
pHead->contLen = htonl(pHead->contLen); pHead->contLen = htonl(pHead->contLen);
assert(pHead->contLen > 0); assert(pHead->contLen > 0);
SVnode *pVnode = vnodeAcquireNotClose(pHead->vgId); SVnode *pVnode = vnodeAcquire(pHead->vgId);
if (pVnode != NULL) { if (pVnode != NULL) {
code = vnodeWriteToRQueue(pVnode, pCont, pHead->contLen, TAOS_QTYPE_RPC, pMsg); code = vnodeWriteToRQueue(pVnode, pCont, pHead->contLen, TAOS_QTYPE_RPC, pMsg);
if (code == TSDB_CODE_SUCCESS) queuedMsgNum++; if (code == TSDB_CODE_SUCCESS) queuedMsgNum++;

7
source/server/vnode/src/vnodeReadMsg.c
View File

@ -14,11 +14,7 @@
*/ */
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "vnodeMain.h"
#include "taosmsg.h"
#include "tglobal.h"
// #include "query.h"
#include "vnodeStatus.h"
#include "vnodeRead.h" #include "vnodeRead.h"
#include "vnodeReadMsg.h" #include "vnodeReadMsg.h"
@ -236,6 +232,7 @@ int32_t vnodeProcessConsumeMsg(SVnode *pVnode, SReadMsg *pRead) {
//fetch or register context //fetch or register context
tqFetchMsg(pHandle, pRead); tqFetchMsg(pHandle, pRead);
//judge mode, tail read or catch up read //judge mode, tail read or catch up read
/*int64_t lastVer = walLastVer(pVnode->wal);*/
//launch new query //launch new query
return 0; return 0;
} }

136
source/server/vnode/src/vnodeStatus.c
View File

@ -1,136 +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 "taosmsg.h"
// #include "query.h"
#include "vnodeRead.h"
#include "vnodeStatus.h"
#include "vnodeWrite.h"
char* vnodeStatus[] = {
"init",
"ready",
"closing",
"updating",
"reset"
};
bool vnodeSetInitStatus(SVnode* pVnode) {
pthread_mutex_lock(&pVnode->statusMutex);
pVnode->status = TAOS_VN_STATUS_INIT;
pthread_mutex_unlock(&pVnode->statusMutex);
return true;
}
bool vnodeSetReadyStatus(SVnode* pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY ||
pVnode->status == TAOS_VN_STATUS_UPDATING) {
pVnode->status = TAOS_VN_STATUS_READY;
set = true;
}
#if 0
qQueryMgmtReOpen(pVnode->qMgmt);
#endif
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
static bool vnodeSetClosingStatusImp(SVnode* pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_READY || pVnode->status == TAOS_VN_STATUS_INIT) {
pVnode->status = TAOS_VN_STATUS_CLOSING;
set = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
bool vnodeSetClosingStatus(SVnode* pVnode) {
if (pVnode->status == TAOS_VN_STATUS_CLOSING)
return true;
while (!vnodeSetClosingStatusImp(pVnode)) {
taosMsleep(1);
}
#if 0
// release local resources only after cutting off outside connections
qQueryMgmtNotifyClosed(pVnode->qMgmt);
#endif
vnodeWaitReadCompleted(pVnode);
vnodeWaitWriteCompleted(pVnode);
return true;
}
bool vnodeSetUpdatingStatus(SVnode* pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_READY) {
pVnode->status = TAOS_VN_STATUS_UPDATING;
set = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
bool vnodeInInitStatus(SVnode* pVnode) {
bool in = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_INIT) {
in = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return in;
}
bool vnodeInReadyStatus(SVnode* pVnode) {
bool in = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_READY) {
in = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return in;
}
bool vnodeInClosingStatus(SVnode* pVnode) {
bool in = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_CLOSING) {
in = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return in;
}

103
source/server/vnode/src/vnodeVersion.c
View File

@ -1,103 +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 "tglobal.h"
#include "vnodeVersion.h"
int32_t vnodeReadVersion(SVnode *pVnode) {
int32_t len = 0;
int32_t maxLen = 100;
char * content = calloc(1, maxLen + 1);
cJSON * root = NULL;
FILE * fp = NULL;
terrno = TSDB_CODE_VND_INVALID_VRESION_FILE;
char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
fp = fopen(file, "r");
if (!fp) {
if (errno != ENOENT) {
vError("vgId:%d, failed to read %s, error:%s", pVnode->vgId, file, strerror(errno));
terrno = TAOS_SYSTEM_ERROR(errno);
} else {
terrno = TSDB_CODE_SUCCESS;
}
goto PARSE_VER_ERROR;
}
len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) {
vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file);
goto PARSE_VER_ERROR;
}
root = cJSON_Parse(content);
if (root == NULL) {
vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file);
goto PARSE_VER_ERROR;
}
cJSON *ver = cJSON_GetObjectItem(root, "version");
if (!ver || ver->type != cJSON_Number) {
vError("vgId:%d, failed to read %s, version not found", pVnode->vgId, file);
goto PARSE_VER_ERROR;
}
#if 0
pVnode->version = (uint64_t)ver->valueint;
terrno = TSDB_CODE_SUCCESS;
vInfo("vgId:%d, read %s successfully, fver:%" PRIu64, pVnode->vgId, file, pVnode->version);
#endif
PARSE_VER_ERROR:
if (content != NULL) free(content);
if (root != NULL) cJSON_Delete(root);
if (fp != NULL) fclose(fp);
return terrno;
}
int32_t vnodeSaveVersion(SVnode *pVnode) {
char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
FILE *fp = fopen(file, "w");
if (!fp) {
vError("vgId:%d, failed to write %s, reason:%s", pVnode->vgId, file, strerror(errno));
return -1;
}
int32_t len = 0;
int32_t maxLen = 100;
char * content = calloc(1, maxLen + 1);
#if 0
len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"version\": %" PRIu64 "\n", pVnode->fversion);
len += snprintf(content + len, maxLen - len, "}\n");
#endif
fwrite(content, 1, len, fp);
taosFsyncFile(fileno(fp));
fclose(fp);
free(content);
terrno = 0;
// vInfo("vgId:%d, successed to write %s, fver:%" PRIu64, pVnode->vgId, file, pVnode->fversion);
return TSDB_CODE_SUCCESS;
}

110
source/server/vnode/src/vnodeWorker.c
View File

@ -1,110 +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 "vnodeWorker.h"
enum { CLEANUP_TASK = 0, DESTROY_TASK = 1, BACKUP_TASK = 2 };
typedef struct {
int32_t vgId;
int32_t code;
int32_t type;
void * rpcHandle;
SVnode *pVnode;
} SVnTask;
static struct {
SWorkerPool pool;
taos_queue pQueue;
} tsVworker = {0};
static void vnodeProcessTaskStart(void *unused, SVnTask *pTask, int32_t qtype) {
pTask->code = 0;
switch (pTask->type) {
case CLEANUP_TASK:
vnodeCleanUp(pTask->pVnode);
break;
case DESTROY_TASK:
vnodeDestroy(pTask->pVnode);
break;
case BACKUP_TASK:
vnodeBackup(pTask->vgId);
break;
default:
break;
}
}
static void vnodeProcessTaskEnd(void *unused, SVnTask *pTask, int32_t qtype, int32_t code) {
if (pTask->rpcHandle != NULL) {
SRpcMsg rpcRsp = {.handle = pTask->rpcHandle, .code = pTask->code};
rpcSendResponse(&rpcRsp);
}
taosFreeQitem(pTask);
}
static int32_t vnodeWriteIntoTaskQueue(SVnode *pVnode, int32_t type, void *rpcHandle) {
SVnTask *pTask = taosAllocateQitem(sizeof(SVnTask));
if (pTask == NULL) return TSDB_CODE_VND_OUT_OF_MEMORY;
pTask->vgId = pVnode->vgId;
pTask->pVnode = pVnode;
pTask->rpcHandle = rpcHandle;
pTask->type = type;
return taosWriteQitem(tsVworker.pQueue, TAOS_QTYPE_RPC, pTask);
}
void vnodeProcessCleanupTask(SVnode *pVnode) {
vnodeWriteIntoTaskQueue(pVnode, CLEANUP_TASK, NULL);
}
void vnodeProcessDestroyTask(SVnode *pVnode) {
vnodeWriteIntoTaskQueue(pVnode, DESTROY_TASK, NULL);
}
void vnodeProcessBackupTask(SVnode *pVnode) {
vnodeWriteIntoTaskQueue(pVnode, BACKUP_TASK, NULL);
}
int32_t vnodeInitWorker() {
SWorkerPool *pPool = &tsVworker.pool;
pPool->name = "vworker";
pPool->startFp = (ProcessStartFp)vnodeProcessTaskStart;
pPool->endFp = (ProcessEndFp)vnodeProcessTaskEnd;
pPool->min = 0;
pPool->max = 1;
if (tWorkerInit(pPool) != 0) {
return TSDB_CODE_VND_OUT_OF_MEMORY;
}
tsVworker.pQueue = tWorkerAllocQueue(pPool, NULL);
vInfo("vworker is initialized, max worker %d", pPool->max);
return TSDB_CODE_SUCCESS;
}
void vnodeCleanupWorker() {
tWorkerFreeQueue(&tsVworker.pool, tsVworker.pQueue);
tWorkerCleanup(&tsVworker.pool);
tsVworker.pQueue = NULL;
vInfo("vworker is closed");
}

12
source/server/vnode/src/vnodeWrite.c
View File

@ -15,12 +15,9 @@
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "os.h"
#include "tglobal.h"
#include "tqueue.h" #include "tqueue.h"
#include "tworker.h" #include "tworker.h"
#include "taosmsg.h" #include "vnodeMain.h"
#include "vnodeStatus.h"
#include "vnodeWrite.h" #include "vnodeWrite.h"
#include "vnodeWriteMsg.h" #include "vnodeWriteMsg.h"
@ -68,11 +65,6 @@ static int32_t vnodeWriteToWQueue(SVnode *pVnode, SWalHead *pHead, int32_t qtype
return TSDB_CODE_WAL_SIZE_LIMIT; return TSDB_CODE_WAL_SIZE_LIMIT;
} }
if (!vnodeInReadyStatus(pVnode)) {
vError("vgId:%d, failed to write into vwqueue, vstatus is %s", pVnode->vgId, vnodeStatus[pVnode->status]);
return TSDB_CODE_APP_NOT_READY;
}
if (tsVwrite.queuedBytes > tsMaxVnodeQueuedBytes) { if (tsVwrite.queuedBytes > tsMaxVnodeQueuedBytes) {
vDebug("vgId:%d, too many bytes:%" PRId64 " in vwqueue, flow control", pVnode->vgId, tsVwrite.queuedBytes); vDebug("vgId:%d, too many bytes:%" PRId64 " in vwqueue, flow control", pVnode->vgId, tsVwrite.queuedBytes);
return TSDB_CODE_VND_IS_FLOWCTRL; return TSDB_CODE_VND_IS_FLOWCTRL;
@ -122,7 +114,7 @@ void vnodeProcessWriteMsg(SRpcMsg *pRpcMsg) {
pMsg->vgId = htonl(pMsg->vgId); pMsg->vgId = htonl(pMsg->vgId);
pMsg->contLen = htonl(pMsg->contLen); pMsg->contLen = htonl(pMsg->contLen);
SVnode *pVnode = vnodeAcquireNotClose(pMsg->vgId); SVnode *pVnode = vnodeAcquire(pMsg->vgId);
if (pVnode == NULL) { if (pVnode == NULL) {
code = TSDB_CODE_VND_INVALID_VGROUP_ID; code = TSDB_CODE_VND_INVALID_VGROUP_ID;
} else { } else {