Merge remote-tracking branch 'origin/3.0' into feature/3.0_liaohj

This commit is contained in:
Haojun Liao 2022-03-15 18:10:49 +08:00
commit 5b36dddec9
79 changed files with 3478 additions and 1497 deletions

View File

@ -33,7 +33,7 @@ typedef enum {
TSDB_SUPER_TABLE = 1, // super table
TSDB_CHILD_TABLE = 2, // table created from super table
TSDB_NORMAL_TABLE = 3, // ordinary table
TSDB_STREAM_TABLE = 4, // table created from stream computing
TSDB_STREAM_TABLE = 4, // table created by stream processing
TSDB_TEMP_TABLE = 5, // temp table created by nest query
TSDB_TABLE_MAX = 6
} ETableType;
@ -50,7 +50,12 @@ typedef enum {
TSDB_CHECK_ITEM_MAX
} ECheckItemType;
typedef enum { TD_ROW_DISCARD_UPDATE = 0, TD_ROW_OVERWRITE_UPDATE = 1, TD_ROW_PARTIAL_UPDATE = 2 } TDUpdateConfig;
typedef enum {
TD_ROW_DISCARD_UPDATE = 0,
TD_ROW_OVERWRITE_UPDATE = 1,
TD_ROW_PARTIAL_UPDATE = 2,
} TDUpdateConfig;
typedef enum {
TSDB_STATIS_OK = 0, // statis part exist and load successfully
TSDB_STATIS_NONE = 1, // statis part not exist
@ -61,6 +66,12 @@ typedef enum {
TSDB_SMA_STAT_EXPIRED = 1, // not ready or expired
} ETsdbSmaStat;
typedef enum {
TSDB_SMA_TYPE_BLOCK = 0, // Block-wise SMA
TSDB_SMA_TYPE_TIME_RANGE = 1, // Time-range-wise SMA
TSDB_SMA_TYPE_ROLLUP = 2, // Rollup SMA
} ETsdbSmaType;
extern char *qtypeStr[];
#define TSDB_PORT_HTTP 11

View File

@ -140,6 +140,23 @@ static FORCE_INLINE void* tDecodeDataBlock(const void* buf, SSDataBlock* pBlock)
return (void*)buf;
}
static FORCE_INLINE void tDeleteSSDataBlock(SSDataBlock* pBlock) {
if (pBlock == NULL) {
return;
}
// int32_t numOfOutput = pBlock->info.numOfCols;
int32_t sz = taosArrayGetSize(pBlock->pDataBlock);
for (int32_t i = 0; i < sz; ++i) {
SColumnInfoData* pColInfoData = (SColumnInfoData*)taosArrayGet(pBlock->pDataBlock, i);
tfree(pColInfoData->pData);
}
taosArrayDestroy(pBlock->pDataBlock);
tfree(pBlock->pBlockAgg);
// tfree(pBlock);
}
static FORCE_INLINE int32_t tEncodeSMqPollRsp(void** buf, const SMqPollRsp* pRsp) {
int32_t tlen = 0;
int32_t sz = 0;
@ -182,23 +199,6 @@ static FORCE_INLINE void* tDecodeSMqPollRsp(void* buf, SMqPollRsp* pRsp) {
return buf;
}
static FORCE_INLINE void tDeleteSSDataBlock(SSDataBlock* pBlock) {
if (pBlock == NULL) {
return;
}
// int32_t numOfOutput = pBlock->info.numOfCols;
int32_t sz = taosArrayGetSize(pBlock->pDataBlock);
for (int32_t i = 0; i < sz; ++i) {
SColumnInfoData* pColInfoData = (SColumnInfoData*)taosArrayGet(pBlock->pDataBlock, i);
tfree(pColInfoData->pData);
}
taosArrayDestroy(pBlock->pDataBlock);
tfree(pBlock->pBlockAgg);
// tfree(pBlock);
}
static FORCE_INLINE void tDeleteSMqConsumeRsp(SMqPollRsp* pRsp) {
if (pRsp->schemas) {
if (pRsp->schemas->nCols) {
@ -208,10 +208,6 @@ static FORCE_INLINE void tDeleteSMqConsumeRsp(SMqPollRsp* pRsp) {
}
taosArrayDestroyEx(pRsp->pBlockData, (void (*)(void*))tDeleteSSDataBlock);
pRsp->pBlockData = NULL;
// for (int32_t i = 0; i < taosArrayGetSize(pRsp->pBlockData); i++) {
// SSDataBlock* pDataBlock = (SSDataBlock*)taosArrayGet(pRsp->pBlockData, i);
// tDeleteSSDataBlock(pDataBlock);
//}
}
//======================================================================================================================

View File

@ -190,7 +190,10 @@ typedef struct SEp {
typedef struct {
int32_t contLen;
int32_t vgId;
union {
int32_t vgId;
int32_t streamTaskId;
};
} SMsgHead;
// Submit message for one table
@ -1139,6 +1142,17 @@ int32_t tSerializeSCMCreateStreamReq(void* buf, int32_t bufLen, const SCMCreateS
int32_t tDeserializeSCMCreateStreamReq(void* buf, int32_t bufLen, SCMCreateStreamReq* pReq);
void tFreeSCMCreateStreamReq(SCMCreateStreamReq* pReq);
typedef struct {
char name[TSDB_TOPIC_FNAME_LEN];
int64_t streamId;
char* sql;
char* executorMsg;
} SMVCreateStreamReq, SMSCreateStreamReq;
typedef struct {
int64_t streamId;
} SMVCreateStreamRsp, SMSCreateStreamRsp;
typedef struct {
char name[TSDB_TOPIC_FNAME_LEN];
int8_t igExists;

View File

@ -23,6 +23,9 @@
extern "C" {
#endif
#define SND_UNIQUE_THREAD_NUM 2
#define SND_SHARED_THREAD_NUM 2
/* ------------------------ TYPES EXPOSED ------------------------ */
typedef struct SDnode SDnode;
typedef struct SSnode SSnode;
@ -80,6 +83,10 @@ int32_t sndGetLoad(SSnode *pSnode, SSnodeLoad *pLoad);
*/
int32_t sndProcessMsg(SSnode *pSnode, SRpcMsg *pMsg, SRpcMsg **pRsp);
int32_t sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg);
int32_t sndProcessSMsg(SSnode *pSnode, SRpcMsg *pMsg);
/**
* @brief Drop a snode.
*

View File

@ -71,6 +71,10 @@ typedef struct SRpcInit {
// call back to keep conn or not
bool (*pfp)(void *parent, tmsg_t msgType);
// to support Send messages multiple times on a link
//
void* (*mfp)(void *parent, tmsg_t msgType);
void *parent;
} SRpcInit;
@ -89,6 +93,9 @@ void rpcSendRecv(void *shandle, SEpSet *pEpSet, SRpcMsg *pReq, SRpcMsg *pRsp)
int rpcReportProgress(void *pConn, char *pCont, int contLen);
void rpcCancelRequest(int64_t rid);
// just release client conn to rpc instance, no close sock
void rpcReleaseHandle(void *handle);
void rpcRefHandle(void *handle, int8_t type);
void rpcUnrefHandle(void *handle, int8_t type);

View File

@ -17,12 +17,16 @@
#define _TD_OS_LOCALE_H_
#include "os.h"
#include "osString.h"
#ifdef __cplusplus
extern "C" {
#endif
// If the error is in a third-party library, place this header file under the third-party library header file.
#ifndef ALLOW_FORBID_FUNC
#define setlocale SETLOCALE_FUNC_TAOS_FORBID
#endif
char *taosCharsetReplace(char *charsetstr);
void taosGetSystemLocale(char *outLocale, char *outCharset);
void taosSetSystemLocale(const char *inLocale, const char *inCharSet);

View File

@ -353,6 +353,7 @@ int32_t* taosGetErrno();
#define TSDB_CODE_TDB_MESSED_MSG TAOS_DEF_ERROR_CODE(0, 0x0614)
#define TSDB_CODE_TDB_IVLD_TAG_VAL TAOS_DEF_ERROR_CODE(0, 0x0615)
#define TSDB_CODE_TDB_NO_CACHE_LAST_ROW TAOS_DEF_ERROR_CODE(0, 0x0616)
#define TSDB_CODE_TDB_NO_SMA_INDEX_IN_META TAOS_DEF_ERROR_CODE(0, 0x0617)
// query
#define TSDB_CODE_QRY_INVALID_QHANDLE TAOS_DEF_ERROR_CODE(0, 0x0700)

39
include/util/tuuid.h Normal file
View File

@ -0,0 +1,39 @@
/*
* 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/>.
*/
#include "os.h"
#include "taoserror.h"
#include "thash.h"
/**
* Generate an non-negative signed 32bit id
*+------------+-----+-----------+---------------+
*| uid|localIp| PId | timestamp | serial number |
*+------------+-----+-----------+---------------+
*| 6bit |6bit | 12bit | 8bit |
*+------------+-----+-----------+---------------+
* @return
*/
int32_t tGenIdPI32(void);
/**
* Generate an non-negative signed 64bit id
*+------------+-----+-----------+---------------+
*| uid|localIp| PId | timestamp | serial number |
*+------------+-----+-----------+---------------+
*| 12bit |12bit|24bit |16bit |
*+------------+-----+-----------+---------------+
* @return
*/
int64_t tGenIdPI64(void);

View File

@ -2666,7 +2666,6 @@ int32_t tSerializeSCMCreateStreamReq(void *buf, int32_t bufLen, const SCMCreateS
if (tEncodeCStr(&encoder, pReq->sql) < 0) return -1;
if (tEncodeCStr(&encoder, pReq->physicalPlan) < 0) return -1;
if (tEncodeCStr(&encoder, pReq->logicalPlan) < 0) return -1;
tEndEncode(&encoder);
int32_t tlen = encoder.pos;

View File

@ -90,9 +90,11 @@ typedef struct {
int32_t refCount;
int8_t deployed;
int8_t dropped;
int8_t uniqueWorkerInUse;
SSnode *pSnode;
SRWLatch latch;
SDnodeWorker writeWorker;
SArray *uniqueWorkers; // SArray<SDnodeWorker*>
SDnodeWorker sharedWorker;
} SSnodeMgmt;
typedef struct {
@ -153,4 +155,4 @@ int32_t dndGetMonitorDiskInfo(SDnode *pDnode, SMonDiskInfo *pInfo);
}
#endif
#endif /*_TD_DND_ENV_H_*/
#endif /*_TD_DND_ENV_H_*/

View File

@ -70,4 +70,4 @@ void dndGetStartup(SDnode *pDnode, SStartupReq *pStartup);
}
#endif
#endif /*_TD_DND_INT_H_*/
#endif /*_TD_DND_INT_H_*/

View File

@ -19,7 +19,20 @@
#include "dndTransport.h"
#include "dndWorker.h"
static void dndProcessSnodeQueue(SDnode *pDnode, SRpcMsg *pMsg);
typedef struct {
int32_t vgId;
int32_t refCount;
int32_t snVersion;
int8_t dropped;
char *path;
SSnode *pImpl;
STaosQueue *pSharedQ;
STaosQueue *pUniqueQ;
} SSnodeObj;
static void dndProcessSnodeSharedQueue(SDnode *pDnode, SRpcMsg *pMsg);
static void dndProcessSnodeUniqueQueue(SDnode *pDnode, STaosQall *qall, int32_t numOfMsgs);
static SSnode *dndAcquireSnode(SDnode *pDnode) {
SSnodeMgmt *pMgmt = &pDnode->smgmt;
@ -152,8 +165,21 @@ static int32_t dndWriteSnodeFile(SDnode *pDnode) {
static int32_t dndStartSnodeWorker(SDnode *pDnode) {
SSnodeMgmt *pMgmt = &pDnode->smgmt;
if (dndInitWorker(pDnode, &pMgmt->writeWorker, DND_WORKER_SINGLE, "snode-write", 0, 1, dndProcessSnodeQueue) != 0) {
dError("failed to start snode write worker since %s", terrstr());
pMgmt->uniqueWorkers = taosArrayInit(0, sizeof(void *));
for (int32_t i = 0; i < SND_UNIQUE_THREAD_NUM; i++) {
SDnodeWorker *pUniqueWorker = malloc(sizeof(SDnodeWorker));
if (pUniqueWorker == NULL) {
return -1;
}
if (dndInitWorker(pDnode, pUniqueWorker, DND_WORKER_MULTI, "snode-unique", 1, 1, dndProcessSnodeSharedQueue) != 0) {
dError("failed to start snode unique worker since %s", terrstr());
return -1;
}
taosArrayPush(pMgmt->uniqueWorkers, &pUniqueWorker);
}
if (dndInitWorker(pDnode, &pMgmt->sharedWorker, DND_WORKER_SINGLE, "snode-shared", SND_SHARED_THREAD_NUM,
SND_SHARED_THREAD_NUM, dndProcessSnodeSharedQueue)) {
dError("failed to start snode shared worker since %s", terrstr());
return -1;
}
@ -169,9 +195,13 @@ static void dndStopSnodeWorker(SDnode *pDnode) {
while (pMgmt->refCount > 0) {
taosMsleep(10);
}
}
dndCleanupWorker(&pMgmt->writeWorker);
for (int32_t i = 0; i < taosArrayGetSize(pMgmt->uniqueWorkers); i++) {
SDnodeWorker *worker = taosArrayGetP(pMgmt->uniqueWorkers, i);
dndCleanupWorker(worker);
}
taosArrayDestroy(pMgmt->uniqueWorkers);
}
static void dndBuildSnodeOption(SDnode *pDnode, SSnodeOpt *pOption) {
@ -292,17 +322,36 @@ int32_t dndProcessDropSnodeReq(SDnode *pDnode, SRpcMsg *pReq) {
}
}
static void dndProcessSnodeQueue(SDnode *pDnode, SRpcMsg *pMsg) {
static void dndProcessSnodeUniqueQueue(SDnode *pDnode, STaosQall *qall, int32_t numOfMsgs) {
SSnodeMgmt *pMgmt = &pDnode->smgmt;
SRpcMsg *pRsp = NULL;
int32_t code = TSDB_CODE_DND_SNODE_NOT_DEPLOYED;
SSnode *pSnode = dndAcquireSnode(pDnode);
if (pSnode != NULL) {
code = sndProcessMsg(pSnode, pMsg, &pRsp);
for (int32_t i = 0; i < numOfMsgs; i++) {
SRpcMsg *pMsg = NULL;
taosGetQitem(qall, (void **)&pMsg);
sndProcessUMsg(pSnode, pMsg);
rpcFreeCont(pMsg->pCont);
taosFreeQitem(pMsg);
}
}
dndReleaseSnode(pDnode, pSnode);
}
static void dndProcessSnodeSharedQueue(SDnode *pDnode, SRpcMsg *pMsg) {
SSnodeMgmt *pMgmt = &pDnode->smgmt;
int32_t code = TSDB_CODE_DND_SNODE_NOT_DEPLOYED;
SSnode *pSnode = dndAcquireSnode(pDnode);
if (pSnode != NULL) {
code = sndProcessSMsg(pSnode, pMsg);
}
dndReleaseSnode(pDnode, pSnode);
#if 0
if (pMsg->msgType & 1u) {
if (pRsp != NULL) {
pRsp->ahandle = pMsg->ahandle;
@ -314,11 +363,58 @@ static void dndProcessSnodeQueue(SDnode *pDnode, SRpcMsg *pMsg) {
rpcSendResponse(&rpcRsp);
}
}
#endif
rpcFreeCont(pMsg->pCont);
taosFreeQitem(pMsg);
}
static FORCE_INLINE int32_t dndGetSWIdFromMsg(SRpcMsg *pMsg) {
SMsgHead *pHead = pMsg->pCont;
pHead->streamTaskId = htonl(pHead->streamTaskId);
return pHead->streamTaskId % SND_UNIQUE_THREAD_NUM;
}
static void dndWriteSnodeMsgToWorkerByMsg(SDnode *pDnode, SRpcMsg *pMsg) {
int32_t code = TSDB_CODE_DND_SNODE_NOT_DEPLOYED;
SSnode *pSnode = dndAcquireSnode(pDnode);
if (pSnode != NULL) {
int32_t index = dndGetSWIdFromMsg(pMsg);
SDnodeWorker *pWorker = taosArrayGetP(pDnode->smgmt.uniqueWorkers, index);
code = dndWriteMsgToWorker(pWorker, pMsg, sizeof(SRpcMsg));
}
dndReleaseSnode(pDnode, pSnode);
if (code != 0) {
if (pMsg->msgType & 1u) {
SRpcMsg rsp = {.handle = pMsg->handle, .ahandle = pMsg->ahandle, .code = code};
rpcSendResponse(&rsp);
}
rpcFreeCont(pMsg->pCont);
}
}
static void dndWriteSnodeMsgToMgmtWorker(SDnode *pDnode, SRpcMsg *pMsg) {
int32_t code = TSDB_CODE_DND_SNODE_NOT_DEPLOYED;
SSnode *pSnode = dndAcquireSnode(pDnode);
if (pSnode != NULL) {
SDnodeWorker *pWorker = taosArrayGet(pDnode->smgmt.uniqueWorkers, 0);
code = dndWriteMsgToWorker(pWorker, pMsg, sizeof(SRpcMsg));
}
dndReleaseSnode(pDnode, pSnode);
if (code != 0) {
if (pMsg->msgType & 1u) {
SRpcMsg rsp = {.handle = pMsg->handle, .ahandle = pMsg->ahandle, .code = code};
rpcSendResponse(&rsp);
}
rpcFreeCont(pMsg->pCont);
}
}
static void dndWriteSnodeMsgToWorker(SDnode *pDnode, SDnodeWorker *pWorker, SRpcMsg *pMsg) {
int32_t code = TSDB_CODE_DND_SNODE_NOT_DEPLOYED;
@ -337,8 +433,16 @@ static void dndWriteSnodeMsgToWorker(SDnode *pDnode, SDnodeWorker *pWorker, SRpc
}
}
void dndProcessSnodeWriteMsg(SDnode *pDnode, SRpcMsg *pMsg, SEpSet *pEpSet) {
dndWriteSnodeMsgToWorker(pDnode, &pDnode->smgmt.writeWorker, pMsg);
void dndProcessSnodeMgmtMsg(SDnode *pDnode, SRpcMsg *pMsg, SEpSet *pEpSet) {
dndWriteSnodeMsgToMgmtWorker(pDnode, pMsg);
}
void dndProcessSnodeUniqueMsg(SDnode *pDnode, SRpcMsg *pMsg, SEpSet *pEpSet) {
dndWriteSnodeMsgToWorkerByMsg(pDnode, pMsg);
}
void dndProcessSnodeSharedMsg(SDnode *pDnode, SRpcMsg *pMsg, SEpSet *pEpSet) {
dndWriteSnodeMsgToWorker(pDnode, &pDnode->smgmt.sharedWorker, pMsg);
}
int32_t dndInitSnode(SDnode *pDnode) {

View File

@ -109,4 +109,4 @@ int32_t dndWriteMsgToWorker(SDnodeWorker *pWorker, void *pCont, int32_t contLen)
}
return 0;
}
}

View File

@ -30,6 +30,7 @@
#include "mndShow.h"
#include "mndSnode.h"
#include "mndStb.h"
#include "mndStream.h"
#include "mndSubscribe.h"
#include "mndSync.h"
#include "mndTelem.h"
@ -220,6 +221,7 @@ static int32_t mndInitSteps(SMnode *pMnode) {
if (mndAllocStep(pMnode, "mnode-user", mndInitUser, mndCleanupUser) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-auth", mndInitAuth, mndCleanupAuth) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-acct", mndInitAcct, mndCleanupAcct) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-stream", mndInitStream, mndCleanupStream) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-topic", mndInitTopic, mndCleanupTopic) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-consumer", mndInitConsumer, mndCleanupConsumer) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-subscribe", mndInitSubscribe, mndCleanupSubscribe) != 0) return -1;

View File

@ -30,41 +30,42 @@ extern "C" {
#endif
enum {
STREAM_STATUS__READY = 1,
STREAM_STATUS__RUNNING = 1,
STREAM_STATUS__STOPPED,
STREAM_STATUS__CREATING,
STREAM_STATUS__STOPING,
STREAM_STATUS__RESUMING,
STREAM_STATUS__RESTORING,
STREAM_STATUS__DELETING,
};
enum {
STREAM_RUNNER__RUNNING = 1,
STREAM_RUNNER__STOP,
STREAM_TASK_STATUS__RUNNING = 1,
STREAM_TASK_STATUS__STOP,
};
typedef struct {
SHashObj* pHash; // taskId -> streamTask
} SStreamMeta;
typedef struct SSnode {
SSnodeOpt cfg;
SStreamMeta* pMeta;
SSnodeOpt cfg;
} SSnode;
typedef struct {
int64_t streamId;
int32_t taskId;
int32_t IdxInLevel;
int32_t level;
} SStreamInfo;
} SStreamTaskInfo;
typedef struct {
SStreamInfo meta;
int8_t status;
void* executor;
STaosQueue* queue;
void* stateStore;
SStreamTaskInfo meta;
int8_t status;
void* executor;
void* stateStore;
// storage handle
} SStreamRunner;
typedef struct {
SHashObj* pHash;
} SStreamMeta;
} SStreamTask;
int32_t sndCreateStream();
int32_t sndDropStream();

View File

@ -14,6 +14,7 @@
*/
#include "sndInt.h"
#include "tuuid.h"
SSnode *sndOpen(const char *path, const SSnodeOpt *pOption) {
SSnode *pSnode = calloc(1, sizeof(SSnode));
@ -31,3 +32,25 @@ int32_t sndProcessMsg(SSnode *pSnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
}
void sndDestroy(const char *path) {}
static int32_t sndDeployTask(SSnode *pSnode, SRpcMsg *pMsg) {
SStreamTask *task = malloc(sizeof(SStreamTask));
if (task == NULL) {
return -1;
}
task->meta.taskId = tGenIdPI32();
taosHashPut(pSnode->pMeta->pHash, &task->meta.taskId, sizeof(int32_t), &task, sizeof(void *));
return 0;
}
int32_t sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg) {
// stream deployment
// stream stop/resume
// operator exec
return 0;
}
int32_t sndProcessSMsg(SSnode *pSnode, SRpcMsg *pMsg) {
// operator exec
return 0;
}

View File

@ -95,6 +95,7 @@ int tsdbCommit(STsdb *pTsdb);
* @return int32_t
*/
int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg);
int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg);
/**
* @brief Insert RSma(Time-range-wise Rollup SMA) data.
@ -105,6 +106,12 @@ int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg);
*/
int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg);
// TODO: This is the basic params, and should wrap the params to a queryHandle.
int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySkey,
int32_t nMaxResult);
// STsdbCfg
int tsdbOptionsInit(STsdbCfg *);
void tsdbOptionsClear(STsdbCfg *);

View File

@ -0,0 +1,44 @@
/*
* 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_TSDB_DB_DEF_H_
#define _TD_TSDB_DB_DEF_H_
#include "db.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef struct SDBFile SDBFile;
typedef DB_ENV* TDBEnv;
struct SDBFile {
DB* pDB;
char* path;
};
int32_t tsdbOpenDBF(TDBEnv pEnv, SDBFile* pDBF);
void tsdbCloseDBF(SDBFile* pDBF);
int32_t tsdbOpenBDBEnv(DB_ENV** ppEnv, const char* path);
void tsdbCloseBDBEnv(DB_ENV* pEnv);
int32_t tsdbSaveSmaToDB(SDBFile* pDBF, void* key, uint32_t keySize, void* data, uint32_t dataSize);
void* tsdbGetSmaDataByKey(SDBFile* pDBF, void* key, uint32_t keySize, uint32_t* valueSize);
#ifdef __cplusplus
}
#endif
#endif /*_TD_TSDB_DB_DEF_H_*/

View File

@ -27,6 +27,7 @@
#include "ttime.h"
#include "tsdb.h"
#include "tsdbDBDef.h"
#include "tsdbCommit.h"
#include "tsdbFS.h"
#include "tsdbFile.h"
@ -37,12 +38,15 @@
#include "tsdbReadImpl.h"
#include "tsdbSma.h"
#ifdef __cplusplus
extern "C" {
#endif
struct STsdb {
int32_t vgId;
bool repoLocked;
pthread_mutex_t mutex;
char * path;
STsdbCfg config;
STsdbMemTable * mem;
@ -52,12 +56,17 @@ struct STsdb {
STsdbFS * fs;
SMeta * pMeta;
STfs * pTfs;
SSmaStat * pSmaStat;
SSmaEnv * pTSmaEnv;
SSmaEnv * pRSmaEnv;
};
#define REPO_ID(r) ((r)->vgId)
#define REPO_CFG(r) (&(r)->config)
#define REPO_FS(r) (r)->fs
#define REPO_ID(r) ((r)->vgId)
#define REPO_CFG(r) (&(r)->config)
#define REPO_FS(r) (r)->fs
#define IS_REPO_LOCKED(r) (r)->repoLocked
int tsdbLockRepo(STsdb *pTsdb);
int tsdbUnlockRepo(STsdb *pTsdb);
static FORCE_INLINE STSchema *tsdbGetTableSchemaImpl(STable *pTable, bool lock, bool copy, int32_t version) {
return pTable->pSchema;

View File

@ -329,21 +329,23 @@ static FORCE_INLINE int tsdbCopyDFile(SDFile* pSrc, SDFile* pDest) {
// =============== SDFileSet
typedef struct {
int fid;
int8_t state; // -128~127
uint8_t ver; // 0~255, DFileSet version
int8_t state; // -128~127
uint8_t ver; // 0~255, DFileSet version
uint16_t reserve;
SDFile files[TSDB_FILE_MAX];
} SDFileSet;
typedef struct {
int fid;
int8_t state;
uint8_t ver;
int fid;
int8_t state;
uint8_t ver;
uint16_t reserve;
#if 0
SDFInfo info;
#endif
STfsFile f;
TdFilePtr pFile;
} SSFile; // files split by days with fid
#define TSDB_LATEST_FSET_VER 0

View File

@ -17,27 +17,29 @@
#define _TD_TSDB_SMA_H_
typedef struct SSmaStat SSmaStat;
typedef struct SSmaEnv SSmaEnv;
// insert/update interface
int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg);
int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg);
struct SSmaEnv {
pthread_rwlock_t lock;
TDBEnv dbEnv;
char * path;
SSmaStat * pStat;
};
#define SMA_ENV_LOCK(env) ((env)->lock)
#define SMA_ENV_ENV(env) ((env)->dbEnv)
#define SMA_ENV_PATH(env) ((env)->path)
#define SMA_ENV_STAT(env) ((env)->pStat)
#define SMA_ENV_STAT_ITEMS(env) ((env)->pStat->smaStatItems)
// query interface
// TODO: This is the basic params, and should wrap the params to a queryHandle.
int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, STimeWindow *queryWin, int32_t nMaxResult);
// management interface
int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, char *msg);
int32_t tsdbDestroySmaState(SSmaStat *pSmaStat);
void tsdbDestroySmaEnv(SSmaEnv *pSmaEnv);
void *tsdbFreeSmaEnv(SSmaEnv *pSmaEnv);
#if 0
int32_t tsdbGetTSmaStatus(STsdb *pTsdb, STSma *param, void *result);
int32_t tsdbRemoveTSmaData(STsdb *pTsdb, STSma *param, STimeWindow *pWin);
#endif
// internal func
static FORCE_INLINE int32_t tsdbEncodeTSmaKey(tb_uid_t tableUid, col_id_t colId, TSKEY tsKey, void **pData) {
int32_t len = 0;
len += taosEncodeFixedI64(pData, tableUid);
@ -46,4 +48,31 @@ static FORCE_INLINE int32_t tsdbEncodeTSmaKey(tb_uid_t tableUid, col_id_t colId,
return len;
}
static FORCE_INLINE int tsdbRLockSma(SSmaEnv *pEnv) {
int code = pthread_rwlock_rdlock(&(pEnv->lock));
if (code != 0) {
terrno = TAOS_SYSTEM_ERROR(code);
return -1;
}
return 0;
}
static FORCE_INLINE int tsdbWLockSma(SSmaEnv *pEnv) {
int code = pthread_rwlock_wrlock(&(pEnv->lock));
if (code != 0) {
terrno = TAOS_SYSTEM_ERROR(code);
return -1;
}
return 0;
}
static FORCE_INLINE int tsdbUnLockSma(SSmaEnv *pEnv) {
int code = pthread_rwlock_unlock(&(pEnv->lock));
if (code != 0) {
terrno = TAOS_SYSTEM_ERROR(code);
return -1;
}
return 0;
}
#endif /* _TD_TSDB_SMA_H_ */

View File

@ -231,30 +231,31 @@ int metaSaveSmaToDB(SMeta *pMeta, STSma *pSmaCfg) {
void *pBuf = NULL, *qBuf = NULL;
DBT key1 = {0}, value1 = {0};
{
// save sma info
int32_t len = tEncodeTSma(NULL, pSmaCfg);
pBuf = calloc(len, 1);
if (pBuf == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
key1.data = (void *)&pSmaCfg->indexUid;
key1.size = sizeof(pSmaCfg->indexUid);
qBuf = pBuf;
tEncodeTSma(&qBuf, pSmaCfg);
value1.data = pBuf;
value1.size = POINTER_DISTANCE(qBuf, pBuf);
value1.app_data = pSmaCfg;
// save sma info
int32_t len = tEncodeTSma(NULL, pSmaCfg);
pBuf = calloc(len, 1);
if (pBuf == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
key1.data = (void *)&pSmaCfg->indexUid;
key1.size = sizeof(pSmaCfg->indexUid);
qBuf = pBuf;
tEncodeTSma(&qBuf, pSmaCfg);
value1.data = pBuf;
value1.size = POINTER_DISTANCE(qBuf, pBuf);
value1.app_data = pSmaCfg;
metaDBWLock(pMeta->pDB);
pMeta->pDB->pSmaDB->put(pMeta->pDB->pSmaDB, NULL, &key1, &value1, 0);
metaDBULock(pMeta->pDB);
// release
tfree(pBuf);
return 0;
}

View File

@ -83,8 +83,8 @@ bool tqNextDataBlock(STqReadHandle* pHandle) {
}
int tqRetrieveDataBlockInfo(STqReadHandle* pHandle, SDataBlockInfo* pBlockInfo) {
/*int32_t sversion = pHandle->pBlock->sversion;*/
/*SSchemaWrapper* pSchema = metaGetTableSchema(pHandle->pMeta, pHandle->pBlock->uid, sversion, false);*/
// currently only rows are used
pBlockInfo->numOfCols = taosArrayGetSize(pHandle->pColIdList);
pBlockInfo->rows = pHandle->pBlock->numOfRows;
pBlockInfo->uid = pHandle->pBlock->uid;

View File

@ -12,3 +12,162 @@
* 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 ALLOW_FORBID_FUNC
#include "db.h"
#include "taoserror.h"
#include "tcoding.h"
#include "thash.h"
#include "tsdbDBDef.h"
#include "tsdbLog.h"
#define IMPL_WITH_LOCK 1
static int tsdbOpenBDBDb(DB **ppDB, DB_ENV *pEnv, const char *pFName, bool isDup);
static void tsdbCloseBDBDb(DB *pDB);
#define BDB_PERR(info, code) fprintf(stderr, "%s:%d " info " reason: %s\n", __FILE__, __LINE__, db_strerror(code))
int32_t tsdbOpenDBF(TDBEnv pEnv, SDBFile *pDBF) {
// TDBEnv is shared by a group of SDBFile
if (!pEnv) {
terrno = TSDB_CODE_INVALID_PTR;
return -1;
}
// Open DBF
if (tsdbOpenBDBDb(&(pDBF->pDB), pEnv, pDBF->path, false) < 0) {
terrno = TSDB_CODE_TDB_INIT_FAILED;
tsdbCloseBDBDb(pDBF->pDB);
return -1;
}
return 0;
}
void tsdbCloseDBF(SDBFile *pDBF) {
if (pDBF->pDB) {
tsdbCloseBDBDb(pDBF->pDB);
pDBF->pDB = NULL;
}
tfree(pDBF->path);
}
int32_t tsdbOpenBDBEnv(DB_ENV **ppEnv, const char *path) {
int ret = 0;
DB_ENV *pEnv = NULL;
if (path == NULL) return 0;
ret = db_env_create(&pEnv, 0);
if (ret != 0) {
BDB_PERR("Failed to create tsdb env", ret);
return -1;
}
ret = pEnv->open(pEnv, path, DB_CREATE | DB_INIT_CDB | DB_INIT_MPOOL, 0);
if (ret != 0) {
// BDB_PERR("Failed to open tsdb env", ret);
tsdbWarn("Failed to open tsdb env for path %s since %d", path ? path : "NULL", ret);
return -1;
}
*ppEnv = pEnv;
return 0;
}
void tsdbCloseBDBEnv(DB_ENV *pEnv) {
if (pEnv) {
pEnv->close(pEnv, 0);
}
}
static int tsdbOpenBDBDb(DB **ppDB, DB_ENV *pEnv, const char *pFName, bool isDup) {
int ret;
DB *pDB;
ret = db_create(&(pDB), pEnv, 0);
if (ret != 0) {
BDB_PERR("Failed to create DBP", ret);
return -1;
}
if (isDup) {
ret = pDB->set_flags(pDB, DB_DUPSORT);
if (ret != 0) {
BDB_PERR("Failed to set DB flags", ret);
return -1;
}
}
ret = pDB->open(pDB, NULL, pFName, NULL, DB_BTREE, DB_CREATE, 0);
if (ret) {
BDB_PERR("Failed to open DBF", ret);
return -1;
}
*ppDB = pDB;
return 0;
}
static void tsdbCloseBDBDb(DB *pDB) {
if (pDB) {
pDB->close(pDB, 0);
}
}
int32_t tsdbSaveSmaToDB(SDBFile *pDBF, void *key, uint32_t keySize, void *data, uint32_t dataSize) {
int ret;
DBT key1 = {0}, value1 = {0};
key1.data = key;
key1.size = keySize;
value1.data = data;
value1.size = dataSize;
// TODO: lock
ret = pDBF->pDB->put(pDBF->pDB, NULL, &key1, &value1, 0);
if (ret) {
BDB_PERR("Failed to put data to DBF", ret);
// TODO: unlock
return -1;
}
// TODO: unlock
return 0;
}
void *tsdbGetSmaDataByKey(SDBFile *pDBF, void* key, uint32_t keySize, uint32_t *valueSize) {
void *result = NULL;
DBT key1 = {0};
DBT value1 = {0};
int ret;
// Set key/value
key1.data = key;
key1.size = keySize;
// Query
// TODO: lock
ret = pDBF->pDB->get(pDBF->pDB, NULL, &key1, &value1, 0);
// TODO: unlock
if (ret != 0) {
return NULL;
}
result = calloc(1, value1.size);
if (result == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
}
*valueSize = value1.size;
memcpy(result, value1.data, value1.size);
return result;
}

View File

@ -80,6 +80,8 @@ static STsdb *tsdbNew(const char *path, int32_t vgId, const STsdbCfg *pTsdbCfg,
pTsdb->pmaf = pMAF;
pTsdb->pMeta = pMeta;
pTsdb->pTfs = pTfs;
pTsdb->pTSmaEnv = NULL;
pTsdb->pRSmaEnv = NULL;
pTsdb->fs = tsdbNewFS(pTsdbCfg);
@ -88,8 +90,9 @@ static STsdb *tsdbNew(const char *path, int32_t vgId, const STsdbCfg *pTsdbCfg,
static void tsdbFree(STsdb *pTsdb) {
if (pTsdb) {
tsdbFreeSmaEnv(pTsdb->pRSmaEnv);
tsdbFreeSmaEnv(pTsdb->pTSmaEnv);
tsdbFreeFS(pTsdb->fs);
tsdbDestroySmaState(pTsdb->pSmaStat);
tfree(pTsdb->path);
free(pTsdb);
}
@ -105,6 +108,30 @@ static void tsdbCloseImpl(STsdb *pTsdb) {
tsdbCloseFS(pTsdb);
// TODO
}
int tsdbLockRepo(STsdb *pTsdb) {
int code = pthread_mutex_lock(&pTsdb->mutex);
if (code != 0) {
tsdbError("vgId:%d failed to lock tsdb since %s", REPO_ID(pTsdb), strerror(errno));
terrno = TAOS_SYSTEM_ERROR(code);
return -1;
}
pTsdb->repoLocked = true;
return 0;
}
int tsdbUnlockRepo(STsdb *pTsdb) {
ASSERT(IS_REPO_LOCKED(pTsdb));
pTsdb->repoLocked = false;
int code = pthread_mutex_unlock(&pTsdb->mutex);
if (code != 0) {
tsdbError("vgId:%d failed to unlock tsdb since %s", REPO_ID(pTsdb), strerror(errno));
terrno = TAOS_SYSTEM_ERROR(code);
return -1;
}
return 0;
}
#if 0
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>

View File

@ -15,7 +15,9 @@
#include "tsdbDef.h"
#undef SMA_PRINT_DEBUG_LOG
#define SMA_STORAGE_TSDB_DAYS 30
#define SMA_STORAGE_TSDB_TIMES 30
#define SMA_STORAGE_SPLIT_HOURS 24
#define SMA_KEY_LEN 18 // tableUid_colId_TSKEY 8+2+8
@ -23,7 +25,7 @@
#define SMA_STATE_ITEM_HASH_SLOT 32
#define SMA_TEST_INDEX_NAME "smaTestIndexName" // TODO: just for test
#define SMA_TEST_INDEX_UID 123456 // TODO: just for test
#define SMA_TEST_INDEX_UID 2000000001 // TODO: just for test
typedef enum {
SMA_STORAGE_LEVEL_TSDB = 0, // use days of self-defined e.g. vnode${N}/tsdb/tsma/sma_index_uid/v2t200.dat
SMA_STORAGE_LEVEL_DFILESET = 1 // use days of TS data e.g. vnode${N}/tsdb/rsma/sma_index_uid/v2r200.dat
@ -31,23 +33,22 @@ typedef enum {
typedef struct {
STsdb * pTsdb;
char * pDFile; // TODO: use the real DFile type, not char*
int32_t interval; // interval with the precision of DB
// TODO
SDBFile dFile;
int32_t interval; // interval with the precision of DB
} STSmaWriteH;
typedef struct {
int32_t iter;
int32_t fid;
} SmaFsIter;
typedef struct {
STsdb * pTsdb;
char * pDFile; // TODO: use the real DFile type, not char*
SDBFile dFile;
int32_t interval; // interval with the precision of DB
int32_t blockSize; // size of SMA block item
int8_t storageLevel;
int8_t days;
SmaFsIter smaFsIter;
// TODO
} STSmaReadH;
typedef struct {
@ -68,18 +69,117 @@ struct SSmaStat {
};
// declaration of static functions
static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, STSmaDataWrapper *pData);
static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, STSmaDataWrapper *pData);
static int32_t tsdbJudgeStorageLevel(int64_t interval, int8_t intervalUnit);
static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *pData);
static int32_t tsdbInsertTSmaBlocks(void *bTree, const char *smaKey, const char *pData, int32_t dataLen);
static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg);
static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg);
// TODO: This is the basic params, and should wrap the params to a queryHandle.
static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySkey,
int32_t nMaxResult);
static int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, int8_t smaType, char *msg);
static int32_t tsdbInitSmaStat(SSmaStat **pSmaStat);
static int32_t tsdbDestroySmaState(SSmaStat *pSmaStat);
static SSmaEnv *tsdbNewSmaEnv(const STsdb *pTsdb, const char *path);
static int32_t tsdbInitSmaEnv(STsdb *pTsdb, const char *path, SSmaEnv **pEnv);
static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, STSmaDataWrapper *pData);
static void tsdbDestroyTSmaWriteH(STSmaWriteH *pSmaH);
static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, int64_t interval, int8_t intervalUnit);
static int32_t tsdbGetSmaStorageLevel(int64_t interval, int8_t intervalUnit);
static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *pData);
static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t keyLen, void *pData, uint32_t dataLen);
static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision);
static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLevel);
static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t storageLevel, int32_t fid);
static int32_t tsdbInitTSmaFile(STSmaReadH *pSmaH, TSKEY skey);
static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, TSKEY *queryKey);
static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, STSmaDataWrapper *pData);
static int32_t tsdbInitTSmaFile(STSmaReadH *pReadH, STimeWindow *queryWin);
static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, STimeWindow *queryWin);
static SSmaEnv *tsdbNewSmaEnv(const STsdb *pTsdb, const char *path) {
SSmaEnv *pEnv = NULL;
pEnv = (SSmaEnv *)calloc(1, sizeof(SSmaEnv));
if (pEnv == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
}
int code = pthread_rwlock_init(&(pEnv->lock), NULL);
if (code) {
terrno = TAOS_SYSTEM_ERROR(code);
free(pEnv);
return NULL;
}
ASSERT(path && (strlen(path) > 0));
pEnv->path = strdup(path);
if (pEnv->path == NULL) {
tsdbFreeSmaEnv(pEnv);
return NULL;
}
if (tsdbInitSmaStat(&pEnv->pStat) != TSDB_CODE_SUCCESS) {
tsdbFreeSmaEnv(pEnv);
return NULL;
}
if (tsdbOpenBDBEnv(&pEnv->dbEnv, pEnv->path) != TSDB_CODE_SUCCESS) {
tsdbFreeSmaEnv(pEnv);
return NULL;
}
return pEnv;
}
static int32_t tsdbInitSmaEnv(STsdb *pTsdb, const char *path, SSmaEnv **pEnv) {
if (!pEnv) {
terrno = TSDB_CODE_INVALID_PTR;
return TSDB_CODE_FAILED;
}
if (pEnv && *pEnv) {
return TSDB_CODE_SUCCESS;
}
if (tsdbLockRepo(pTsdb) != 0) {
return TSDB_CODE_FAILED;
}
if (*pEnv == NULL) {
if ((*pEnv = tsdbNewSmaEnv(pTsdb, path)) == NULL) {
tsdbUnlockRepo(pTsdb);
return TSDB_CODE_FAILED;
}
}
if (tsdbUnlockRepo(pTsdb) != 0) {
tsdbFreeSmaEnv(*pEnv);
return TSDB_CODE_FAILED;
}
return TSDB_CODE_SUCCESS;
}
/**
* @brief Release resources allocated for its member fields, not including itself.
*
* @param pSmaEnv
* @return int32_t
*/
void tsdbDestroySmaEnv(SSmaEnv *pSmaEnv) {
if (pSmaEnv) {
tsdbDestroySmaState(pSmaEnv->pStat);
tfree(pSmaEnv->pStat);
tfree(pSmaEnv->path);
pthread_rwlock_destroy(&(pSmaEnv->lock));
tsdbCloseBDBEnv(pSmaEnv->dbEnv);
}
}
void *tsdbFreeSmaEnv(SSmaEnv *pSmaEnv) {
tsdbDestroySmaEnv(pSmaEnv);
tfree(pSmaEnv);
return NULL;
}
static int32_t tsdbInitSmaStat(SSmaStat **pSmaStat) {
ASSERT(pSmaStat != NULL);
@ -125,6 +225,12 @@ static SSmaStatItem *tsdbNewSmaStatItem(int8_t state) {
return pItem;
}
/**
* @brief Release resources allocated for its member fields, not including itself.
*
* @param pSmaStat
* @return int32_t
*/
int32_t tsdbDestroySmaState(SSmaStat *pSmaStat) {
if (pSmaStat) {
// TODO: use taosHashSetFreeFp when taosHashSetFreeFp is ready.
@ -135,7 +241,6 @@ int32_t tsdbDestroySmaState(SSmaStat *pSmaStat) {
item = taosHashIterate(pSmaStat->smaStatItems, item);
}
taosHashCleanup(pSmaStat->smaStatItems);
free(pSmaStat);
}
}
@ -143,22 +248,35 @@ int32_t tsdbDestroySmaState(SSmaStat *pSmaStat) {
* @brief Update expired window according to msg from stream computing module.
*
* @param pTsdb
* @param smaType ETsdbSmaType
* @param msg
* @return int32_t
*/
int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, char *msg) {
if (msg == NULL) {
int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, int8_t smaType, char *msg) {
STsdbCfg *pCfg = REPO_CFG(pTsdb);
SSmaEnv * pEnv = NULL;
if (!msg || !pTsdb->pMeta) {
terrno = TSDB_CODE_INVALID_PTR;
return TSDB_CODE_FAILED;
}
// lazy mode
if (tsdbInitSmaStat(&pTsdb->pSmaStat) != TSDB_CODE_SUCCESS) {
char smaPath[TSDB_FILENAME_LEN] = "/proj/.sma/";
if (tsdbInitSmaEnv(pTsdb, smaPath, &pEnv) != TSDB_CODE_SUCCESS) {
return TSDB_CODE_FAILED;
}
if (smaType == TSDB_SMA_TYPE_TIME_RANGE) {
pTsdb->pTSmaEnv = pEnv;
} else if (smaType == TSDB_SMA_TYPE_ROLLUP) {
pTsdb->pRSmaEnv = pEnv;
} else {
ASSERT(0);
}
// TODO: decode the msg => start
int64_t indexUid = SMA_TEST_INDEX_UID;
const char * indexName = SMA_TEST_INDEX_NAME;
int64_t indexUid = SMA_TEST_INDEX_UID;
// const char * indexName = SMA_TEST_INDEX_NAME;
const int32_t SMA_TEST_EXPIRED_WINDOW_SIZE = 10;
TSKEY expiredWindows[SMA_TEST_EXPIRED_WINDOW_SIZE];
int64_t now = taosGetTimestampMs();
@ -167,9 +285,9 @@ int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, char *msg) {
}
// TODO: decode the msg <= end
SHashObj *pItemsHash = pTsdb->pSmaStat->smaStatItems;
SHashObj *pItemsHash = SMA_ENV_STAT_ITEMS(pEnv);
SSmaStatItem *pItem = (SSmaStatItem *)taosHashGet(pItemsHash, indexName, strlen(indexName));
SSmaStatItem *pItem = (SSmaStatItem *)taosHashGet(pItemsHash, &indexUid, sizeof(indexUid));
if (pItem == NULL) {
pItem = tsdbNewSmaStatItem(TSDB_SMA_STAT_EXPIRED); // TODO use the real state
if (pItem == NULL) {
@ -181,20 +299,28 @@ int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, char *msg) {
// cache smaMeta
STSma *pSma = metaGetSmaInfoByIndex(pTsdb->pMeta, indexUid);
if (pSma == NULL) {
terrno = TSDB_CODE_TDB_NO_SMA_INDEX_IN_META;
taosHashCleanup(pItem->expiredWindows);
free(pItem);
tsdbWarn("vgId:%d update expired window failed for smaIndex %" PRIi64 " since %s", REPO_ID(pTsdb), indexUid,
tstrerror(terrno));
return TSDB_CODE_FAILED;
}
pItem->pSma = pSma;
// TODO: change indexName to indexUid
if (taosHashPut(pItemsHash, indexName, strnlen(indexName, TSDB_INDEX_NAME_LEN), &pItem, sizeof(pItem)) != 0) {
if (taosHashPut(pItemsHash, &indexUid, sizeof(indexUid), &pItem, sizeof(pItem)) != 0) {
// If error occurs during put smaStatItem, free the resources of pItem
taosHashCleanup(pItem->expiredWindows);
free(pItem);
return TSDB_CODE_FAILED;
}
}
#if 0
SSmaStatItem *pItem1 = (SSmaStatItem *)taosHashGet(pItemsHash, &indexUid, sizeof(indexUid));
int size1 = taosHashGetSize(pItem1->expiredWindows);
tsdbWarn("vgId:%d smaIndex %" PRIi64 " size is %d before hashPut", REPO_ID(pTsdb), indexUid, size1);
#endif
int8_t state = TSDB_SMA_STAT_EXPIRED;
for (int32_t i = 0; i < SMA_TEST_EXPIRED_WINDOW_SIZE; ++i) {
@ -207,21 +333,28 @@ int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, char *msg) {
// windows failed to put into hash table.
taosHashCleanup(pItem->expiredWindows);
tfree(pItem->pSma);
taosHashRemove(pItemsHash, indexName, sizeof(indexName));
taosHashRemove(pItemsHash, &indexUid, sizeof(indexUid));
return TSDB_CODE_FAILED;
}
}
#if 0
SSmaStatItem *pItem2 = (SSmaStatItem *)taosHashGet(pItemsHash, &indexUid, sizeof(indexUid));
int size2 = taosHashGetSize(pItem1->expiredWindows);
tsdbWarn("vgId:%d smaIndex %" PRIi64 " size is %d after hashPut", REPO_ID(pTsdb), indexUid, size2);
#endif
return TSDB_CODE_SUCCESS;
}
static int32_t tsdbResetExpiredWindow(STsdb *pTsdb, int64_t indexUid, TSKEY skey) {
static int32_t tsdbResetExpiredWindow(SSmaStat *pStat, int64_t indexUid, TSKEY skey) {
SSmaStatItem *pItem = NULL;
if (pTsdb->pSmaStat && pTsdb->pSmaStat->smaStatItems) {
pItem = (SSmaStatItem *)taosHashGet(pTsdb->pSmaStat->smaStatItems, &indexUid, sizeof(indexUid));
// TODO: If HASH_ENTRY_LOCK used, whether rwlock needed to handle cases of removing hashNode?
if (pStat && pStat->smaStatItems) {
pItem = (SSmaStatItem *)taosHashGet(pStat->smaStatItems, &indexUid, sizeof(indexUid));
}
#if 0
if (pItem != NULL) {
// TODO: reset time window for the sma data blocks
if (taosHashRemove(pItem->expiredWindows, &skey, sizeof(TSKEY)) != 0) {
@ -231,6 +364,7 @@ static int32_t tsdbResetExpiredWindow(STsdb *pTsdb, int64_t indexUid, TSKEY skey
} else {
// error handling
}
#endif
return TSDB_CODE_SUCCESS;
}
@ -241,7 +375,7 @@ static int32_t tsdbResetExpiredWindow(STsdb *pTsdb, int64_t indexUid, TSKEY skey
* @param intervalUnit
* @return int32_t
*/
static int32_t tsdbJudgeStorageLevel(int64_t interval, int8_t intervalUnit) {
static int32_t tsdbGetSmaStorageLevel(int64_t interval, int8_t intervalUnit) {
// TODO: configurable for SMA_STORAGE_SPLIT_HOURS?
switch (intervalUnit) {
case TD_TIME_UNIT_HOUR:
@ -281,18 +415,35 @@ static int32_t tsdbJudgeStorageLevel(int64_t interval, int8_t intervalUnit) {
}
/**
* @brief Insert TSma data blocks to B+Tree
* @brief Insert TSma data blocks to DB File build by B+Tree
*
* @param bTree
* @param pSmaH
* @param smaKey
* @param keyLen
* @param pData
* @param dataLen
* @return int32_t
*/
static int32_t tsdbInsertTSmaBlocks(void *bTree, const char *smaKey, const char *pData, int32_t dataLen) {
static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t keyLen, void *pData, uint32_t dataLen) {
SDBFile *pDBFile = &pSmaH->dFile;
// TODO: insert sma data blocks into B+Tree
tsdbDebug("insert sma data blocks into B+Tree: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", dataLen %d",
*(uint64_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8), *(int64_t *)POINTER_SHIFT(smaKey, 10), dataLen);
tsdbDebug("vgId:%d insert sma data blocks into %s: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", dataLen %d",
REPO_ID(pSmaH->pTsdb), pDBFile->path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
*(int64_t *)POINTER_SHIFT(smaKey, 10), dataLen);
if (tsdbSaveSmaToDB(pDBFile, smaKey, keyLen, pData, dataLen) != 0) {
return TSDB_CODE_FAILED;
}
#ifdef SMA_PRINT_DEBUG_LOG
uint32_t valueSize = 0;
void * data = tsdbGetSmaDataByKey(pDBFile, smaKey, keyLen, &valueSize);
ASSERT(data != NULL);
for (uint32_t v = 0; v < valueSize; v += 8) {
tsdbWarn("vgId:%d sma data - val[%d] is %" PRIi64, REPO_ID(pSmaH->pTsdb), v, *(int64_t *)POINTER_SHIFT(data, v));
}
#endif
return TSDB_CODE_SUCCESS;
}
@ -324,41 +475,41 @@ static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit
}
}
switch (intervalUnit) {
case TD_TIME_UNIT_MILLISEC:
if (TSDB_TIME_PRECISION_MILLI == precision) {
return interval;
} else if (TSDB_TIME_PRECISION_MICRO == precision) {
return interval * 1e3;
} else { // nano second
return interval * 1e6;
}
break;
case TD_TIME_UNIT_MICROSEC:
if (TSDB_TIME_PRECISION_MILLI == precision) {
switch (precision) {
case TSDB_TIME_PRECISION_MILLI:
if (TD_TIME_UNIT_MICROSEC == intervalUnit) { // us
return interval / 1e3;
} else if (TSDB_TIME_PRECISION_MICRO == precision) {
return interval;
} else { // nano second
return interval * 1e3;
}
break;
case TD_TIME_UNIT_NANOSEC:
if (TSDB_TIME_PRECISION_MILLI == precision) {
} else if (TD_TIME_UNIT_NANOSEC == intervalUnit) { // nano second
return interval / 1e6;
} else if (TSDB_TIME_PRECISION_MICRO == precision) {
return interval / 1e3;
} else { // nano second
} else {
return interval;
}
break;
default:
if (TSDB_TIME_PRECISION_MILLI == precision) {
case TSDB_TIME_PRECISION_MICRO:
if (TD_TIME_UNIT_MICROSEC == intervalUnit) { // us
return interval;
} else if (TD_TIME_UNIT_NANOSEC == intervalUnit) { // nano second
return interval / 1e3;
} else {
return interval * 1e3;
} else if (TSDB_TIME_PRECISION_MICRO == precision) {
}
break;
case TSDB_TIME_PRECISION_NANO:
if (TD_TIME_UNIT_MICROSEC == intervalUnit) {
return interval * 1e3;
} else if (TD_TIME_UNIT_NANOSEC == intervalUnit) { // nano second
return interval;
} else {
return interval * 1e6;
} else { // nano second
return interval * 1e9;
}
break;
default: // ms
if (TD_TIME_UNIT_MICROSEC == intervalUnit) { // us
return interval / 1e3;
} else if (TD_TIME_UNIT_NANOSEC == intervalUnit) { // nano second
return interval / 1e6;
} else {
return interval;
}
break;
}
@ -381,8 +532,6 @@ static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *p
// TODO: check the data integrity
void *bTree = pSmaH->pDFile;
int32_t len = 0;
while (true) {
if (len >= pData->dataLen) {
@ -405,7 +554,7 @@ static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *p
pData->indexUid, pData->skey, pTbData->tableUid, pColData->colId);
#endif
tsdbEncodeTSmaKey(pTbData->tableUid, pColData->colId, pData->skey, (void **)&pSmaKey);
if (tsdbInsertTSmaBlocks(bTree, smaKey, pColData->data, pColData->blockSize) < 0) {
if (tsdbInsertTSmaBlocks(pSmaH, smaKey, SMA_KEY_LEN, pColData->data, pColData->blockSize) < 0) {
tsdbWarn("vgId:%d insert tSma blocks failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
tbLen += (sizeof(STSmaColData) + pColData->blockSize);
@ -419,15 +568,43 @@ static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *p
static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, STSmaDataWrapper *pData) {
pSmaH->pTsdb = pTsdb;
pSmaH->interval = tsdbGetIntervalByPrecision(pData->interval, pData->intervalUnit, REPO_CFG(pTsdb)->precision);
return TSDB_CODE_SUCCESS;
}
static void tsdbDestroyTSmaWriteH(STSmaWriteH *pSmaH) {
if (pSmaH) {
tsdbCloseDBF(&pSmaH->dFile);
}
}
static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t storageLevel, int32_t fid) {
// TODO
pSmaH->pDFile = "tSma_interval_file_name";
STsdb *pTsdb = pSmaH->pTsdb;
ASSERT(pSmaH->dFile.path == NULL && pSmaH->dFile.pDB == NULL);
char tSmaFile[TSDB_FILENAME_LEN] = {0};
snprintf(tSmaFile, TSDB_FILENAME_LEN, "v%df%d.tsma", REPO_ID(pTsdb), fid);
pSmaH->dFile.path = strdup(tSmaFile);
return TSDB_CODE_SUCCESS;
}
}
/**
* @brief
*
* @param pTsdb
* @param interval Interval calculated by DB's precision
* @param storageLevel
* @return int32_t
*/
static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLevel) {
STsdbCfg *pCfg = REPO_CFG(pTsdb);
int32_t daysPerFile = pCfg->daysPerFile;
if (storageLevel == SMA_STORAGE_LEVEL_TSDB) {
int32_t days = SMA_STORAGE_TSDB_TIMES * (interval / tsTickPerDay[pCfg->precision]);
daysPerFile = days > SMA_STORAGE_TSDB_DAYS ? days : SMA_STORAGE_TSDB_DAYS;
}
return daysPerFile;
}
/**
* @brief Insert/Update Time-range-wise SMA data.
@ -441,51 +618,72 @@ static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData,
* @param msg
* @return int32_t
*/
int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
STsdbCfg * pCfg = REPO_CFG(pTsdb);
static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
STsdbCfg * pCfg = REPO_CFG(pTsdb);
STSmaDataWrapper *pData = (STSmaDataWrapper *)msg;
STSmaWriteH tSmaH = {0};
tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData);
if (!pTsdb->pTSmaEnv) {
terrno = TSDB_CODE_INVALID_PTR;
tsdbWarn("vgId:%d insert tSma data failed since pTSmaEnv is NULL", REPO_ID(pTsdb));
return terrno;
}
if (pData->dataLen <= 0) {
TASSERT(0);
terrno = TSDB_CODE_INVALID_PARA;
return terrno;
return TSDB_CODE_FAILED;
}
// Step 1: Judge the storage level
int32_t storageLevel = tsdbJudgeStorageLevel(pData->interval, pData->intervalUnit);
int32_t daysPerFile = storageLevel == SMA_STORAGE_LEVEL_TSDB ? SMA_STORAGE_TSDB_DAYS : pCfg->daysPerFile;
STSmaWriteH tSmaH = {0};
if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData) != 0) {
return TSDB_CODE_FAILED;
}
// Step 1: Judge the storage level and days
int32_t storageLevel = tsdbGetSmaStorageLevel(pData->interval, pData->intervalUnit);
int32_t daysPerFile = tsdbGetTSmaDays(pTsdb, tSmaH.interval, storageLevel);
int32_t fid = (int32_t)(TSDB_KEY_FID(pData->skey, daysPerFile, pCfg->precision));
// Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
// - Set and open the DFile or the B+Tree file
int32_t fid = (int32_t)(TSDB_KEY_FID(pData->skey, daysPerFile, pCfg->precision));
// Save all the TSma data to one file
// TODO: tsdbStartTSmaCommit();
tsdbSetTSmaDataFile(&tSmaH, pData, storageLevel, fid);
tsdbInsertTSmaDataSection(&tSmaH, pData);
if (tsdbOpenDBF(pTsdb->pTSmaEnv->dbEnv, &tSmaH.dFile) != 0) {
tsdbWarn("vgId:%d open DB file %s failed since %s", REPO_ID(pTsdb),
tSmaH.dFile.path ? tSmaH.dFile.path : "path is NULL", tstrerror(terrno));
tsdbDestroyTSmaWriteH(&tSmaH);
return TSDB_CODE_FAILED;
}
if (tsdbInsertTSmaDataSection(&tSmaH, pData) != 0) {
tsdbWarn("vgId:%d insert tSma data section failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
tsdbDestroyTSmaWriteH(&tSmaH);
return TSDB_CODE_FAILED;
}
// TODO:tsdbEndTSmaCommit();
// reset the SSmaStat
tsdbResetExpiredWindow(pTsdb, pData->indexUid, pData->skey);
// Step 3: reset the SSmaStat
tsdbResetExpiredWindow(SMA_ENV_STAT(pTsdb->pTSmaEnv), pData->indexUid, pData->skey);
tsdbDestroyTSmaWriteH(&tSmaH);
return TSDB_CODE_SUCCESS;
}
static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t fid) {
// TODO
pSmaH->pDFile = "rSma_interval_file_name";
STsdb *pTsdb = pSmaH->pTsdb;
char tSmaFile[TSDB_FILENAME_LEN] = {0};
snprintf(tSmaFile, TSDB_FILENAME_LEN, "v%df%d.rsma", REPO_ID(pTsdb), fid);
pSmaH->dFile.path = strdup(tSmaFile);
return TSDB_CODE_SUCCESS;
}
int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
STsdbCfg * pCfg = REPO_CFG(pTsdb);
static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
STsdbCfg * pCfg = REPO_CFG(pTsdb);
STSmaDataWrapper *pData = (STSmaDataWrapper *)msg;
STSmaWriteH tSmaH = {0};
STSmaWriteH tSmaH = {0};
tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData);
@ -496,7 +694,7 @@ int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
}
// Step 1: Judge the storage level
int32_t storageLevel = tsdbJudgeStorageLevel(pData->interval, pData->intervalUnit);
int32_t storageLevel = tsdbGetSmaStorageLevel(pData->interval, pData->intervalUnit);
int32_t daysPerFile = storageLevel == SMA_STORAGE_LEVEL_TSDB ? SMA_STORAGE_TSDB_DAYS : pCfg->daysPerFile;
// Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
@ -507,45 +705,46 @@ int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
// Save all the TSma data to one file
// TODO: tsdbStartTSmaCommit();
tsdbSetTSmaDataFile(&tSmaH, pData, storageLevel, fid);
tsdbInsertTSmaDataSection(&tSmaH, pData);
// TODO:tsdbEndTSmaCommit();
// reset the SSmaStat
tsdbResetExpiredWindow(pTsdb, pData->indexUid, pData->skey);
tsdbResetExpiredWindow(SMA_ENV_STAT(pTsdb->pRSmaEnv), pData->indexUid, pData->skey);
return TSDB_CODE_SUCCESS;
}
/**
* @brief Init of tSma ReadH
* @brief
*
* @param pSmaH
* @param pTsdb
* @param param
* @param pData
* @param interval
* @param intervalUnit
* @return int32_t
*/
static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, STSmaDataWrapper *pData) {
static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, int64_t interval, int8_t intervalUnit) {
pSmaH->pTsdb = pTsdb;
pSmaH->interval = tsdbGetIntervalByPrecision(pData->interval, pData->intervalUnit, REPO_CFG(pTsdb)->precision);
// pSmaH->blockSize = param->numOfFuncIds * sizeof(int64_t);
pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision);
pSmaH->storageLevel = tsdbGetSmaStorageLevel(interval, intervalUnit);
pSmaH->days = tsdbGetTSmaDays(pTsdb, pSmaH->interval, pSmaH->storageLevel);
}
/**
* @brief Init of tSma FS
*
* @param pReadH
* @param param
* @param queryWin
* @param skey
* @return int32_t
*/
static int32_t tsdbInitTSmaFile(STSmaReadH *pReadH, STimeWindow *queryWin) {
int32_t storageLevel = 0; //tsdbJudgeStorageLevel(param->interval, param->intervalUnit);
int32_t daysPerFile =
storageLevel == SMA_STORAGE_LEVEL_TSDB ? SMA_STORAGE_TSDB_DAYS : REPO_CFG(pReadH->pTsdb)->daysPerFile;
pReadH->storageLevel = storageLevel;
pReadH->days = daysPerFile;
pReadH->smaFsIter.iter = 0;
static int32_t tsdbInitTSmaFile(STSmaReadH *pSmaH, TSKEY skey) {
int32_t fid = (int32_t)(TSDB_KEY_FID(skey, pSmaH->days, REPO_CFG(pSmaH->pTsdb)->precision));
char tSmaFile[TSDB_FILENAME_LEN] = {0};
snprintf(tSmaFile, TSDB_FILENAME_LEN, "v%df%d.tsma", REPO_ID(pSmaH->pTsdb), fid);
pSmaH->dFile.path = strdup(tSmaFile);
pSmaH->smaFsIter.iter = 0;
pSmaH->smaFsIter.fid = fid;
}
/**
@ -557,17 +756,18 @@ static int32_t tsdbInitTSmaFile(STSmaReadH *pReadH, STimeWindow *queryWin) {
* @return true
* @return false
*/
static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, STimeWindow *queryWin) {
static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, TSKEY *queryKey) {
SArray *smaFs = pReadH->pTsdb->fs->cstatus->sf;
int32_t nSmaFs = taosArrayGetSize(smaFs);
pReadH->pDFile = NULL;
tsdbCloseDBF(&pReadH->dFile);
#if 0
while (pReadH->smaFsIter.iter < nSmaFs) {
void *pSmaFile = taosArrayGet(smaFs, pReadH->smaFsIter.iter);
if (pSmaFile) { // match(indexName, queryWindow)
// TODO: select the file by index_name ...
pReadH->pDFile = pSmaFile;
pReadH->dFile = pSmaFile;
++pReadH->smaFsIter.iter;
break;
}
@ -578,41 +778,83 @@ static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, STimeWindow *queryWin) {
tsdbDebug("vg%d: smaFile %s matched", REPO_ID(pReadH->pTsdb), "[pSmaFile dir]");
return true;
}
#endif
return false;
}
/**
* @brief Return the data between queryWin and fill the pData.
* @brief
*
* @param pTsdb
* @param param
* @param pTsdb Return the data between queryWin and fill the pData.
* @param pData
* @param queryWin
* @param indexUid
* @param interval
* @param intervalUnit
* @param tableUid
* @param colId
* @param pQuerySKey
* @param nMaxResult The query invoker should control the nMaxResult need to return to avoid OOM.
* @return int32_t
*/
int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, STimeWindow *queryWin, int32_t nMaxResult) {
SSmaStatItem *pItem =
(SSmaStatItem *)taosHashGet(pTsdb->pSmaStat->smaStatItems, &pData->indexUid, sizeof(pData->indexUid));
static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySkey,
int32_t nMaxResult) {
SSmaStatItem *pItem = (SSmaStatItem *)taosHashGet(SMA_ENV_STAT_ITEMS(pTsdb->pTSmaEnv), &indexUid, sizeof(indexUid));
if (pItem == NULL) {
// mark all window as expired and notify query module to query raw TS data.
return TSDB_CODE_SUCCESS;
}
int32_t nQueryWin = 0;
#if 0
int32_t nQueryWin = taosArrayGetSize(pQuerySKey);
for (int32_t n = 0; n < nQueryWin; ++n) {
TSKEY thisWindow = n;
if (taosHashGet(pItem->expiredWindows, &thisWindow, sizeof(thisWindow)) != NULL) {
TSKEY skey = taosArrayGet(pQuerySKey, n);
if (taosHashGet(pItem->expiredWindows, &skey, sizeof(TSKEY)) != NULL) {
// TODO: mark this window as expired.
}
}
#endif
#if 0
if (taosHashGet(pItem->expiredWindows, &querySkey, sizeof(TSKEY)) != NULL) {
// TODO: mark this window as expired.
}
#endif
STSmaReadH tReadH = {0};
tsdbInitTSmaReadH(&tReadH, pTsdb, pData);
tsdbInitTSmaReadH(&tReadH, pTsdb, interval, intervalUnit);
tsdbCloseDBF(&tReadH.dFile);
tsdbInitTSmaFile(&tReadH, queryWin);
tsdbInitTSmaFile(&tReadH, querySkey);
if (tsdbOpenDBF(SMA_ENV_ENV(pTsdb->pTSmaEnv), &tReadH.dFile) != 0) {
tsdbWarn("vgId:%d open DBF %s failed since %s", REPO_ID(pTsdb), tReadH.dFile.path, tstrerror(terrno));
return TSDB_CODE_FAILED;
}
char smaKey[SMA_KEY_LEN] = {0};
void *pSmaKey = &smaKey;
tsdbEncodeTSmaKey(tableUid, colId, querySkey, (void **)&pSmaKey);
tsdbDebug("vgId:%d get sma data from %s: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", keyLen %d", REPO_ID(pTsdb),
tReadH.dFile.path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
*(int64_t *)POINTER_SHIFT(smaKey, 10), SMA_KEY_LEN);
void * result = NULL;
uint32_t valueSize = 0;
if ((result = tsdbGetSmaDataByKey(&tReadH.dFile, smaKey, SMA_KEY_LEN, &valueSize)) == NULL) {
tsdbWarn("vgId:%d get sma data failed from smaIndex %" PRIi64 ", smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64
" since %s",
REPO_ID(pTsdb), indexUid, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
*(int64_t *)POINTER_SHIFT(smaKey, 10), tstrerror(terrno));
tsdbCloseDBF(&tReadH.dFile);
return TSDB_CODE_FAILED;
}
tfree(result);
#ifdef SMA_PRINT_DEBUG_LOG
for (uint32_t v = 0; v < valueSize; v += 8) {
tsdbWarn("vgId:%d v[%d]=%" PRIi64, REPO_ID(pTsdb), v, *(int64_t *)POINTER_SHIFT(result, v));
}
#endif
#if 0
int32_t nResult = 0;
int64_t lastKey = 0;
@ -634,8 +876,9 @@ int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, STimeWindow *
}
}
}
#endif
// read data from file and fill the result
tsdbCloseDBF(&tReadH.dFile);
return TSDB_CODE_SUCCESS;
}
@ -673,4 +916,55 @@ int32_t tsdbRemoveTSmaData(STsdb *pTsdb, void *smaIndex, STimeWindow *pWin) {
// }
return TSDB_CODE_SUCCESS;
}
#endif
#endif
/**
* @brief Insert/Update tSma(Time-range-wise SMA) data from stream computing engine
*
* @param pTsdb
* @param param
* @param msg
* @return int32_t
* TODO: Who is responsible for resource allocate and release?
*/
int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
int32_t code = TSDB_CODE_SUCCESS;
if ((code = tsdbInsertTSmaDataImpl(pTsdb, msg)) < 0) {
tsdbWarn("vgId:%d insert tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
return code;
}
int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg) {
int32_t code = TSDB_CODE_SUCCESS;
if ((code = tsdbUpdateExpiredWindow(pTsdb, smaType, msg)) < 0) {
tsdbWarn("vgId:%d update expired sma window failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
return code;
}
/**
* @brief Insert Time-range-wise Rollup Sma(RSma) data
*
* @param pTsdb
* @param param
* @param msg
* @return int32_t
*/
int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
int32_t code = TSDB_CODE_SUCCESS;
if ((code = tsdbInsertRSmaDataImpl(pTsdb, msg)) < 0) {
tsdbWarn("vgId:%d insert rSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
return code;
}
int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
tb_uid_t tableUid, col_id_t colId, TSKEY querySkey, int32_t nMaxResult) {
int32_t code = TSDB_CODE_SUCCESS;
if ((code = tsdbGetTSmaDataImpl(pTsdb, pData, indexUid, interval, intervalUnit, tableUid, colId, querySkey,
nMaxResult)) < 0) {
tsdbWarn("vgId:%d get tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
return code;
}

View File

@ -34,6 +34,7 @@ int tsdbInsertData(STsdb *pTsdb, SSubmitReq *pMsg, SSubmitRsp *pRsp) {
return tsdbMemTableInsert(pTsdb, pTsdb->mem, pMsg, NULL);
}
#if 0
/**
* @brief Insert/Update tSma(Time-range-wise SMA) data from stream computing engine
*
@ -51,6 +52,14 @@ int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
return code;
}
int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg) {
int32_t code = TSDB_CODE_SUCCESS;
if ((code = tsdbUpdateExpiredWindow(pTsdb, smaType, msg)) < 0) {
tsdbWarn("vgId:%d update expired sma window failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
return code;
}
/**
* @brief Insert Time-range-wise Rollup Sma(RSma) data
*
@ -65,4 +74,6 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
tsdbWarn("vgId:%d insert rSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
}
return code;
}
}
#endif

View File

@ -33,7 +33,7 @@ int main(int argc, char **argv) {
return RUN_ALL_TESTS();
}
TEST(testCase, tSmaEncodeDecodeTest) {
TEST(testCase, tSma_Meta_Encode_Decode_Test) {
// encode
STSma tSma = {0};
tSma.version = 0;
@ -87,8 +87,9 @@ TEST(testCase, tSmaEncodeDecodeTest) {
tdDestroyTSma(&tSma);
tdDestroyTSmaWrapper(&dstTSmaWrapper);
}
#if 1
TEST(testCase, tSma_DB_Put_Get_Del_Test) {
TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
const char * smaIndexName1 = "sma_index_test_1";
const char * smaIndexName2 = "sma_index_test_2";
const char * timezone = "Asia/Shanghai";
@ -220,13 +221,84 @@ TEST(testCase, tSma_DB_Put_Get_Del_Test) {
#endif
#if 1
TEST(testCase, tSmaInsertTest) {
const int64_t indexUid = 2000000002;
TEST(testCase, tSma_Data_Insert_Query_Test) {
// step 1: prepare meta
const char * smaIndexName1 = "sma_index_test_1";
const char * timezone = "Asia/Shanghai";
const char * expr = "select count(a,b, top 20), from table interval 1d, sliding 1h;";
const char * tagsFilter = "where tags.location='Beijing' and tags.district='ChaoYang'";
const char * smaTestDir = "./smaTest";
const tb_uid_t tbUid = 1234567890;
const int64_t indexUid1 = 2000000001;
const int64_t interval1 = 1;
const int8_t intervalUnit1 = TD_TIME_UNIT_DAY;
const uint32_t nCntTSma = 2;
TSKEY skey1 = 1646987196;
const int64_t testSmaData1 = 100;
const int64_t testSmaData2 = 200;
// encode
STSma tSma = {0};
tSma.version = 0;
tSma.intervalUnit = TD_TIME_UNIT_DAY;
tSma.interval = 1;
tSma.slidingUnit = TD_TIME_UNIT_HOUR;
tSma.sliding = 0;
tSma.indexUid = indexUid1;
tstrncpy(tSma.indexName, smaIndexName1, TSDB_INDEX_NAME_LEN);
tstrncpy(tSma.timezone, timezone, TD_TIMEZONE_LEN);
tSma.tableUid = tbUid;
tSma.exprLen = strlen(expr);
tSma.expr = (char *)calloc(tSma.exprLen + 1, 1);
tstrncpy(tSma.expr, expr, tSma.exprLen + 1);
tSma.tagsFilterLen = strlen(tagsFilter);
tSma.tagsFilter = (char *)calloc(tSma.tagsFilterLen + 1, 1);
tstrncpy(tSma.tagsFilter, tagsFilter, tSma.tagsFilterLen + 1);
SMeta * pMeta = NULL;
STSma * pSmaCfg = &tSma;
const SMetaCfg *pMetaCfg = &defaultMetaOptions;
taosRemoveDir(smaTestDir);
pMeta = metaOpen(smaTestDir, pMetaCfg, NULL);
assert(pMeta != NULL);
// save index 1
EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
// step 2: insert data
STSmaDataWrapper *pSmaData = NULL;
STsdb tsdb = {0};
STsdbCfg * pCfg = &tsdb.config;
pCfg->daysPerFile = 1;
tsdb.pMeta = pMeta;
tsdb.vgId = 2;
tsdb.config.daysPerFile = 10; // default days is 10
tsdb.config.keep1 = 30;
tsdb.config.keep2 = 90;
tsdb.config.keep = 365;
tsdb.config.precision = TSDB_TIME_PRECISION_MILLI;
tsdb.config.update = TD_ROW_OVERWRITE_UPDATE;
tsdb.config.compression = TWO_STAGE_COMP;
switch (tsdb.config.precision) {
case TSDB_TIME_PRECISION_MILLI:
skey1 *= 1e3;
break;
case TSDB_TIME_PRECISION_MICRO:
skey1 *= 1e6;
break;
case TSDB_TIME_PRECISION_NANO:
skey1 *= 1e9;
break;
default: // ms
skey1 *= 1e3;
break;
}
char *msg = (char *)calloc(100, 1);
EXPECT_EQ(tsdbUpdateSmaWindow(&tsdb, TSDB_SMA_TYPE_TIME_RANGE, msg), 0);
// init
int32_t allocCnt = 0;
@ -235,21 +307,21 @@ TEST(testCase, tSmaInsertTest) {
void * buf = NULL;
EXPECT_EQ(tsdbMakeRoom(&buf, allocStep), 0);
int32_t bufSize = taosTSizeof(buf);
int32_t numOfTables = 25;
int32_t numOfTables = 10;
col_id_t numOfCols = 4096;
EXPECT_GT(numOfCols, 0);
pSmaData = (STSmaDataWrapper *)buf;
printf(">> allocate [%d] time to %d and addr is %p\n", ++allocCnt, bufSize, pSmaData);
pSmaData->skey = 1646987196;
pSmaData->interval = 10;
pSmaData->intervalUnit = TD_TIME_UNIT_MINUTE;
pSmaData->indexUid = indexUid;
pSmaData->skey = skey1;
pSmaData->interval = interval1;
pSmaData->intervalUnit = intervalUnit1;
pSmaData->indexUid = indexUid1;
int32_t len = sizeof(STSmaDataWrapper);
for (int32_t t = 0; t < numOfTables; ++t) {
STSmaTbData *pTbData = (STSmaTbData *)POINTER_SHIFT(pSmaData, len);
pTbData->tableUid = t;
pTbData->tableUid = tbUid + t;
int32_t tableDataLen = sizeof(STSmaTbData);
for (col_id_t c = 0; c < numOfCols; ++c) {
@ -262,8 +334,17 @@ TEST(testCase, tSmaInsertTest) {
}
STSmaColData *pColData = (STSmaColData *)POINTER_SHIFT(pSmaData, len + tableDataLen);
pColData->colId = c + PRIMARYKEY_TIMESTAMP_COL_ID;
pColData->blockSize = ((c & 1) == 0) ? 8 : 16;
// TODO: fill col data
if ((c & 1) == 0) {
pColData->blockSize = 8;
memcpy(pColData->data, &testSmaData1, 8);
} else {
pColData->blockSize = 16;
memcpy(pColData->data, &testSmaData1, 8);
memcpy(POINTER_SHIFT(pColData->data, 8), &testSmaData2, 8);
}
tableDataLen += (sizeof(STSmaColData) + pColData->blockSize);
}
pTbData->dataLen = (tableDataLen - sizeof(STSmaTbData));
@ -277,8 +358,24 @@ TEST(testCase, tSmaInsertTest) {
// execute
EXPECT_EQ(tsdbInsertTSmaData(&tsdb, (char *)pSmaData), TSDB_CODE_SUCCESS);
// release
// step 3: query
uint32_t checkDataCnt = 0;
for (int32_t t = 0; t < numOfTables; ++t) {
for (col_id_t c = 0; c < numOfCols; ++c) {
EXPECT_EQ(tsdbGetTSmaData(&tsdb, NULL, indexUid1, interval1, intervalUnit1, tbUid + t,
c + PRIMARYKEY_TIMESTAMP_COL_ID, skey1, 1),
TSDB_CODE_SUCCESS);
++checkDataCnt;
}
}
printf("%s:%d The sma data check count for insert and query is %" PRIu32 "\n", __FILE__, __LINE__, checkDataCnt);
// release data
taosTZfree(buf);
// release meta
tdDestroyTSma(&tSma);
metaClose(pMeta);
}
#endif

File diff suppressed because it is too large Load Diff

View File

@ -39,7 +39,6 @@ extern "C" {
// /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode);
int32_t syncNodeElect(SSyncNode* pSyncNode);
int32_t syncNodeRequestVote(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncRequestVote* pMsg);

View File

@ -29,6 +29,7 @@ extern "C" {
#include "ttimer.h"
#define TIMER_MAX_MS 0x7FFFFFFF
#define ENV_TICK_TIMER_MS 1000
#define PING_TIMER_MS 1000
#define ELECT_TIMER_MS_MIN 150
#define ELECT_TIMER_MS_MAX 300
@ -38,17 +39,28 @@ extern "C" {
#define EMPTY_RAFT_ID ((SRaftId){.addr = 0, .vgId = 0})
typedef struct SSyncEnv {
tmr_h pEnvTickTimer;
// tick timer
tmr_h pEnvTickTimer;
int32_t envTickTimerMS;
uint64_t envTickTimerLogicClock; // if use queue, should pass logic clock into queue item
uint64_t envTickTimerLogicClockUser;
TAOS_TMR_CALLBACK FpEnvTickTimer; // Timer Fp
uint64_t envTickTimerCounter;
// timer manager
tmr_h pTimerManager;
char name[128];
// other resources shared by SyncNodes
// ...
} SSyncEnv;
extern SSyncEnv* gSyncEnv;
int32_t syncEnvStart();
int32_t syncEnvStop();
tmr_h syncEnvStartTimer(TAOS_TMR_CALLBACK fp, int mseconds, void* param);
void syncEnvStopTimer(tmr_h* pTimer);
int32_t syncEnvStartTimer();
int32_t syncEnvStopTimer();
#ifdef __cplusplus
}

View File

@ -29,22 +29,28 @@ extern "C" {
#include "tqueue.h"
#include "trpc.h"
#define TICK_Q_TIMER_MS 1000
#define TICK_Ping_TIMER_MS 1000
typedef struct SSyncIO {
STaosQueue *pMsgQ;
STaosQset * pQset;
STaosQset *pQset;
pthread_t consumerTid;
void * serverRpc;
void * clientRpc;
void *serverRpc;
void *clientRpc;
SEpSet myAddr;
void *ioTimerTickQ;
void *ioTimerTickPing;
void *ioTimerManager;
tmr_h qTimer;
int32_t qTimerMS;
tmr_h pingTimer;
int32_t pingTimerMS;
tmr_h timerMgr;
void *pSyncNode;
int32_t (*FpOnSyncPing)(SSyncNode *pSyncNode, SyncPing *pMsg);
int32_t (*FpOnSyncPingReply)(SSyncNode *pSyncNode, SyncPingReply *pMsg);
int32_t (*FpOnSyncClientRequest)(SSyncNode *pSyncNode, SyncClientRequest *pMsg);
int32_t (*FpOnSyncRequestVote)(SSyncNode *pSyncNode, SyncRequestVote *pMsg);
int32_t (*FpOnSyncRequestVoteReply)(SSyncNode *pSyncNode, SyncRequestVoteReply *pMsg);
int32_t (*FpOnSyncAppendEntries)(SSyncNode *pSyncNode, SyncAppendEntries *pMsg);
@ -59,11 +65,14 @@ extern SSyncIO *gSyncIO;
int32_t syncIOStart(char *host, uint16_t port);
int32_t syncIOStop();
int32_t syncIOTickQ();
int32_t syncIOTickPing();
int32_t syncIOSendMsg(void *clientRpc, const SEpSet *pEpSet, SRpcMsg *pMsg);
int32_t syncIOEqMsg(void *queue, SRpcMsg *pMsg);
int32_t syncIOQTimerStart();
int32_t syncIOQTimerStop();
int32_t syncIOPingTimerStart();
int32_t syncIOPingTimerStop();
#ifdef __cplusplus
}
#endif

View File

@ -67,12 +67,12 @@ extern "C" {
} \
}
struct SRaft;
typedef struct SRaft SRaft;
struct SyncTimeout;
typedef struct SyncTimeout SyncTimeout;
struct SyncClientRequest;
typedef struct SyncClientRequest SyncClientRequest;
struct SyncPing;
typedef struct SyncPing SyncPing;
@ -117,8 +117,10 @@ typedef struct SSyncNode {
SSyncCfg syncCfg;
char path[TSDB_FILENAME_LEN];
char raftStorePath[TSDB_FILENAME_LEN * 2];
SWal* pWal;
void* rpcClient;
// sync io
SWal* pWal;
void* rpcClient;
int32_t (*FpSendMsg)(void* rpcClient, const SEpSet* pEpSet, SRpcMsg* pMsg);
void* queue;
int32_t (*FpEqMsg)(void* queue, SRpcMsg* pMsg);
@ -164,7 +166,7 @@ typedef struct SSyncNode {
int32_t pingTimerMS;
uint64_t pingTimerLogicClock;
uint64_t pingTimerLogicClockUser;
TAOS_TMR_CALLBACK FpPingTimer; // Timer Fp
TAOS_TMR_CALLBACK FpPingTimerCB; // Timer Fp
uint64_t pingTimerCounter;
// elect timer
@ -172,7 +174,7 @@ typedef struct SSyncNode {
int32_t electTimerMS;
uint64_t electTimerLogicClock;
uint64_t electTimerLogicClockUser;
TAOS_TMR_CALLBACK FpElectTimer; // Timer Fp
TAOS_TMR_CALLBACK FpElectTimerCB; // Timer Fp
uint64_t electTimerCounter;
// heartbeat timer
@ -180,12 +182,13 @@ typedef struct SSyncNode {
int32_t heartbeatTimerMS;
uint64_t heartbeatTimerLogicClock;
uint64_t heartbeatTimerLogicClockUser;
TAOS_TMR_CALLBACK FpHeartbeatTimer; // Timer Fp
TAOS_TMR_CALLBACK FpHeartbeatTimerCB; // Timer Fp
uint64_t heartbeatTimerCounter;
// callback
int32_t (*FpOnPing)(SSyncNode* ths, SyncPing* pMsg);
int32_t (*FpOnPingReply)(SSyncNode* ths, SyncPingReply* pMsg);
int32_t (*FpOnClientRequest)(SSyncNode* ths, SyncClientRequest* pMsg);
int32_t (*FpOnRequestVote)(SSyncNode* ths, SyncRequestVote* pMsg);
int32_t (*FpOnRequestVoteReply)(SSyncNode* ths, SyncRequestVoteReply* pMsg);
int32_t (*FpOnAppendEntries)(SSyncNode* ths, SyncAppendEntries* pMsg);
@ -194,26 +197,47 @@ typedef struct SSyncNode {
} SSyncNode;
// open/close --------------
SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo);
void syncNodeClose(SSyncNode* pSyncNode);
int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg);
int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg);
// ping --------------
int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg);
int32_t syncNodePingAll(SSyncNode* pSyncNode);
int32_t syncNodePingPeers(SSyncNode* pSyncNode);
int32_t syncNodePingSelf(SSyncNode* pSyncNode);
int32_t syncNodePingPeers(SSyncNode* pSyncNode);
int32_t syncNodePingAll(SSyncNode* pSyncNode);
// timer control --------------
int32_t syncNodeStartPingTimer(SSyncNode* pSyncNode);
int32_t syncNodeStopPingTimer(SSyncNode* pSyncNode);
int32_t syncNodeStartElectTimer(SSyncNode* pSyncNode, int32_t ms);
int32_t syncNodeStopElectTimer(SSyncNode* pSyncNode);
int32_t syncNodeRestartElectTimer(SSyncNode* pSyncNode, int32_t ms);
int32_t syncNodeResetElectTimer(SSyncNode* pSyncNode);
int32_t syncNodeStartHeartbeatTimer(SSyncNode* pSyncNode);
int32_t syncNodeStopHeartbeatTimer(SSyncNode* pSyncNode);
// utils --------------
int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg);
int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg);
cJSON* syncNode2Json(const SSyncNode* pSyncNode);
char* syncNode2Str(const SSyncNode* pSyncNode);
// raft state change --------------
void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term);
void syncNodeBecomeFollower(SSyncNode* pSyncNode);
void syncNodeBecomeLeader(SSyncNode* pSyncNode);
void syncNodeCandidate2Leader(SSyncNode* pSyncNode);
void syncNodeFollower2Candidate(SSyncNode* pSyncNode);
void syncNodeLeader2Follower(SSyncNode* pSyncNode);
void syncNodeCandidate2Follower(SSyncNode* pSyncNode);
// raft vote --------------
void syncNodeVoteForTerm(SSyncNode* pSyncNode, SyncTerm term, SRaftId* pRaftId);
void syncNodeVoteForSelf(SSyncNode* pSyncNode);
void syncNodeMaybeAdvanceCommitIndex(SSyncNode* pSyncNode);
// for debug --------------
void syncNodePrint(SSyncNode* pObj);
void syncNodePrint2(char* s, SSyncNode* pObj);

View File

@ -39,6 +39,7 @@ typedef enum ESyncMessageType {
SYNC_REQUEST_VOTE_REPLY = 111,
SYNC_APPEND_ENTRIES = 113,
SYNC_APPEND_ENTRIES_REPLY = 115,
SYNC_RESPONSE = 119,
} ESyncMessageType;
@ -195,7 +196,7 @@ typedef struct SyncRequestVote {
SRaftId srcId;
SRaftId destId;
// private data
SyncTerm currentTerm;
SyncTerm term;
SyncIndex lastLogIndex;
SyncTerm lastLogTerm;
} SyncRequestVote;
@ -254,6 +255,7 @@ typedef struct SyncAppendEntries {
SRaftId srcId;
SRaftId destId;
// private data
SyncTerm term;
SyncIndex prevLogIndex;
SyncTerm prevLogTerm;
SyncIndex commitIndex;
@ -286,6 +288,7 @@ typedef struct SyncAppendEntriesReply {
SRaftId srcId;
SRaftId destId;
// private data
SyncTerm term;
bool success;
SyncIndex matchIndex;
} SyncAppendEntriesReply;

View File

@ -27,6 +27,9 @@ extern "C" {
#include "syncRaftEntry.h"
#include "taosdef.h"
#define SYNC_INDEX_BEGIN 0
#define SYNC_INDEX_INVALID -1
typedef struct SSyncLogStoreData {
SSyncNode* pSyncNode;
SWal* pWal;

View File

@ -43,6 +43,12 @@ int32_t raftStorePersist(SRaftStore *pRaftStore);
int32_t raftStoreSerialize(SRaftStore *pRaftStore, char *buf, size_t len);
int32_t raftStoreDeserialize(SRaftStore *pRaftStore, char *buf, size_t len);
bool raftStoreHasVoted(SRaftStore *pRaftStore);
void raftStoreVote(SRaftStore *pRaftStore, SRaftId *pRaftId);
void raftStoreClearVote(SRaftStore *pRaftStore);
void raftStoreNextTerm(SRaftStore *pRaftStore);
void raftStoreSetTerm(SRaftStore *pRaftStore, SyncTerm term);
// for debug -------------------
void raftStorePrint(SRaftStore *pObj);
void raftStorePrint2(char *s, SRaftStore *pObj);

View File

@ -52,7 +52,6 @@ extern "C" {
// /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode);
int32_t syncNodeReplicate(SSyncNode* pSyncNode);
int32_t syncNodeAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncAppendEntries* pMsg);

View File

@ -34,6 +34,7 @@ void syncUtilnodeInfo2EpSet(const SNodeInfo* pNodeInfo, SEpSet* pEpSet);
void syncUtilraftId2EpSet(const SRaftId* raftId, SEpSet* pEpSet);
void syncUtilnodeInfo2raftId(const SNodeInfo* pNodeInfo, SyncGroupId vgId, SRaftId* raftId);
bool syncUtilSameId(const SRaftId* pId1, const SRaftId* pId2);
bool syncUtilEmptyId(const SRaftId* pId);
// ---- SSyncBuffer ----
void syncUtilbufBuild(SSyncBuffer* syncBuf, size_t len);
@ -52,6 +53,8 @@ const char* syncUtilState2String(ESyncState state);
bool syncUtilCanPrint(char c);
char* syncUtilprintBin(char* ptr, uint32_t len);
char* syncUtilprintBin2(char* ptr, uint32_t len);
SyncIndex syncUtilMinIndex(SyncIndex a, SyncIndex b);
SyncIndex syncUtilMaxIndex(SyncIndex a, SyncIndex b);
#ifdef __cplusplus
}

View File

@ -14,6 +14,11 @@
*/
#include "syncAppendEntries.h"
#include "syncInt.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
// TLA+ Spec
// HandleAppendEntriesRequest(i, j, m) ==
@ -80,4 +85,121 @@
// /\ UNCHANGED <<serverVars, commitIndex, messages>>
// /\ UNCHANGED <<candidateVars, leaderVars>>
//
int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {}
int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
int32_t ret = 0;
syncAppendEntriesLog2("==syncNodeOnAppendEntriesCb==", pMsg);
if (pMsg->term > ths->pRaftStore->currentTerm) {
syncNodeUpdateTerm(ths, pMsg->term);
}
assert(pMsg->term <= ths->pRaftStore->currentTerm);
if (pMsg->term == ths->pRaftStore->currentTerm) {
ths->leaderCache = pMsg->srcId;
syncNodeResetElectTimer(ths);
}
assert(pMsg->dataLen >= 0);
SyncTerm localPreLogTerm = 0;
if (pMsg->prevLogTerm >= SYNC_INDEX_BEGIN && pMsg->prevLogTerm <= ths->pLogStore->getLastIndex(ths->pLogStore)) {
SSyncRaftEntry* pEntry = logStoreGetEntry(ths->pLogStore, pMsg->prevLogTerm);
assert(pEntry != NULL);
localPreLogTerm = pEntry->term;
syncEntryDestory(pEntry);
}
bool logOK =
(pMsg->prevLogIndex == SYNC_INDEX_INVALID) ||
((pMsg->prevLogIndex >= SYNC_INDEX_BEGIN) &&
(pMsg->prevLogIndex <= ths->pLogStore->getLastIndex(ths->pLogStore)) && (pMsg->prevLogIndex == localPreLogTerm));
// reject
if ((pMsg->term < ths->pRaftStore->currentTerm) ||
((pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) && !logOK)) {
SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->success = false;
pReply->matchIndex = SYNC_INDEX_INVALID;
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncAppendEntriesReplyDestroy(pReply);
return ret;
}
// return to follower state
if (pMsg->term == ths->pRaftStore->currentTerm && ths->state == TAOS_SYNC_STATE_CANDIDATE) {
syncNodeBecomeFollower(ths);
}
// accept request
if (pMsg->term == ths->pRaftStore->currentTerm && ths->state == TAOS_SYNC_STATE_FOLLOWER && logOK) {
bool matchSuccess = false;
if (pMsg->prevLogIndex == SYNC_INDEX_INVALID &&
ths->pLogStore->getLastIndex(ths->pLogStore) == SYNC_INDEX_INVALID) {
matchSuccess = true;
}
if (pMsg->prevLogIndex >= SYNC_INDEX_BEGIN && pMsg->prevLogIndex <= ths->pLogStore->getLastIndex(ths->pLogStore)) {
SSyncRaftEntry* pEntry = logStoreGetEntry(ths->pLogStore, pMsg->prevLogTerm);
assert(pEntry != NULL);
if (pMsg->prevLogTerm == pEntry->term) {
matchSuccess = true;
}
syncEntryDestory(pEntry);
}
if (matchSuccess) {
// delete conflict entries
if (ths->pLogStore->getLastIndex(ths->pLogStore) > pMsg->prevLogIndex) {
SyncIndex fromIndex = pMsg->prevLogIndex + 1;
ths->pLogStore->truncate(ths->pLogStore, fromIndex);
}
// append one entry
if (pMsg->dataLen > 0) {
SSyncRaftEntry* pEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
syncEntryDestory(pEntry);
}
SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->success = true;
pReply->matchIndex = pMsg->prevLogIndex + 1;
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncAppendEntriesReplyDestroy(pReply);
} else {
SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->success = false;
pReply->matchIndex = SYNC_INDEX_INVALID;
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncAppendEntriesReplyDestroy(pReply);
}
if (pMsg->commitIndex > ths->commitIndex) {
if (pMsg->commitIndex <= ths->pLogStore->getLastIndex(ths->pLogStore)) {
// commit
ths->commitIndex = pMsg->commitIndex;
ths->pLogStore->updateCommitIndex(ths->pLogStore, ths->commitIndex);
}
}
}
return ret;
}

View File

@ -14,6 +14,12 @@
*/
#include "syncAppendEntriesReply.h"
#include "syncIndexMgr.h"
#include "syncInt.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
// TLA+ Spec
// HandleAppendEntriesResponse(i, j, m) ==
@ -28,4 +34,41 @@
// /\ Discard(m)
// /\ UNCHANGED <<serverVars, candidateVars, logVars, elections>>
//
int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg) {}
int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg) {
int32_t ret = 0;
syncAppendEntriesReplyLog2("==syncNodeOnAppendEntriesReplyCb==", pMsg);
if (pMsg->term < ths->pRaftStore->currentTerm) {
sTrace("DropStaleResponse, receive term:%lu, current term:%lu", pMsg->term, ths->pRaftStore->currentTerm);
return ret;
}
// no need this code, because if I receive reply.term, then I must have sent for that term.
// if (pMsg->term > ths->pRaftStore->currentTerm) {
// syncNodeUpdateTerm(ths, pMsg->term);
// }
assert(pMsg->term == ths->pRaftStore->currentTerm);
if (pMsg->success) {
// nextIndex = reply.matchIndex + 1
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), pMsg->matchIndex + 1);
// matchIndex = reply.matchIndex
syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), pMsg->matchIndex);
// maybe commit
syncNodeMaybeAdvanceCommitIndex(ths);
} else {
SyncIndex nextIndex = syncIndexMgrGetIndex(ths->pNextIndex, &(pMsg->srcId));
if (nextIndex > SYNC_INDEX_BEGIN) {
--nextIndex;
} else {
nextIndex = SYNC_INDEX_BEGIN;
}
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), nextIndex);
}
return ret;
}

View File

@ -16,6 +16,7 @@
#include "syncElection.h"
#include "syncMessage.h"
#include "syncRaftStore.h"
#include "syncVoteMgr.h"
// TLA+ Spec
// RequestVote(i, j) ==
@ -37,7 +38,7 @@ int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode) {
SyncRequestVote* pMsg = syncRequestVoteBuild();
pMsg->srcId = pSyncNode->myRaftId;
pMsg->destId = pSyncNode->peersId[i];
pMsg->currentTerm = pSyncNode->pRaftStore->currentTerm;
pMsg->term = pSyncNode->pRaftStore->currentTerm;
pMsg->lastLogIndex = pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore);
pMsg->lastLogTerm = pSyncNode->pLogStore->getLastTerm(pSyncNode->pLogStore);
@ -49,10 +50,22 @@ int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode) {
}
int32_t syncNodeElect(SSyncNode* pSyncNode) {
if (pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER) {
syncNodeFollower2Candidate(pSyncNode);
}
assert(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
// start election
raftStoreNextTerm(pSyncNode->pRaftStore);
raftStoreClearVote(pSyncNode->pRaftStore);
voteGrantedReset(pSyncNode->pVotesGranted, pSyncNode->pRaftStore->currentTerm);
votesRespondReset(pSyncNode->pVotesRespond, pSyncNode->pRaftStore->currentTerm);
syncNodeVoteForSelf(pSyncNode);
int32_t ret = syncNodeRequestVotePeers(pSyncNode);
assert(ret == 0);
syncNodeResetElectTimer(pSyncNode);
return ret;
}

View File

@ -19,19 +19,18 @@
SSyncEnv *gSyncEnv = NULL;
// local function -----------------
static void syncEnvTick(void *param, void *tmrId);
static int32_t doSyncEnvStart(SSyncEnv *pSyncEnv);
static int32_t doSyncEnvStop(SSyncEnv *pSyncEnv);
static tmr_h doSyncEnvStartTimer(SSyncEnv *pSyncEnv, TAOS_TMR_CALLBACK fp, int mseconds, void *param);
static void doSyncEnvStopTimer(SSyncEnv *pSyncEnv, tmr_h *pTimer);
static SSyncEnv *doSyncEnvStart();
static int32_t doSyncEnvStop(SSyncEnv *pSyncEnv);
static int32_t doSyncEnvStartTimer(SSyncEnv *pSyncEnv);
static int32_t doSyncEnvStopTimer(SSyncEnv *pSyncEnv);
static void syncEnvTick(void *param, void *tmrId);
// --------------------------------
int32_t syncEnvStart() {
int32_t ret;
int32_t ret = 0;
taosSeedRand(taosGetTimestampSec());
gSyncEnv = (SSyncEnv *)malloc(sizeof(SSyncEnv));
gSyncEnv = doSyncEnvStart(gSyncEnv);
assert(gSyncEnv != NULL);
ret = doSyncEnvStart(gSyncEnv);
return ret;
}
@ -40,31 +39,52 @@ int32_t syncEnvStop() {
return ret;
}
tmr_h syncEnvStartTimer(TAOS_TMR_CALLBACK fp, int mseconds, void *param) {
return doSyncEnvStartTimer(gSyncEnv, fp, mseconds, param);
int32_t syncEnvStartTimer() {
int32_t ret = doSyncEnvStartTimer(gSyncEnv);
return ret;
}
void syncEnvStopTimer(tmr_h *pTimer) { doSyncEnvStopTimer(gSyncEnv, pTimer); }
int32_t syncEnvStopTimer() {
int32_t ret = doSyncEnvStopTimer(gSyncEnv);
return ret;
}
// local function -----------------
static void syncEnvTick(void *param, void *tmrId) {
SSyncEnv *pSyncEnv = (SSyncEnv *)param;
sTrace("syncEnvTick ... name:%s ", pSyncEnv->name);
if (atomic_load_64(&pSyncEnv->envTickTimerLogicClockUser) <= atomic_load_64(&pSyncEnv->envTickTimerLogicClock)) {
++(pSyncEnv->envTickTimerCounter);
sTrace(
"syncEnvTick do ... envTickTimerLogicClockUser:%lu, envTickTimerLogicClock:%lu, envTickTimerCounter:%lu, "
"envTickTimerMS:%d, tmrId:%p",
pSyncEnv->envTickTimerLogicClockUser, pSyncEnv->envTickTimerLogicClock, pSyncEnv->envTickTimerCounter,
pSyncEnv->envTickTimerMS, tmrId);
pSyncEnv->pEnvTickTimer = taosTmrStart(syncEnvTick, 1000, pSyncEnv, pSyncEnv->pTimerManager);
// do something, tick ...
taosTmrReset(syncEnvTick, pSyncEnv->envTickTimerMS, pSyncEnv, pSyncEnv->pTimerManager, &pSyncEnv->pEnvTickTimer);
} else {
sTrace(
"syncEnvTick pass ... envTickTimerLogicClockUser:%lu, envTickTimerLogicClock:%lu, envTickTimerCounter:%lu, "
"envTickTimerMS:%d, tmrId:%p",
pSyncEnv->envTickTimerLogicClockUser, pSyncEnv->envTickTimerLogicClock, pSyncEnv->envTickTimerCounter,
pSyncEnv->envTickTimerMS, tmrId);
}
}
static int32_t doSyncEnvStart(SSyncEnv *pSyncEnv) {
snprintf(pSyncEnv->name, sizeof(pSyncEnv->name), "SyncEnv_%p", pSyncEnv);
static SSyncEnv *doSyncEnvStart() {
SSyncEnv *pSyncEnv = (SSyncEnv *)malloc(sizeof(SSyncEnv));
assert(pSyncEnv != NULL);
memset(pSyncEnv, 0, sizeof(pSyncEnv));
pSyncEnv->envTickTimerCounter = 0;
pSyncEnv->envTickTimerMS = ENV_TICK_TIMER_MS;
pSyncEnv->FpEnvTickTimer = syncEnvTick;
atomic_store_64(&pSyncEnv->envTickTimerLogicClock, 0);
atomic_store_64(&pSyncEnv->envTickTimerLogicClockUser, 0);
// start tmr thread
pSyncEnv->pTimerManager = taosTmrInit(1000, 50, 10000, "SYNC-ENV");
// pSyncEnv->pEnvTickTimer = taosTmrStart(syncEnvTick, 1000, pSyncEnv, pSyncEnv->pTimerManager);
sTrace("SyncEnv start ok, name:%s", pSyncEnv->name);
return 0;
return pSyncEnv;
}
static int32_t doSyncEnvStop(SSyncEnv *pSyncEnv) {
@ -72,8 +92,18 @@ static int32_t doSyncEnvStop(SSyncEnv *pSyncEnv) {
return 0;
}
static tmr_h doSyncEnvStartTimer(SSyncEnv *pSyncEnv, TAOS_TMR_CALLBACK fp, int mseconds, void *param) {
return taosTmrStart(fp, mseconds, pSyncEnv, pSyncEnv->pTimerManager);
static int32_t doSyncEnvStartTimer(SSyncEnv *pSyncEnv) {
int32_t ret = 0;
taosTmrReset(pSyncEnv->FpEnvTickTimer, pSyncEnv->envTickTimerMS, pSyncEnv, pSyncEnv->pTimerManager,
&pSyncEnv->pEnvTickTimer);
atomic_store_64(&pSyncEnv->envTickTimerLogicClock, pSyncEnv->envTickTimerLogicClockUser);
return ret;
}
static void doSyncEnvStopTimer(SSyncEnv *pSyncEnv, tmr_h *pTimer) {}
static int32_t doSyncEnvStopTimer(SSyncEnv *pSyncEnv) {
int32_t ret = 0;
atomic_add_fetch_64(&pSyncEnv->envTickTimerLogicClockUser, 1);
taosTmrStop(pSyncEnv->pEnvTickTimer);
pSyncEnv->pEnvTickTimer = NULL;
return ret;
}

View File

@ -16,6 +16,7 @@
#include "syncIO.h"
#include <tdatablock.h>
#include "syncMessage.h"
#include "syncUtil.h"
#include "tglobal.h"
#include "ttimer.h"
#include "tutil.h"
@ -23,33 +24,36 @@
SSyncIO *gSyncIO = NULL;
// local function ------------
static int32_t syncIOStartInternal(SSyncIO *io);
static int32_t syncIOStopInternal(SSyncIO *io);
static SSyncIO *syncIOCreate(char *host, uint16_t port);
static int32_t syncIODestroy(SSyncIO *io);
static int32_t syncIOStartInternal(SSyncIO *io);
static int32_t syncIOStopInternal(SSyncIO *io);
static void *syncIOConsumerFunc(void *param);
static int syncIOAuth(void *parent, char *meterId, char *spi, char *encrypt, char *secret, char *ckey);
static void syncIOProcessRequest(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet);
static void syncIOProcessReply(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet);
static void *syncIOConsumerFunc(void *param);
static void syncIOProcessRequest(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet);
static void syncIOProcessReply(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet);
static int32_t syncIOAuth(void *parent, char *meterId, char *spi, char *encrypt, char *secret, char *ckey);
static int32_t syncIOTickQInternal(SSyncIO *io);
static void syncIOTickQFunc(void *param, void *tmrId);
static int32_t syncIOTickPingInternal(SSyncIO *io);
static void syncIOTickPingFunc(void *param, void *tmrId);
static int32_t syncIOStartQ(SSyncIO *io);
static int32_t syncIOStopQ(SSyncIO *io);
static int32_t syncIOStartPing(SSyncIO *io);
static int32_t syncIOStopPing(SSyncIO *io);
static void syncIOTickQ(void *param, void *tmrId);
static void syncIOTickPing(void *param, void *tmrId);
// ----------------------------
// public function ------------
int32_t syncIOStart(char *host, uint16_t port) {
int32_t ret = 0;
gSyncIO = syncIOCreate(host, port);
assert(gSyncIO != NULL);
taosSeedRand(taosGetTimestampSec());
int32_t ret = syncIOStartInternal(gSyncIO);
ret = syncIOStartInternal(gSyncIO);
assert(ret == 0);
sTrace("syncIOStart ok, gSyncIO:%p gSyncIO->clientRpc:%p", gSyncIO, gSyncIO->clientRpc);
return 0;
sTrace("syncIOStart ok, gSyncIO:%p", gSyncIO);
return ret;
}
int32_t syncIOStop() {
@ -61,37 +65,25 @@ int32_t syncIOStop() {
return ret;
}
int32_t syncIOTickQ() {
int32_t ret = syncIOTickQInternal(gSyncIO);
assert(ret == 0);
return ret;
}
int32_t syncIOTickPing() {
int32_t ret = syncIOTickPingInternal(gSyncIO);
assert(ret == 0);
return ret;
}
int32_t syncIOSendMsg(void *clientRpc, const SEpSet *pEpSet, SRpcMsg *pMsg) {
sTrace(
"<--- syncIOSendMsg ---> clientRpc:%p, numOfEps:%d, inUse:%d, destAddr:%s-%u, pMsg->ahandle:%p, pMsg->handle:%p, "
"pMsg->msgType:%d, pMsg->contLen:%d",
clientRpc, pEpSet->numOfEps, pEpSet->inUse, pEpSet->eps[0].fqdn, pEpSet->eps[0].port, pMsg->ahandle, pMsg->handle,
pMsg->msgType, pMsg->contLen);
{
cJSON *pJson = syncRpcMsg2Json(pMsg);
char * serialized = cJSON_Print(pJson);
sTrace("process syncMessage send: pMsg:%s ", serialized);
free(serialized);
cJSON_Delete(pJson);
}
assert(pEpSet->inUse == 0);
assert(pEpSet->numOfEps == 1);
int32_t ret = 0;
char logBuf[256];
snprintf(logBuf, sizeof(logBuf), "==syncIOSendMsg== %s:%d", pEpSet->eps[0].fqdn, pEpSet->eps[0].port);
syncRpcMsgPrint2(logBuf, pMsg);
pMsg->handle = NULL;
rpcSendRequest(clientRpc, pEpSet, pMsg, NULL);
return 0;
return ret;
}
int32_t syncIOEqMsg(void *queue, SRpcMsg *pMsg) {
int32_t ret = 0;
char logBuf[128];
syncRpcMsgPrint2((char *)"==syncIOEqMsg==", pMsg);
SRpcMsg *pTemp;
pTemp = taosAllocateQitem(sizeof(SRpcMsg));
memcpy(pTemp, pMsg, sizeof(SRpcMsg));
@ -99,11 +91,75 @@ int32_t syncIOEqMsg(void *queue, SRpcMsg *pMsg) {
STaosQueue *pMsgQ = queue;
taosWriteQitem(pMsgQ, pTemp);
return 0;
return ret;
}
int32_t syncIOQTimerStart() {
int32_t ret = syncIOStartQ(gSyncIO);
assert(ret == 0);
return ret;
}
int32_t syncIOQTimerStop() {
int32_t ret = syncIOStopQ(gSyncIO);
assert(ret == 0);
return ret;
}
int32_t syncIOPingTimerStart() {
int32_t ret = syncIOStartPing(gSyncIO);
assert(ret == 0);
return ret;
}
int32_t syncIOPingTimerStop() {
int32_t ret = syncIOStopPing(gSyncIO);
assert(ret == 0);
return ret;
}
// local function ------------
static SSyncIO *syncIOCreate(char *host, uint16_t port) {
SSyncIO *io = (SSyncIO *)malloc(sizeof(SSyncIO));
memset(io, 0, sizeof(*io));
io->pMsgQ = taosOpenQueue();
io->pQset = taosOpenQset();
taosAddIntoQset(io->pQset, io->pMsgQ, NULL);
io->myAddr.inUse = 0;
io->myAddr.numOfEps = 0;
addEpIntoEpSet(&io->myAddr, host, port);
io->qTimerMS = TICK_Q_TIMER_MS;
io->pingTimerMS = TICK_Ping_TIMER_MS;
return io;
}
static int32_t syncIODestroy(SSyncIO *io) {
int32_t ret = 0;
int8_t start = atomic_load_8(&io->isStart);
assert(start == 0);
if (io->serverRpc != NULL) {
rpcClose(io->serverRpc);
io->serverRpc = NULL;
}
if (io->clientRpc != NULL) {
rpcClose(io->clientRpc);
io->clientRpc = NULL;
}
taosCloseQueue(io->pMsgQ);
taosCloseQset(io->pQset);
return ret;
}
static int32_t syncIOStartInternal(SSyncIO *io) {
int32_t ret = 0;
taosBlockSIGPIPE();
rpcInit();
@ -163,58 +219,24 @@ static int32_t syncIOStartInternal(SSyncIO *io) {
}
// start tmr thread
io->ioTimerManager = taosTmrInit(1000, 50, 10000, "SYNC");
io->timerMgr = taosTmrInit(1000, 50, 10000, "SYNC-IO");
return 0;
atomic_store_8(&io->isStart, 1);
return ret;
}
static int32_t syncIOStopInternal(SSyncIO *io) {
int32_t ret = 0;
atomic_store_8(&io->isStart, 0);
pthread_join(io->consumerTid, NULL);
return 0;
}
static SSyncIO *syncIOCreate(char *host, uint16_t port) {
SSyncIO *io = (SSyncIO *)malloc(sizeof(SSyncIO));
memset(io, 0, sizeof(*io));
io->pMsgQ = taosOpenQueue();
io->pQset = taosOpenQset();
taosAddIntoQset(io->pQset, io->pMsgQ, NULL);
io->myAddr.inUse = 0;
addEpIntoEpSet(&io->myAddr, host, port);
return io;
}
static int32_t syncIODestroy(SSyncIO *io) {
int8_t start = atomic_load_8(&io->isStart);
assert(start == 0);
if (io->serverRpc != NULL) {
free(io->serverRpc);
io->serverRpc = NULL;
}
if (io->clientRpc != NULL) {
free(io->clientRpc);
io->clientRpc = NULL;
}
taosCloseQueue(io->pMsgQ);
taosCloseQset(io->pQset);
return 0;
taosTmrCleanUp(io->timerMgr);
return ret;
}
static void *syncIOConsumerFunc(void *param) {
SSyncIO *io = param;
SSyncIO *io = param;
STaosQall *qall;
SRpcMsg * pRpcMsg, rpcMsg;
int type;
SRpcMsg *pRpcMsg, rpcMsg;
qall = taosAllocateQall();
while (1) {
@ -226,77 +248,74 @@ static void *syncIOConsumerFunc(void *param) {
for (int i = 0; i < numOfMsgs; ++i) {
taosGetQitem(qall, (void **)&pRpcMsg);
syncRpcMsgLog2((char *)"==syncIOConsumerFunc==", pRpcMsg);
char *s = syncRpcMsg2Str(pRpcMsg);
sTrace("syncIOConsumerFunc get item from queue: msgType:%d contLen:%d msg:%s", pRpcMsg->msgType, pRpcMsg->contLen,
s);
free(s);
// use switch case instead of if else
if (pRpcMsg->msgType == SYNC_PING) {
if (io->FpOnSyncPing != NULL) {
SyncPing *pSyncMsg;
SyncPing *pSyncMsg = syncPingFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
io->FpOnSyncPing(io->pSyncNode, pSyncMsg);
syncPingDestroy(pSyncMsg);
/*
pSyncMsg = syncPingBuild(pRpcMsg->contLen);
syncPingFromRpcMsg(pRpcMsg, pSyncMsg);
// memcpy(pSyncMsg, tmpRpcMsg.pCont, tmpRpcMsg.contLen);
io->FpOnSyncPing(io->pSyncNode, pSyncMsg);
syncPingDestroy(pSyncMsg);
*/
}
} else if (pRpcMsg->msgType == SYNC_PING_REPLY) {
if (io->FpOnSyncPingReply != NULL) {
SyncPingReply *pSyncMsg;
pSyncMsg = syncPingReplyBuild(pRpcMsg->contLen);
syncPingReplyFromRpcMsg(pRpcMsg, pSyncMsg);
SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pRpcMsg);
io->FpOnSyncPingReply(io->pSyncNode, pSyncMsg);
syncPingReplyDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == SYNC_CLIENT_REQUEST) {
if (io->FpOnSyncClientRequest != NULL) {
SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pRpcMsg);
io->FpOnSyncClientRequest(io->pSyncNode, pSyncMsg);
syncClientRequestDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == SYNC_REQUEST_VOTE) {
if (io->FpOnSyncRequestVote != NULL) {
SyncRequestVote *pSyncMsg;
pSyncMsg = syncRequestVoteBuild(pRpcMsg->contLen);
syncRequestVoteFromRpcMsg(pRpcMsg, pSyncMsg);
SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pRpcMsg);
io->FpOnSyncRequestVote(io->pSyncNode, pSyncMsg);
syncRequestVoteDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == SYNC_REQUEST_VOTE_REPLY) {
if (io->FpOnSyncRequestVoteReply != NULL) {
SyncRequestVoteReply *pSyncMsg;
pSyncMsg = syncRequestVoteReplyBuild();
syncRequestVoteReplyFromRpcMsg(pRpcMsg, pSyncMsg);
SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pRpcMsg);
io->FpOnSyncRequestVoteReply(io->pSyncNode, pSyncMsg);
syncRequestVoteReplyDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == SYNC_APPEND_ENTRIES) {
if (io->FpOnSyncAppendEntries != NULL) {
SyncAppendEntries *pSyncMsg;
pSyncMsg = syncAppendEntriesBuild(pRpcMsg->contLen);
syncAppendEntriesFromRpcMsg(pRpcMsg, pSyncMsg);
SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pRpcMsg);
io->FpOnSyncAppendEntries(io->pSyncNode, pSyncMsg);
syncAppendEntriesDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == SYNC_APPEND_ENTRIES_REPLY) {
if (io->FpOnSyncAppendEntriesReply != NULL) {
SyncAppendEntriesReply *pSyncMsg;
pSyncMsg = syncAppendEntriesReplyBuild();
syncAppendEntriesReplyFromRpcMsg(pRpcMsg, pSyncMsg);
SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pRpcMsg);
io->FpOnSyncAppendEntriesReply(io->pSyncNode, pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == SYNC_TIMEOUT) {
if (io->FpOnSyncTimeout != NULL) {
SyncTimeout *pSyncMsg;
pSyncMsg = syncTimeoutBuild();
syncTimeoutFromRpcMsg(pRpcMsg, pSyncMsg);
SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pRpcMsg);
io->FpOnSyncTimeout(io->pSyncNode, pSyncMsg);
syncTimeoutDestroy(pSyncMsg);
}
} else {
;
sTrace("unknown msgType:%d, no operator", pRpcMsg->msgType);
}
}
@ -306,15 +325,16 @@ static void *syncIOConsumerFunc(void *param) {
rpcFreeCont(pRpcMsg->pCont);
if (pRpcMsg->handle != NULL) {
int msgSize = 128;
int msgSize = 32;
memset(&rpcMsg, 0, sizeof(rpcMsg));
rpcMsg.msgType = SYNC_RESPONSE;
rpcMsg.pCont = rpcMallocCont(msgSize);
rpcMsg.contLen = msgSize;
snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", "give a reply");
rpcMsg.handle = pRpcMsg->handle;
rpcMsg.code = 0;
sTrace("syncIOConsumerFunc rpcSendResponse ... msgType:%d contLen:%d", pRpcMsg->msgType, rpcMsg.contLen);
syncRpcMsgPrint2((char *)"syncIOConsumerFunc rpcSendResponse --> ", &rpcMsg);
rpcSendResponse(&rpcMsg);
}
@ -326,71 +346,95 @@ static void *syncIOConsumerFunc(void *param) {
return NULL;
}
static int syncIOAuth(void *parent, char *meterId, char *spi, char *encrypt, char *secret, char *ckey) {
// app shall retrieve the auth info based on meterID from DB or a data file
// demo code here only for simple demo
int ret = 0;
return ret;
}
static void syncIOProcessRequest(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet) {
sTrace("<-- syncIOProcessRequest --> type:%d, contLen:%d, cont:%s", pMsg->msgType, pMsg->contLen,
(char *)pMsg->pCont);
syncRpcMsgPrint2((char *)"==syncIOProcessRequest==", pMsg);
SSyncIO *io = pParent;
SRpcMsg *pTemp;
pTemp = taosAllocateQitem(sizeof(SRpcMsg));
memcpy(pTemp, pMsg, sizeof(SRpcMsg));
taosWriteQitem(io->pMsgQ, pTemp);
}
static void syncIOProcessReply(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet) {
sTrace("syncIOProcessReply: type:%d, contLen:%d msg:%s", pMsg->msgType, pMsg->contLen, (char *)pMsg->pCont);
if (pMsg->msgType == SYNC_RESPONSE) {
sTrace("==syncIOProcessReply==");
} else {
syncRpcMsgPrint2((char *)"==syncIOProcessReply==", pMsg);
}
rpcFreeCont(pMsg->pCont);
}
static int32_t syncIOTickQInternal(SSyncIO *io) {
io->ioTimerTickQ = taosTmrStart(syncIOTickQFunc, 1000, io, io->ioTimerManager);
return 0;
static int32_t syncIOAuth(void *parent, char *meterId, char *spi, char *encrypt, char *secret, char *ckey) {
// app shall retrieve the auth info based on meterID from DB or a data file
// demo code here only for simple demo
int32_t ret = 0;
return ret;
}
static void syncIOTickQFunc(void *param, void *tmrId) {
static int32_t syncIOStartQ(SSyncIO *io) {
int32_t ret = 0;
taosTmrReset(syncIOTickQ, io->qTimerMS, io, io->timerMgr, &io->qTimer);
return ret;
}
static int32_t syncIOStopQ(SSyncIO *io) {
int32_t ret = 0;
taosTmrStop(io->qTimer);
io->qTimer = NULL;
return ret;
}
static int32_t syncIOStartPing(SSyncIO *io) {
int32_t ret = 0;
taosTmrReset(syncIOTickPing, io->pingTimerMS, io, io->timerMgr, &io->pingTimer);
return ret;
}
static int32_t syncIOStopPing(SSyncIO *io) {
int32_t ret = 0;
taosTmrStop(io->pingTimer);
io->pingTimer = NULL;
return ret;
}
static void syncIOTickQ(void *param, void *tmrId) {
SSyncIO *io = (SSyncIO *)param;
sTrace("<-- syncIOTickQFunc -->");
SRaftId srcId, destId;
srcId.addr = syncUtilAddr2U64(io->myAddr.eps[0].fqdn, io->myAddr.eps[0].port);
srcId.vgId = -1;
destId.addr = syncUtilAddr2U64(io->myAddr.eps[0].fqdn, io->myAddr.eps[0].port);
destId.vgId = -1;
SyncPingReply *pMsg = syncPingReplyBuild2(&srcId, &destId, "syncIOTickQ");
SRpcMsg rpcMsg;
rpcMsg.contLen = 64;
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", "syncIOTickQ");
rpcMsg.handle = NULL;
rpcMsg.msgType = 55;
syncPingReply2RpcMsg(pMsg, &rpcMsg);
SRpcMsg *pTemp;
pTemp = taosAllocateQitem(sizeof(SRpcMsg));
memcpy(pTemp, &rpcMsg, sizeof(SRpcMsg));
syncRpcMsgPrint2((char *)"==syncIOTickQ==", &rpcMsg);
taosWriteQitem(io->pMsgQ, pTemp);
taosTmrReset(syncIOTickQFunc, 1000, io, io->ioTimerManager, &io->ioTimerTickQ);
syncPingReplyDestroy(pMsg);
taosTmrReset(syncIOTickQ, io->qTimerMS, io, io->timerMgr, &io->qTimer);
}
static int32_t syncIOTickPingInternal(SSyncIO *io) {
io->ioTimerTickPing = taosTmrStart(syncIOTickPingFunc, 1000, io, io->ioTimerManager);
return 0;
}
static void syncIOTickPingFunc(void *param, void *tmrId) {
static void syncIOTickPing(void *param, void *tmrId) {
SSyncIO *io = (SSyncIO *)param;
sTrace("<-- syncIOTickPingFunc -->");
SRaftId srcId, destId;
srcId.addr = syncUtilAddr2U64(io->myAddr.eps[0].fqdn, io->myAddr.eps[0].port);
srcId.vgId = -1;
destId.addr = syncUtilAddr2U64(io->myAddr.eps[0].fqdn, io->myAddr.eps[0].port);
destId.vgId = -1;
SyncPing *pMsg = syncPingBuild2(&srcId, &destId, "syncIOTickPing");
// SyncPing *pMsg = syncPingBuild3(&srcId, &destId);
SRpcMsg rpcMsg;
rpcMsg.contLen = 64;
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", "syncIOTickPing");
rpcMsg.handle = NULL;
rpcMsg.msgType = 77;
syncPing2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"==syncIOTickPing==", &rpcMsg);
rpcSendRequest(io->clientRpc, &io->myAddr, &rpcMsg, NULL);
taosTmrReset(syncIOTickPingFunc, 1000, io, io->ioTimerManager, &io->ioTimerTickPing);
syncPingDestroy(pMsg);
taosTmrReset(syncIOTickPing, io->pingTimerMS, io, io->timerMgr, &io->pingTimer);
}

View File

@ -17,11 +17,14 @@
#include "sync.h"
#include "syncAppendEntries.h"
#include "syncAppendEntriesReply.h"
#include "syncElection.h"
#include "syncEnv.h"
#include "syncIndexMgr.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncReplication.h"
#include "syncRequestVote.h"
#include "syncRequestVoteReply.h"
#include "syncTimeout.h"
@ -31,33 +34,32 @@
static int32_t tsNodeRefId = -1;
// ------ local funciton ---------
// enqueue message ----
static void syncNodeEqPingTimer(void* param, void* tmrId);
static void syncNodeEqElectTimer(void* param, void* tmrId);
static void syncNodeEqHeartbeatTimer(void* param, void* tmrId);
// on message ----
static int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg);
static int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg);
static void UpdateTerm(SSyncNode* pSyncNode, SyncTerm term);
static void syncNodeBecomeFollower(SSyncNode* pSyncNode);
static void syncNodeBecomeLeader(SSyncNode* pSyncNode);
static void syncNodeFollower2Candidate(SSyncNode* pSyncNode);
static void syncNodeCandidate2Leader(SSyncNode* pSyncNode);
static void syncNodeLeader2Follower(SSyncNode* pSyncNode);
static void syncNodeCandidate2Follower(SSyncNode* pSyncNode);
static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg);
// ---------------------------------
int32_t syncInit() {
sTrace("syncInit ok");
return 0;
int32_t ret = syncEnvStart();
return ret;
}
void syncCleanUp() { sTrace("syncCleanUp ok"); }
void syncCleanUp() {
int32_t ret = syncEnvStop();
assert(ret == 0);
}
int64_t syncStart(const SSyncInfo* pSyncInfo) {
int32_t ret = 0;
SSyncNode* pSyncNode = syncNodeOpen(pSyncInfo);
assert(pSyncNode != NULL);
return 0;
return ret;
}
void syncStop(int64_t rid) {
@ -65,9 +67,13 @@ void syncStop(int64_t rid) {
syncNodeClose(pSyncNode);
}
int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) { return 0; }
int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) {
int32_t ret = 0;
return ret;
}
int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
int32_t ret = 0;
SSyncNode* pSyncNode = NULL; // get pointer from rid
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
SyncClientRequest* pSyncMsg = syncClientRequestBuild2(pMsg, 0, isWeak);
@ -75,11 +81,13 @@ int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
syncClientRequest2RpcMsg(pSyncMsg, &rpcMsg);
pSyncNode->FpEqMsg(pSyncNode->queue, &rpcMsg);
syncClientRequestDestroy(pSyncMsg);
ret = 0;
} else {
sTrace("syncForwardToPeer not leader, %s", syncUtilState2String(pSyncNode->state));
return -1; // need define err code !!
ret = -1; // need define err code !!
}
return 0;
return ret;
}
ESyncState syncGetMyRole(int64_t rid) {
@ -89,6 +97,7 @@ ESyncState syncGetMyRole(int64_t rid) {
void syncGetNodesRole(int64_t rid, SNodesRole* pNodeRole) {}
// open/close --------------
SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
SSyncNode* pSyncNode = (SSyncNode*)malloc(sizeof(SSyncNode));
assert(pSyncNode != NULL);
@ -162,7 +171,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
pSyncNode->pingTimerMS = PING_TIMER_MS;
atomic_store_64(&pSyncNode->pingTimerLogicClock, 0);
atomic_store_64(&pSyncNode->pingTimerLogicClockUser, 0);
pSyncNode->FpPingTimer = syncNodeEqPingTimer;
pSyncNode->FpPingTimerCB = syncNodeEqPingTimer;
pSyncNode->pingTimerCounter = 0;
// init elect timer
@ -170,7 +179,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
pSyncNode->electTimerMS = syncUtilElectRandomMS();
atomic_store_64(&pSyncNode->electTimerLogicClock, 0);
atomic_store_64(&pSyncNode->electTimerLogicClockUser, 0);
pSyncNode->FpElectTimer = syncNodeEqElectTimer;
pSyncNode->FpElectTimerCB = syncNodeEqElectTimer;
pSyncNode->electTimerCounter = 0;
// init heartbeat timer
@ -178,12 +187,13 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
pSyncNode->heartbeatTimerMS = HEARTBEAT_TIMER_MS;
atomic_store_64(&pSyncNode->heartbeatTimerLogicClock, 0);
atomic_store_64(&pSyncNode->heartbeatTimerLogicClockUser, 0);
pSyncNode->FpHeartbeatTimer = syncNodeEqHeartbeatTimer;
pSyncNode->FpHeartbeatTimerCB = syncNodeEqHeartbeatTimer;
pSyncNode->heartbeatTimerCounter = 0;
// init callback
pSyncNode->FpOnPing = syncNodeOnPingCb;
pSyncNode->FpOnPingReply = syncNodeOnPingReplyCb;
pSyncNode->FpOnClientRequest = syncNodeOnClientRequestCb;
pSyncNode->FpOnRequestVote = syncNodeOnRequestVoteCb;
pSyncNode->FpOnRequestVoteReply = syncNodeOnRequestVoteReplyCb;
pSyncNode->FpOnAppendEntries = syncNodeOnAppendEntriesCb;
@ -194,10 +204,153 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
}
void syncNodeClose(SSyncNode* pSyncNode) {
int32_t ret;
assert(pSyncNode != NULL);
ret = raftStoreClose(pSyncNode->pRaftStore);
assert(ret == 0);
voteGrantedDestroy(pSyncNode->pVotesGranted);
votesRespondDestory(pSyncNode->pVotesRespond);
syncIndexMgrDestroy(pSyncNode->pNextIndex);
syncIndexMgrDestroy(pSyncNode->pMatchIndex);
logStoreDestory(pSyncNode->pLogStore);
syncNodeStopPingTimer(pSyncNode);
syncNodeStopElectTimer(pSyncNode);
syncNodeStopHeartbeatTimer(pSyncNode);
free(pSyncNode);
}
// ping --------------
int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg) {
syncPingLog2((char*)"==syncNodePing==", pMsg);
int32_t ret = 0;
SRpcMsg rpcMsg;
syncPing2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgLog2((char*)"==syncNodePing==", &rpcMsg);
ret = syncNodeSendMsgById(destRaftId, pSyncNode, &rpcMsg);
return ret;
}
int32_t syncNodePingSelf(SSyncNode* pSyncNode) {
int32_t ret = 0;
SyncPing* pMsg = syncPingBuild3(&pSyncNode->myRaftId, &pSyncNode->myRaftId);
ret = syncNodePing(pSyncNode, &pMsg->destId, pMsg);
assert(ret == 0);
syncPingDestroy(pMsg);
return ret;
}
int32_t syncNodePingPeers(SSyncNode* pSyncNode) {
int32_t ret = 0;
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SRaftId destId;
syncUtilnodeInfo2raftId(&pSyncNode->peersNodeInfo[i], pSyncNode->vgId, &destId);
SyncPing* pMsg = syncPingBuild3(&pSyncNode->myRaftId, &destId);
ret = syncNodePing(pSyncNode, &destId, pMsg);
assert(ret == 0);
syncPingDestroy(pMsg);
}
return ret;
}
int32_t syncNodePingAll(SSyncNode* pSyncNode) {
int32_t ret = 0;
for (int i = 0; i < pSyncNode->syncCfg.replicaNum; ++i) {
SRaftId destId;
syncUtilnodeInfo2raftId(&pSyncNode->syncCfg.nodeInfo[i], pSyncNode->vgId, &destId);
SyncPing* pMsg = syncPingBuild3(&pSyncNode->myRaftId, &destId);
ret = syncNodePing(pSyncNode, &destId, pMsg);
assert(ret == 0);
syncPingDestroy(pMsg);
}
return ret;
}
// timer control --------------
int32_t syncNodeStartPingTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
taosTmrReset(pSyncNode->FpPingTimerCB, pSyncNode->pingTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pPingTimer);
atomic_store_64(&pSyncNode->pingTimerLogicClock, pSyncNode->pingTimerLogicClockUser);
return ret;
}
int32_t syncNodeStopPingTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
atomic_add_fetch_64(&pSyncNode->pingTimerLogicClockUser, 1);
taosTmrStop(pSyncNode->pPingTimer);
pSyncNode->pPingTimer = NULL;
return ret;
}
int32_t syncNodeStartElectTimer(SSyncNode* pSyncNode, int32_t ms) {
int32_t ret = 0;
pSyncNode->electTimerMS = ms;
taosTmrReset(pSyncNode->FpElectTimerCB, pSyncNode->electTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pElectTimer);
atomic_store_64(&pSyncNode->electTimerLogicClock, pSyncNode->electTimerLogicClockUser);
return ret;
}
int32_t syncNodeStopElectTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
atomic_add_fetch_64(&pSyncNode->electTimerLogicClockUser, 1);
taosTmrStop(pSyncNode->pElectTimer);
pSyncNode->pElectTimer = NULL;
return ret;
}
int32_t syncNodeRestartElectTimer(SSyncNode* pSyncNode, int32_t ms) {
int32_t ret = 0;
syncNodeStopElectTimer(pSyncNode);
syncNodeStartElectTimer(pSyncNode, ms);
return ret;
}
int32_t syncNodeResetElectTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
int32_t electMS = syncUtilElectRandomMS();
ret = syncNodeRestartElectTimer(pSyncNode, electMS);
return ret;
}
int32_t syncNodeStartHeartbeatTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
taosTmrReset(pSyncNode->FpHeartbeatTimerCB, pSyncNode->heartbeatTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pHeartbeatTimer);
atomic_store_64(&pSyncNode->heartbeatTimerLogicClock, pSyncNode->heartbeatTimerLogicClockUser);
return ret;
}
int32_t syncNodeStopHeartbeatTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
atomic_add_fetch_64(&pSyncNode->heartbeatTimerLogicClockUser, 1);
taosTmrStop(pSyncNode->pHeartbeatTimer);
pSyncNode->pHeartbeatTimer = NULL;
return ret;
}
// utils --------------
int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg) {
SEpSet epSet;
syncUtilraftId2EpSet(destRaftId, &epSet);
pSyncNode->FpSendMsg(pSyncNode->rpcClient, &epSet, pMsg);
return 0;
}
int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg) {
SEpSet epSet;
syncUtilnodeInfo2EpSet(nodeInfo, &epSet);
pSyncNode->FpSendMsg(pSyncNode->rpcClient, &epSet, pMsg);
return 0;
}
cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
char u64buf[128];
cJSON* pRoot = cJSON_CreateObject();
@ -253,12 +406,22 @@ cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
// tla+ server vars
cJSON_AddNumberToObject(pRoot, "state", pSyncNode->state);
cJSON_AddStringToObject(pRoot, "state_str", syncUtilState2String(pSyncNode->state));
char tmpBuf[RAFT_STORE_BLOCK_SIZE];
raftStoreSerialize(pSyncNode->pRaftStore, tmpBuf, sizeof(tmpBuf));
cJSON_AddStringToObject(pRoot, "pRaftStore", tmpBuf);
// tla+ candidate vars
cJSON_AddItemToObject(pRoot, "pVotesGranted", voteGranted2Json(pSyncNode->pVotesGranted));
cJSON_AddItemToObject(pRoot, "pVotesRespond", votesRespond2Json(pSyncNode->pVotesRespond));
// tla+ leader vars
cJSON_AddItemToObject(pRoot, "pNextIndex", syncIndexMgr2Json(pSyncNode->pNextIndex));
cJSON_AddItemToObject(pRoot, "pMatchIndex", syncIndexMgr2Json(pSyncNode->pMatchIndex));
// tla+ log vars
cJSON_AddItemToObject(pRoot, "pLogStore", logStore2Json(pSyncNode->pLogStore));
snprintf(u64buf, sizeof(u64buf), "%ld", pSyncNode->commitIndex);
cJSON_AddStringToObject(pRoot, "commitIndex", u64buf);
// ping timer
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->pPingTimer);
@ -268,8 +431,8 @@ cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
cJSON_AddStringToObject(pRoot, "pingTimerLogicClock", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSyncNode->pingTimerLogicClockUser);
cJSON_AddStringToObject(pRoot, "pingTimerLogicClockUser", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpPingTimer);
cJSON_AddStringToObject(pRoot, "FpPingTimer", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpPingTimerCB);
cJSON_AddStringToObject(pRoot, "FpPingTimerCB", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSyncNode->pingTimerCounter);
cJSON_AddStringToObject(pRoot, "pingTimerCounter", u64buf);
@ -281,8 +444,8 @@ cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
cJSON_AddStringToObject(pRoot, "electTimerLogicClock", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSyncNode->electTimerLogicClockUser);
cJSON_AddStringToObject(pRoot, "electTimerLogicClockUser", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpElectTimer);
cJSON_AddStringToObject(pRoot, "FpElectTimer", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpElectTimerCB);
cJSON_AddStringToObject(pRoot, "FpElectTimerCB", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSyncNode->electTimerCounter);
cJSON_AddStringToObject(pRoot, "electTimerCounter", u64buf);
@ -294,8 +457,8 @@ cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
cJSON_AddStringToObject(pRoot, "heartbeatTimerLogicClock", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSyncNode->heartbeatTimerLogicClockUser);
cJSON_AddStringToObject(pRoot, "heartbeatTimerLogicClockUser", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpHeartbeatTimer);
cJSON_AddStringToObject(pRoot, "FpHeartbeatTimer", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpHeartbeatTimerCB);
cJSON_AddStringToObject(pRoot, "FpHeartbeatTimerCB", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSyncNode->heartbeatTimerCounter);
cJSON_AddStringToObject(pRoot, "heartbeatTimerCounter", u64buf);
@ -327,6 +490,107 @@ char* syncNode2Str(const SSyncNode* pSyncNode) {
return serialized;
}
// raft state change --------------
void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
if (term > pSyncNode->pRaftStore->currentTerm) {
raftStoreSetTerm(pSyncNode->pRaftStore, term);
syncNodeBecomeFollower(pSyncNode);
raftStoreClearVote(pSyncNode->pRaftStore);
}
}
void syncNodeBecomeFollower(SSyncNode* pSyncNode) {
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
pSyncNode->leaderCache = EMPTY_RAFT_ID;
}
pSyncNode->state = TAOS_SYNC_STATE_FOLLOWER;
syncNodeStopHeartbeatTimer(pSyncNode);
int32_t electMS = syncUtilElectRandomMS();
syncNodeRestartElectTimer(pSyncNode, electMS);
}
// TLA+ Spec
// \* Candidate i transitions to leader.
// BecomeLeader(i) ==
// /\ state[i] = Candidate
// /\ votesGranted[i] \in Quorum
// /\ state' = [state EXCEPT ![i] = Leader]
// /\ nextIndex' = [nextIndex EXCEPT ![i] =
// [j \in Server |-> Len(log[i]) + 1]]
// /\ matchIndex' = [matchIndex EXCEPT ![i] =
// [j \in Server |-> 0]]
// /\ elections' = elections \cup
// {[eterm |-> currentTerm[i],
// eleader |-> i,
// elog |-> log[i],
// evotes |-> votesGranted[i],
// evoterLog |-> voterLog[i]]}
// /\ UNCHANGED <<messages, currentTerm, votedFor, candidateVars, logVars>>
//
void syncNodeBecomeLeader(SSyncNode* pSyncNode) {
pSyncNode->state = TAOS_SYNC_STATE_LEADER;
pSyncNode->leaderCache = pSyncNode->myRaftId;
for (int i = 0; i < pSyncNode->pNextIndex->replicaNum; ++i) {
pSyncNode->pNextIndex->index[i] = pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore) + 1;
}
for (int i = 0; i < pSyncNode->pMatchIndex->replicaNum; ++i) {
pSyncNode->pMatchIndex->index[i] = SYNC_INDEX_INVALID;
}
syncNodeStopElectTimer(pSyncNode);
syncNodeStartHeartbeatTimer(pSyncNode);
syncNodeReplicate(pSyncNode);
}
void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
assert(voteGrantedMajority(pSyncNode->pVotesGranted));
syncNodeBecomeLeader(pSyncNode);
}
void syncNodeFollower2Candidate(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER);
pSyncNode->state = TAOS_SYNC_STATE_CANDIDATE;
}
void syncNodeLeader2Follower(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_LEADER);
syncNodeBecomeFollower(pSyncNode);
}
void syncNodeCandidate2Follower(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
syncNodeBecomeFollower(pSyncNode);
}
// raft vote --------------
void syncNodeVoteForTerm(SSyncNode* pSyncNode, SyncTerm term, SRaftId* pRaftId) {
assert(term == pSyncNode->pRaftStore->currentTerm);
assert(!raftStoreHasVoted(pSyncNode->pRaftStore));
raftStoreVote(pSyncNode->pRaftStore, pRaftId);
}
void syncNodeVoteForSelf(SSyncNode* pSyncNode) {
syncNodeVoteForTerm(pSyncNode, pSyncNode->pRaftStore->currentTerm, &(pSyncNode->myRaftId));
SyncRequestVoteReply* pMsg = syncRequestVoteReplyBuild();
pMsg->srcId = pSyncNode->myRaftId;
pMsg->destId = pSyncNode->myRaftId;
pMsg->term = pSyncNode->pRaftStore->currentTerm;
pMsg->voteGranted = true;
voteGrantedVote(pSyncNode->pVotesGranted, pMsg);
votesRespondAdd(pSyncNode->pVotesRespond, pMsg);
syncRequestVoteReplyDestroy(pMsg);
}
void syncNodeMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {}
// for debug --------------
void syncNodePrint(SSyncNode* pObj) {
char* serialized = syncNode2Str(pObj);
@ -354,199 +618,24 @@ void syncNodeLog2(char* s, SSyncNode* pObj) {
free(serialized);
}
int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg) {
SEpSet epSet;
syncUtilraftId2EpSet(destRaftId, &epSet);
pSyncNode->FpSendMsg(pSyncNode->rpcClient, &epSet, pMsg);
return 0;
}
int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg) {
SEpSet epSet;
syncUtilnodeInfo2EpSet(nodeInfo, &epSet);
pSyncNode->FpSendMsg(pSyncNode->rpcClient, &epSet, pMsg);
return 0;
}
int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg) {
sTrace("syncNodePing pSyncNode:%p ", pSyncNode);
int32_t ret = 0;
SRpcMsg rpcMsg;
syncPing2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(destRaftId, pSyncNode, &rpcMsg);
{
cJSON* pJson = syncPing2Json(pMsg);
char* serialized = cJSON_Print(pJson);
sTrace("syncNodePing pMsg:%s ", serialized);
free(serialized);
cJSON_Delete(pJson);
}
{
SyncPing* pMsg2 = rpcMsg.pCont;
cJSON* pJson = syncPing2Json(pMsg2);
char* serialized = cJSON_Print(pJson);
sTrace("syncNodePing rpcMsg.pCont:%s ", serialized);
free(serialized);
cJSON_Delete(pJson);
}
return ret;
}
int32_t syncNodePingAll(SSyncNode* pSyncNode) {
sTrace("syncNodePingAll pSyncNode:%p ", pSyncNode);
int32_t ret = 0;
for (int i = 0; i < pSyncNode->syncCfg.replicaNum; ++i) {
SRaftId destId;
syncUtilnodeInfo2raftId(&pSyncNode->syncCfg.nodeInfo[i], pSyncNode->vgId, &destId);
SyncPing* pMsg = syncPingBuild3(&pSyncNode->myRaftId, &destId);
ret = syncNodePing(pSyncNode, &destId, pMsg);
assert(ret == 0);
syncPingDestroy(pMsg);
}
}
int32_t syncNodePingPeers(SSyncNode* pSyncNode) {
int32_t ret = 0;
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SRaftId destId;
syncUtilnodeInfo2raftId(&pSyncNode->peersNodeInfo[i], pSyncNode->vgId, &destId);
SyncPing* pMsg = syncPingBuild3(&pSyncNode->myRaftId, &destId);
ret = syncNodePing(pSyncNode, &destId, pMsg);
assert(ret == 0);
syncPingDestroy(pMsg);
}
}
int32_t syncNodePingSelf(SSyncNode* pSyncNode) {
int32_t ret;
SyncPing* pMsg = syncPingBuild3(&pSyncNode->myRaftId, &pSyncNode->myRaftId);
ret = syncNodePing(pSyncNode, &pMsg->destId, pMsg);
assert(ret == 0);
syncPingDestroy(pMsg);
}
int32_t syncNodeStartPingTimer(SSyncNode* pSyncNode) {
atomic_store_64(&pSyncNode->pingTimerLogicClock, pSyncNode->pingTimerLogicClockUser);
pSyncNode->pingTimerMS = PING_TIMER_MS;
if (pSyncNode->pPingTimer == NULL) {
pSyncNode->pPingTimer =
taosTmrStart(pSyncNode->FpPingTimer, pSyncNode->pingTimerMS, pSyncNode, gSyncEnv->pTimerManager);
} else {
taosTmrReset(pSyncNode->FpPingTimer, pSyncNode->pingTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pPingTimer);
}
return 0;
}
int32_t syncNodeStopPingTimer(SSyncNode* pSyncNode) {
atomic_add_fetch_64(&pSyncNode->pingTimerLogicClockUser, 1);
pSyncNode->pingTimerMS = TIMER_MAX_MS;
return 0;
}
int32_t syncNodeStartElectTimer(SSyncNode* pSyncNode, int32_t ms) {
pSyncNode->electTimerMS = ms;
atomic_store_64(&pSyncNode->electTimerLogicClock, pSyncNode->electTimerLogicClockUser);
if (pSyncNode->pElectTimer == NULL) {
pSyncNode->pElectTimer =
taosTmrStart(pSyncNode->FpElectTimer, pSyncNode->electTimerMS, pSyncNode, gSyncEnv->pTimerManager);
} else {
taosTmrReset(pSyncNode->FpElectTimer, pSyncNode->electTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pElectTimer);
}
return 0;
}
int32_t syncNodeStopElectTimer(SSyncNode* pSyncNode) {
atomic_add_fetch_64(&pSyncNode->electTimerLogicClockUser, 1);
pSyncNode->electTimerMS = TIMER_MAX_MS;
return 0;
}
int32_t syncNodeRestartElectTimer(SSyncNode* pSyncNode, int32_t ms) {
syncNodeStopElectTimer(pSyncNode);
syncNodeStartElectTimer(pSyncNode, ms);
return 0;
}
int32_t syncNodeStartHeartbeatTimer(SSyncNode* pSyncNode) {
atomic_store_64(&pSyncNode->heartbeatTimerLogicClock, pSyncNode->heartbeatTimerLogicClockUser);
if (pSyncNode->pHeartbeatTimer == NULL) {
pSyncNode->pHeartbeatTimer =
taosTmrStart(pSyncNode->FpHeartbeatTimer, pSyncNode->heartbeatTimerMS, pSyncNode, gSyncEnv->pTimerManager);
} else {
taosTmrReset(pSyncNode->FpHeartbeatTimer, pSyncNode->heartbeatTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pHeartbeatTimer);
}
return 0;
}
int32_t syncNodeStopHeartbeatTimer(SSyncNode* pSyncNode) {
atomic_add_fetch_64(&pSyncNode->heartbeatTimerLogicClockUser, 1);
pSyncNode->heartbeatTimerMS = TIMER_MAX_MS;
return 0;
}
// ------ local funciton ---------
static int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg) {
int32_t ret = 0;
sTrace("<-- syncNodeOnPingCb -->");
{
cJSON* pJson = syncPing2Json(pMsg);
char* serialized = cJSON_Print(pJson);
sTrace("process syncMessage recv: syncNodeOnPingCb pMsg:%s ", serialized);
free(serialized);
cJSON_Delete(pJson);
}
SyncPingReply* pMsgReply = syncPingReplyBuild3(&ths->myRaftId, &pMsg->srcId);
SRpcMsg rpcMsg;
syncPingReply2RpcMsg(pMsgReply, &rpcMsg);
syncNodeSendMsgById(&pMsgReply->destId, ths, &rpcMsg);
return ret;
}
static int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg) {
int32_t ret = 0;
sTrace("<-- syncNodeOnPingReplyCb -->");
{
cJSON* pJson = syncPingReply2Json(pMsg);
char* serialized = cJSON_Print(pJson);
sTrace("process syncMessage recv: syncNodeOnPingReplyCb pMsg:%s ", serialized);
free(serialized);
cJSON_Delete(pJson);
}
return ret;
}
// enqueue message ----
static void syncNodeEqPingTimer(void* param, void* tmrId) {
sTrace("<-- syncNodeEqPingTimer -->");
SSyncNode* pSyncNode = (SSyncNode*)param;
if (atomic_load_64(&pSyncNode->pingTimerLogicClockUser) <= atomic_load_64(&pSyncNode->pingTimerLogicClock)) {
SyncTimeout* pSyncMsg = syncTimeoutBuild2(SYNC_TIMEOUT_PING, atomic_load_64(&pSyncNode->pingTimerLogicClock),
pSyncNode->pingTimerMS, pSyncNode);
SRpcMsg rpcMsg;
syncTimeout2RpcMsg(pSyncMsg, &rpcMsg);
syncRpcMsgLog2((char*)"==syncNodeEqPingTimer==", &rpcMsg);
pSyncNode->FpEqMsg(pSyncNode->queue, &rpcMsg);
syncTimeoutDestroy(pSyncMsg);
// reset timer ms
// pSyncNode->pingTimerMS += 100;
taosTmrReset(syncNodeEqPingTimer, pSyncNode->pingTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pPingTimer);
} else {
sTrace("syncNodeEqPingTimer: pingTimerLogicClock:%lu, pingTimerLogicClockUser:%lu", pSyncNode->pingTimerLogicClock,
pSyncNode->pingTimerLogicClockUser);
sTrace("==syncNodeEqPingTimer== pingTimerLogicClock:%lu, pingTimerLogicClockUser:%lu",
pSyncNode->pingTimerLogicClock, pSyncNode->pingTimerLogicClockUser);
}
}
@ -555,19 +644,18 @@ static void syncNodeEqElectTimer(void* param, void* tmrId) {
if (atomic_load_64(&pSyncNode->electTimerLogicClockUser) <= atomic_load_64(&pSyncNode->electTimerLogicClock)) {
SyncTimeout* pSyncMsg = syncTimeoutBuild2(SYNC_TIMEOUT_ELECTION, atomic_load_64(&pSyncNode->electTimerLogicClock),
pSyncNode->electTimerMS, pSyncNode);
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncTimeout2RpcMsg(pSyncMsg, &rpcMsg);
syncRpcMsgLog2((char*)"==syncNodeEqElectTimer==", &rpcMsg);
pSyncNode->FpEqMsg(pSyncNode->queue, &rpcMsg);
syncTimeoutDestroy(pSyncMsg);
// reset timer ms
pSyncNode->electTimerMS = syncUtilElectRandomMS();
taosTmrReset(syncNodeEqPingTimer, pSyncNode->pingTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pPingTimer);
} else {
sTrace("syncNodeEqElectTimer: electTimerLogicClock:%lu, electTimerLogicClockUser:%lu",
sTrace("==syncNodeEqElectTimer== electTimerLogicClock:%lu, electTimerLogicClockUser:%lu",
pSyncNode->electTimerLogicClock, pSyncNode->electTimerLogicClockUser);
}
}
@ -579,80 +667,50 @@ static void syncNodeEqHeartbeatTimer(void* param, void* tmrId) {
SyncTimeout* pSyncMsg =
syncTimeoutBuild2(SYNC_TIMEOUT_HEARTBEAT, atomic_load_64(&pSyncNode->heartbeatTimerLogicClock),
pSyncNode->heartbeatTimerMS, pSyncNode);
SRpcMsg rpcMsg;
syncTimeout2RpcMsg(pSyncMsg, &rpcMsg);
syncRpcMsgLog2((char*)"==syncNodeEqHeartbeatTimer==", &rpcMsg);
pSyncNode->FpEqMsg(pSyncNode->queue, &rpcMsg);
syncTimeoutDestroy(pSyncMsg);
// reset timer ms
// pSyncNode->heartbeatTimerMS += 100;
taosTmrReset(syncNodeEqHeartbeatTimer, pSyncNode->heartbeatTimerMS, pSyncNode, gSyncEnv->pTimerManager,
&pSyncNode->pHeartbeatTimer);
} else {
sTrace("syncNodeEqHeartbeatTimer: heartbeatTimerLogicClock:%lu, heartbeatTimerLogicClockUser:%lu",
sTrace("==syncNodeEqHeartbeatTimer== heartbeatTimerLogicClock:%lu, heartbeatTimerLogicClockUser:%lu",
pSyncNode->heartbeatTimerLogicClock, pSyncNode->heartbeatTimerLogicClockUser);
}
}
static void UpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
if (term > pSyncNode->pRaftStore->currentTerm) {
pSyncNode->pRaftStore->currentTerm = term;
pSyncNode->pRaftStore->voteFor = EMPTY_RAFT_ID;
raftStorePersist(pSyncNode->pRaftStore);
syncNodeBecomeFollower(pSyncNode);
}
// on message ----
static int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg) {
int32_t ret = 0;
syncPingLog2("==syncNodeOnPingCb==", pMsg);
SyncPingReply* pMsgReply = syncPingReplyBuild3(&ths->myRaftId, &pMsg->srcId);
SRpcMsg rpcMsg;
syncPingReply2RpcMsg(pMsgReply, &rpcMsg);
syncNodeSendMsgById(&pMsgReply->destId, ths, &rpcMsg);
return ret;
}
static void syncNodeBecomeFollower(SSyncNode* pSyncNode) {
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
pSyncNode->leaderCache = EMPTY_RAFT_ID;
static int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg) {
int32_t ret = 0;
syncPingReplyLog2("==syncNodeOnPingReplyCb==", pMsg);
return ret;
}
static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg) {
int32_t ret = 0;
syncClientRequestLog2("==syncNodeOnClientRequestCb==", pMsg);
if (ths->state == TAOS_SYNC_STATE_LEADER) {
SSyncRaftEntry* pEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
syncNodeReplicate(ths);
syncEntryDestory(pEntry);
} else {
// ths->pFsm->FpCommitCb(-1)
}
syncNodeStopHeartbeatTimer(pSyncNode);
int32_t electMS = syncUtilElectRandomMS();
syncNodeStartElectTimer(pSyncNode, electMS);
}
// TLA+ Spec
// \* Candidate i transitions to leader.
// BecomeLeader(i) ==
// /\ state[i] = Candidate
// /\ votesGranted[i] \in Quorum
// /\ state' = [state EXCEPT ![i] = Leader]
// /\ nextIndex' = [nextIndex EXCEPT ![i] =
// [j \in Server |-> Len(log[i]) + 1]]
// /\ matchIndex' = [matchIndex EXCEPT ![i] =
// [j \in Server |-> 0]]
// /\ elections' = elections \cup
// {[eterm |-> currentTerm[i],
// eleader |-> i,
// elog |-> log[i],
// evotes |-> votesGranted[i],
// evoterLog |-> voterLog[i]]}
// /\ UNCHANGED <<messages, currentTerm, votedFor, candidateVars, logVars>>
//
static void syncNodeBecomeLeader(SSyncNode* pSyncNode) {
pSyncNode->state = TAOS_SYNC_STATE_LEADER;
pSyncNode->leaderCache = pSyncNode->myRaftId;
// next Index +=1
// match Index = 0;
syncNodeStopElectTimer(pSyncNode);
syncNodeStartHeartbeatTimer(pSyncNode);
// appendEntries;
}
static void syncNodeFollower2Candidate(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER);
pSyncNode->state = TAOS_SYNC_STATE_CANDIDATE;
}
static void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {}
static void syncNodeLeader2Follower(SSyncNode* pSyncNode) {}
static void syncNodeCandidate2Follower(SSyncNode* pSyncNode) {}
return ret;
}

View File

@ -65,10 +65,32 @@ cJSON* syncRpcMsg2Json(SRpcMsg* pRpcMsg) {
pRoot = syncAppendEntriesReply2Json(pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == SYNC_RESPONSE) {
pRoot = cJSON_CreateObject();
char* s;
s = syncUtilprintBin((char*)(pRpcMsg->pCont), pRpcMsg->contLen);
cJSON_AddStringToObject(pRoot, "pCont", s);
free(s);
s = syncUtilprintBin2((char*)(pRpcMsg->pCont), pRpcMsg->contLen);
cJSON_AddStringToObject(pRoot, "pCont2", s);
free(s);
} else {
pRoot = syncRpcUnknownMsg2Json();
char* s;
s = syncUtilprintBin((char*)(pRpcMsg->pCont), pRpcMsg->contLen);
cJSON_AddStringToObject(pRoot, "pCont", s);
free(s);
s = syncUtilprintBin2((char*)(pRpcMsg->pCont), pRpcMsg->contLen);
cJSON_AddStringToObject(pRoot, "pCont2", s);
free(s);
}
cJSON_AddNumberToObject(pRoot, "msgType", pRpcMsg->msgType);
cJSON_AddNumberToObject(pRoot, "contLen", pRpcMsg->contLen);
cJSON_AddNumberToObject(pRoot, "code", pRpcMsg->code);
// cJSON_AddNumberToObject(pRoot, "persist", pRpcMsg->persist);
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "RpcMsg", pRoot);
return pJson;
@ -77,7 +99,7 @@ cJSON* syncRpcMsg2Json(SRpcMsg* pRpcMsg) {
cJSON* syncRpcUnknownMsg2Json() {
cJSON* pRoot = cJSON_CreateObject();
cJSON_AddNumberToObject(pRoot, "msgType", SYNC_UNKNOWN);
cJSON_AddStringToObject(pRoot, "data", "known message");
cJSON_AddStringToObject(pRoot, "data", "unknown message");
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SyncUnknown", pRoot);
@ -798,8 +820,8 @@ cJSON* syncRequestVote2Json(const SyncRequestVote* pMsg) {
cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
cJSON_AddItemToObject(pRoot, "destId", pDestId);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->currentTerm);
cJSON_AddStringToObject(pRoot, "currentTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastLogIndex);
cJSON_AddStringToObject(pRoot, "lastLogIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastLogTerm);
@ -1086,6 +1108,9 @@ cJSON* syncAppendEntries2Json(const SyncAppendEntries* pMsg) {
cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
cJSON_AddItemToObject(pRoot, "destId", pDestId);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->prevLogIndex);
cJSON_AddStringToObject(pRoot, "pre_log_index", u64buf);
@ -1242,9 +1267,11 @@ cJSON* syncAppendEntriesReply2Json(const SyncAppendEntriesReply* pMsg) {
cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
cJSON_AddItemToObject(pRoot, "destId", pDestId);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
cJSON_AddNumberToObject(pRoot, "success", pMsg->success);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->matchIndex);
cJSON_AddStringToObject(pRoot, "match_index", u64buf);
cJSON_AddStringToObject(pRoot, "matchIndex", u64buf);
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SyncAppendEntriesReply", pRoot);
@ -1283,4 +1310,4 @@ void syncAppendEntriesReplyLog2(char* s, const SyncAppendEntriesReply* pMsg) {
char* serialized = syncAppendEntriesReply2Str(pMsg);
sTrace("syncAppendEntriesReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
free(serialized);
}
}

View File

@ -130,7 +130,7 @@ cJSON* logStore2Json(SSyncLogStore* pLogStore) {
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pData->pWal);
cJSON_AddStringToObject(pRoot, "pWal", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastIndex(pLogStore));
snprintf(u64buf, sizeof(u64buf), "%ld", logStoreLastIndex(pLogStore));
cJSON_AddStringToObject(pRoot, "LastIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastTerm(pLogStore));
cJSON_AddStringToObject(pRoot, "LastTerm", u64buf);

View File

@ -15,6 +15,8 @@
#include "syncRaftStore.h"
#include "cJSON.h"
#include "syncEnv.h"
#include "syncUtil.h"
// private function
static int32_t raftStoreInit(SRaftStore *pRaftStore);
@ -135,6 +137,33 @@ int32_t raftStoreDeserialize(SRaftStore *pRaftStore, char *buf, size_t len) {
return 0;
}
bool raftStoreHasVoted(SRaftStore *pRaftStore) {
bool b = syncUtilEmptyId(&(pRaftStore->voteFor));
return b;
}
void raftStoreVote(SRaftStore *pRaftStore, SRaftId *pRaftId) {
assert(!raftStoreHasVoted(pRaftStore));
assert(!syncUtilEmptyId(pRaftId));
pRaftStore->voteFor = *pRaftId;
raftStorePersist(pRaftStore);
}
void raftStoreClearVote(SRaftStore *pRaftStore) {
pRaftStore->voteFor = EMPTY_RAFT_ID;
raftStorePersist(pRaftStore);
}
void raftStoreNextTerm(SRaftStore *pRaftStore) {
++(pRaftStore->currentTerm);
raftStorePersist(pRaftStore);
}
void raftStoreSetTerm(SRaftStore *pRaftStore, SyncTerm term) {
pRaftStore->currentTerm = term;
raftStorePersist(pRaftStore);
}
// for debug -------------------
void raftStorePrint(SRaftStore *pObj) {
char serialized[RAFT_STORE_BLOCK_SIZE];

View File

@ -14,7 +14,11 @@
*/
#include "syncReplication.h"
#include "syncIndexMgr.h"
#include "syncMessage.h"
#include "syncRaftEntry.h"
#include "syncRaftLog.h"
#include "syncUtil.h"
// TLA+ Spec
// AppendEntries(i, j) ==
@ -42,7 +46,39 @@
// /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_LEADER);
int32_t ret = 0;
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SRaftId* pDestId = &(pSyncNode->peersId[i]);
SyncIndex nextIndex = syncIndexMgrGetIndex(pSyncNode->pNextIndex, pDestId);
SyncIndex preLogIndex = nextIndex - 1;
SyncTerm preLogTerm = 0;
if (preLogIndex >= SYNC_INDEX_BEGIN) {
SSyncRaftEntry* pPreEntry = pSyncNode->pLogStore->getEntry(pSyncNode->pLogStore, preLogIndex);
preLogTerm = pPreEntry->term;
}
SyncIndex lastIndex = syncUtilMinIndex(pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore), nextIndex);
assert(nextIndex == lastIndex);
SSyncRaftEntry* pEntry = logStoreGetEntry(pSyncNode->pLogStore, nextIndex);
assert(pEntry != NULL);
SyncAppendEntries* pMsg = syncAppendEntriesBuild(pEntry->bytes);
pMsg->srcId = pSyncNode->myRaftId;
pMsg->destId = *pDestId;
pMsg->prevLogIndex = preLogIndex;
pMsg->prevLogTerm = preLogTerm;
pMsg->commitIndex = pSyncNode->commitIndex;
pMsg->dataLen = pEntry->bytes;
// add pEntry into msg
syncNodeAppendEntries(pSyncNode, pDestId, pMsg);
}
return ret;
}

View File

@ -14,6 +14,10 @@
*/
#include "syncRequestVote.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
// TLA+ Spec
// HandleRequestVoteRequest(i, j, m) ==
@ -37,4 +41,34 @@
// m)
// /\ UNCHANGED <<state, currentTerm, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg) {}
int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg) {
int32_t ret = 0;
syncRequestVoteLog2("==syncNodeOnRequestVoteCb==", pMsg);
if (pMsg->term > ths->pRaftStore->currentTerm) {
syncNodeUpdateTerm(ths, pMsg->term);
}
assert(pMsg->term <= ths->pRaftStore->currentTerm);
bool logOK = (pMsg->lastLogTerm > ths->pLogStore->getLastTerm(ths->pLogStore)) ||
((pMsg->lastLogTerm == ths->pLogStore->getLastTerm(ths->pLogStore)) &&
(pMsg->lastLogIndex >= ths->pLogStore->getLastIndex(ths->pLogStore)));
bool grant = (pMsg->term == ths->pRaftStore->currentTerm) && logOK &&
((!raftStoreHasVoted(ths->pRaftStore)) || (syncUtilSameId(&(ths->pRaftStore->voteFor), &(pMsg->srcId))));
if (grant) {
raftStoreVote(ths->pRaftStore, &(pMsg->srcId));
}
SyncRequestVoteReply* pReply = syncRequestVoteReplyBuild();
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->voteGranted = grant;
SRpcMsg rpcMsg;
syncRequestVoteReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncRequestVoteReplyDestroy(pReply);
return ret;
}

View File

@ -14,6 +14,10 @@
*/
#include "syncRequestVoteReply.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
// TLA+ Spec
// HandleRequestVoteResponse(i, j, m) ==
@ -32,4 +36,33 @@
// /\ Discard(m)
// /\ UNCHANGED <<serverVars, votedFor, leaderVars, logVars>>
//
int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg) {}
int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg) {
int32_t ret = 0;
syncRequestVoteReplyLog2("==syncNodeOnRequestVoteReplyCb==", pMsg);
if (pMsg->term < ths->pRaftStore->currentTerm) {
sTrace("DropStaleResponse, receive term:%lu, current term:%lu", pMsg->term, ths->pRaftStore->currentTerm);
return ret;
}
// no need this code, because if I receive reply.term, then I must have sent for that term.
// if (pMsg->term > ths->pRaftStore->currentTerm) {
// syncNodeUpdateTerm(ths, pMsg->term);
// }
assert(pMsg->term == ths->pRaftStore->currentTerm);
if (ths->state == TAOS_SYNC_STATE_CANDIDATE) {
votesRespondAdd(ths->pVotesRespond, pMsg);
if (pMsg->voteGranted) {
voteGrantedVote(ths->pVotesGranted, pMsg);
if (voteGrantedMajority(ths->pVotesGranted)) {
if (ths->pVotesGranted->toLeader) {
syncNodeCandidate2Leader(ths);
ths->pVotesGranted->toLeader = true;
}
}
}
}
return ret;
}

View File

@ -19,15 +19,7 @@
int32_t syncNodeOnTimeoutCb(SSyncNode* ths, SyncTimeout* pMsg) {
int32_t ret = 0;
sTrace("<-- syncNodeOnTimeoutCb -->");
{
cJSON* pJson = syncTimeout2Json(pMsg);
char* serialized = cJSON_Print(pJson);
sTrace("process syncMessage recv: syncNodeOnTimeoutCb pMsg:%s ", serialized);
free(serialized);
cJSON_Delete(pJson);
}
syncTimeoutLog2("==syncNodeOnTimeoutCb==", pMsg);
if (pMsg->timeoutType == SYNC_TIMEOUT_PING) {
if (atomic_load_64(&ths->pingTimerLogicClockUser) <= pMsg->logicClock) {

View File

@ -74,6 +74,8 @@ bool syncUtilSameId(const SRaftId* pId1, const SRaftId* pId2) {
return ret;
}
bool syncUtilEmptyId(const SRaftId* pId) { return (pId->addr == 0 && pId->vgId == 0); }
// ---- SSyncBuffer -----
void syncUtilbufBuild(SSyncBuffer* syncBuf, size_t len) {
syncBuf->len = len;
@ -184,4 +186,14 @@ char* syncUtilprintBin2(char* ptr, uint32_t len) {
p += n;
}
return s;
}
SyncIndex syncUtilMinIndex(SyncIndex a, SyncIndex b) {
SyncIndex r = a < b ? a : b;
return r;
}
SyncIndex syncUtilMaxIndex(SyncIndex a, SyncIndex b) {
SyncIndex r = a > b ? a : b;
return r;
}

View File

@ -4,8 +4,8 @@ add_executable(syncPingTimerTest "")
add_executable(syncIOTickQTest "")
add_executable(syncIOTickPingTest "")
add_executable(syncIOSendMsgTest "")
add_executable(syncIOSendMsgClientTest "")
add_executable(syncIOSendMsgServerTest "")
add_executable(syncIOClientTest "")
add_executable(syncIOServerTest "")
add_executable(syncRaftStoreTest "")
add_executable(syncEnqTest "")
add_executable(syncIndexTest "")
@ -51,13 +51,13 @@ target_sources(syncIOSendMsgTest
PRIVATE
"syncIOSendMsgTest.cpp"
)
target_sources(syncIOSendMsgClientTest
target_sources(syncIOClientTest
PRIVATE
"syncIOSendMsgClientTest.cpp"
"syncIOClientTest.cpp"
)
target_sources(syncIOSendMsgServerTest
target_sources(syncIOServerTest
PRIVATE
"syncIOSendMsgServerTest.cpp"
"syncIOServerTest.cpp"
)
target_sources(syncRaftStoreTest
PRIVATE
@ -167,12 +167,12 @@ target_include_directories(syncIOSendMsgTest
"${CMAKE_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncIOSendMsgClientTest
target_include_directories(syncIOClientTest
PUBLIC
"${CMAKE_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncIOSendMsgServerTest
target_include_directories(syncIOServerTest
PUBLIC
"${CMAKE_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
@ -298,11 +298,11 @@ target_link_libraries(syncIOSendMsgTest
sync
gtest_main
)
target_link_libraries(syncIOSendMsgClientTest
target_link_libraries(syncIOClientTest
sync
gtest_main
)
target_link_libraries(syncIOSendMsgServerTest
target_link_libraries(syncIOServerTest
sync
gtest_main
)

View File

@ -1,9 +1,10 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
void logTest() {
sTrace("--- sync log test: trace");
@ -14,64 +15,69 @@ void logTest() {
sFatal("--- sync log test: fatal");
}
uint16_t ports[3] = {7010, 7110, 7210};
uint16_t ports[] = {7010, 7110, 7210, 7310, 7410};
int32_t replicaNum = 5;
int32_t myIndex = 0;
SSyncNode* doSync(int myIndex) {
SSyncFSM* pFsm;
SRaftId ids[TSDB_MAX_REPLICA];
SSyncInfo syncInfo;
SSyncFSM* pFsm;
SSyncInfo syncInfo;
syncInfo.vgId = 1;
SSyncNode* syncNodeInit() {
syncInfo.vgId = 1234;
syncInfo.rpcClient = gSyncIO->clientRpc;
syncInfo.FpSendMsg = syncIOSendMsg;
syncInfo.queue = gSyncIO->pMsgQ;
syncInfo.FpEqMsg = syncIOEqMsg;
syncInfo.pFsm = pFsm;
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s", "./test_sync_ping");
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s", "./");
SSyncCfg* pCfg = &syncInfo.syncCfg;
pCfg->myIndex = myIndex;
pCfg->replicaNum = 3;
pCfg->replicaNum = replicaNum;
pCfg->nodeInfo[0].nodePort = ports[0];
snprintf(pCfg->nodeInfo[0].nodeFqdn, sizeof(pCfg->nodeInfo[0].nodeFqdn), "%s", "127.0.0.1");
// taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
pCfg->nodeInfo[1].nodePort = ports[1];
snprintf(pCfg->nodeInfo[1].nodeFqdn, sizeof(pCfg->nodeInfo[1].nodeFqdn), "%s", "127.0.0.1");
// taosGetFqdn(pCfg->nodeInfo[1].nodeFqdn);
pCfg->nodeInfo[2].nodePort = ports[2];
snprintf(pCfg->nodeInfo[2].nodeFqdn, sizeof(pCfg->nodeInfo[2].nodeFqdn), "%s", "127.0.0.1");
// taosGetFqdn(pCfg->nodeInfo[2].nodeFqdn);
for (int i = 0; i < replicaNum; ++i) {
pCfg->nodeInfo[i].nodePort = ports[i];
snprintf(pCfg->nodeInfo[i].nodeFqdn, sizeof(pCfg->nodeInfo[i].nodeFqdn), "%s", "127.0.0.1");
// taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
}
SSyncNode* pSyncNode = syncNodeOpen(&syncInfo);
assert(pSyncNode != NULL);
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->FpOnSyncRequestVote = pSyncNode->FpOnRequestVote;
gSyncIO->FpOnSyncRequestVoteReply = pSyncNode->FpOnRequestVoteReply;
gSyncIO->FpOnSyncAppendEntries = pSyncNode->FpOnAppendEntries;
gSyncIO->FpOnSyncAppendEntriesReply = pSyncNode->FpOnAppendEntriesReply;
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->FpOnSyncTimeout = pSyncNode->FpOnTimeout;
gSyncIO->pSyncNode = pSyncNode;
return pSyncNode;
}
void timerPingAll(void* param, void* tmrId) {
SSyncNode* pSyncNode = (SSyncNode*)param;
syncNodePingAll(pSyncNode);
SSyncNode* syncInitTest() { return syncNodeInit(); }
void initRaftId(SSyncNode* pSyncNode) {
for (int i = 0; i < replicaNum; ++i) {
ids[i] = pSyncNode->replicasId[i];
char* s = syncUtilRaftId2Str(&ids[i]);
printf("raftId[%d] : %s\n", i, s);
free(s);
}
}
int main(int argc, char** argv) {
// taosInitLog((char*)"syncPingTest.log", 100000, 10);
// taosInitLog((char *)"syncTest.log", 100000, 10);
tsAsyncLog = 0;
sDebugFlag = 143 + 64;
logTest();
int myIndex = 0;
myIndex = 0;
if (argc >= 2) {
myIndex = atoi(argv[1]);
if (myIndex > 2 || myIndex < 0) {
fprintf(stderr, "myIndex:%d error. should be 0 - 2", myIndex);
return 1;
}
}
int32_t ret = syncIOStart((char*)"127.0.0.1", ports[myIndex]);
@ -80,21 +86,22 @@ int main(int argc, char** argv) {
ret = syncEnvStart();
assert(ret == 0);
SSyncNode* pSyncNode = doSync(myIndex);
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
SSyncNode* pSyncNode = syncInitTest();
assert(pSyncNode != NULL);
syncNodePrint2((char*)"syncInitTest", pSyncNode);
initRaftId(pSyncNode);
//--------------------------------------------------------------
for (int i = 0; i < 10; ++i) {
SyncPingReply* pSyncMsg = syncPingReplyBuild3(&pSyncNode->myRaftId, &pSyncNode->myRaftId);
SyncPingReply* pSyncMsg = syncPingReplyBuild2(&pSyncNode->myRaftId, &pSyncNode->myRaftId, "syncEnqTest");
SRpcMsg rpcMsg;
syncPingReply2RpcMsg(pSyncMsg, &rpcMsg);
pSyncNode->FpEqMsg(pSyncNode->queue, &rpcMsg);
taosMsleep(1000);
}
while (1) {
taosMsleep(1000);
}
return 0;
}

View File

@ -14,15 +14,6 @@ void logTest() {
sFatal("--- sync log test: fatal");
}
void *pTimer = NULL;
void *pTimerMgr = NULL;
int g = 300;
static void timerFp(void *param, void *tmrId) {
printf("param:%p, tmrId:%p, pTimer:%p, pTimerMgr:%p \n", param, tmrId, pTimer, pTimerMgr);
taosTmrReset(timerFp, 1000, param, pTimerMgr, &pTimer);
}
int main() {
// taosInitLog((char*)"syncEnvTest.log", 100000, 10);
tsAsyncLog = 0;
@ -34,13 +25,20 @@ int main() {
ret = syncEnvStart();
assert(ret == 0);
// timer
pTimerMgr = taosTmrInit(1000, 50, 10000, "SYNC-ENV-TEST");
taosTmrStart(timerFp, 1000, &g, pTimerMgr);
for (int i = 0; i < 5; ++i) {
ret = syncEnvStartTimer();
assert(ret == 0);
while (1) {
taosMsleep(1000);
taosMsleep(5000);
ret = syncEnvStopTimer();
assert(ret == 0);
taosMsleep(5000);
}
ret = syncEnvStop();
assert(ret == 0);
return 0;
}

View File

@ -2,7 +2,8 @@
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncMessage.h"
#include "syncUtil.h"
void logTest() {
sTrace("--- sync log test: trace");
@ -22,7 +23,7 @@ int main() {
int32_t ret;
ret = syncIOStart((char *)"127.0.0.1", 7010);
ret = syncIOStart((char*)"127.0.0.1", 7010);
assert(ret == 0);
for (int i = 0; i < 10; ++i) {
@ -31,20 +32,19 @@ int main() {
epSet.numOfEps = 0;
addEpIntoEpSet(&epSet, "127.0.0.1", 7030);
SRpcMsg rpcMsg;
rpcMsg.contLen = 64;
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
snprintf((char *)rpcMsg.pCont, rpcMsg.contLen, "%s", "syncIOSendMsgTest");
rpcMsg.handle = NULL;
rpcMsg.msgType = 77;
SRaftId srcId, destId;
srcId.addr = syncUtilAddr2U64("127.0.0.1", 1234);
srcId.vgId = 100;
destId.addr = syncUtilAddr2U64("127.0.0.1", 5678);
destId.vgId = 100;
SyncPingReply* pSyncMsg = syncPingReplyBuild2(&srcId, &destId, "syncIOClientTest");
SRpcMsg rpcMsg;
syncPingReply2RpcMsg(pSyncMsg, &rpcMsg);
syncIOSendMsg(gSyncIO->clientRpc, &epSet, &rpcMsg);
taosSsleep(1);
}
while (1) {
taosSsleep(1);
}
return 0;
}

View File

@ -1,8 +1,10 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
void logTest() {
sTrace("--- sync log test: trace");
@ -13,37 +15,96 @@ void logTest() {
sFatal("--- sync log test: fatal");
}
int main() {
uint16_t ports[] = {7010, 7110, 7210, 7310, 7410};
int32_t replicaNum = 5;
int32_t myIndex = 0;
SRaftId ids[TSDB_MAX_REPLICA];
SSyncInfo syncInfo;
SSyncFSM* pFsm;
SSyncNode* syncNodeInit() {
syncInfo.vgId = 1234;
syncInfo.rpcClient = gSyncIO->clientRpc;
syncInfo.FpSendMsg = syncIOSendMsg;
syncInfo.queue = gSyncIO->pMsgQ;
syncInfo.FpEqMsg = syncIOEqMsg;
syncInfo.pFsm = pFsm;
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s", "./");
SSyncCfg* pCfg = &syncInfo.syncCfg;
pCfg->myIndex = myIndex;
pCfg->replicaNum = replicaNum;
for (int i = 0; i < replicaNum; ++i) {
pCfg->nodeInfo[i].nodePort = ports[i];
snprintf(pCfg->nodeInfo[i].nodeFqdn, sizeof(pCfg->nodeInfo[i].nodeFqdn), "%s", "127.0.0.1");
// taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
}
SSyncNode* pSyncNode = syncNodeOpen(&syncInfo);
assert(pSyncNode != NULL);
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->FpOnSyncRequestVote = pSyncNode->FpOnRequestVote;
gSyncIO->FpOnSyncRequestVoteReply = pSyncNode->FpOnRequestVoteReply;
gSyncIO->FpOnSyncAppendEntries = pSyncNode->FpOnAppendEntries;
gSyncIO->FpOnSyncAppendEntriesReply = pSyncNode->FpOnAppendEntriesReply;
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->FpOnSyncTimeout = pSyncNode->FpOnTimeout;
gSyncIO->pSyncNode = pSyncNode;
return pSyncNode;
}
SSyncNode* syncInitTest() { return syncNodeInit(); }
void initRaftId(SSyncNode* pSyncNode) {
for (int i = 0; i < replicaNum; ++i) {
ids[i] = pSyncNode->replicasId[i];
char* s = syncUtilRaftId2Str(&ids[i]);
printf("raftId[%d] : %s\n", i, s);
free(s);
}
}
int main(int argc, char** argv) {
// taosInitLog((char *)"syncTest.log", 100000, 10);
tsAsyncLog = 0;
sDebugFlag = 143 + 64;
logTest();
int32_t ret;
ret = syncIOStart((char *)"127.0.0.1", 7010);
assert(ret == 0);
for (int i = 0; i < 10; ++i) {
SEpSet epSet;
epSet.inUse = 0;
epSet.numOfEps = 0;
addEpIntoEpSet(&epSet, "127.0.0.1", 7010);
SRpcMsg rpcMsg;
rpcMsg.contLen = 64;
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
snprintf((char *)rpcMsg.pCont, rpcMsg.contLen, "%s", "syncIOSendMsgTest");
rpcMsg.handle = NULL;
rpcMsg.msgType = 77;
syncIOSendMsg(gSyncIO->clientRpc, &epSet, &rpcMsg);
taosSsleep(1);
myIndex = 0;
if (argc >= 2) {
myIndex = atoi(argv[1]);
}
while (1) {
taosSsleep(1);
int32_t ret = syncIOStart((char*)"127.0.0.1", ports[myIndex]);
assert(ret == 0);
ret = syncEnvStart();
assert(ret == 0);
SSyncNode* pSyncNode = syncInitTest();
assert(pSyncNode != NULL);
syncNodePrint2((char*)"syncInitTest", pSyncNode);
initRaftId(pSyncNode);
//--------------------------------------------------------------
for (int i = 0; i < 10; ++i) {
SyncPingReply* pSyncMsg = syncPingReplyBuild2(&pSyncNode->myRaftId, &pSyncNode->myRaftId, "syncIOSendMsgTest");
SRpcMsg rpcMsg;
syncPingReply2RpcMsg(pSyncMsg, &rpcMsg);
SEpSet epSet;
syncUtilnodeInfo2EpSet(&pSyncNode->myNodeInfo, &epSet);
pSyncNode->FpSendMsg(pSyncNode->rpcClient, &epSet, &rpcMsg);
taosMsleep(1000);
}
return 0;

View File

@ -25,8 +25,15 @@ int main() {
ret = syncIOStart((char*)"127.0.0.1", 7010);
assert(ret == 0);
ret = syncIOTickPing();
assert(ret == 0);
for (int i = 0; i < 3; ++i) {
ret = syncIOPingTimerStart();
assert(ret == 0);
taosMsleep(5000);
ret = syncIOPingTimerStop();
assert(ret == 0);
taosMsleep(5000);
}
while (1) {
taosSsleep(1);

View File

@ -25,11 +25,18 @@ int main() {
ret = syncIOStart((char*)"127.0.0.1", 7010);
assert(ret == 0);
ret = syncIOTickQ();
for (int i = 0; i < 3; ++i) {
ret = syncIOQTimerStart();
assert(ret == 0);
taosMsleep(5000);
ret = syncIOQTimerStop();
assert(ret == 0);
taosMsleep(5000);
}
ret = syncIOStop();
assert(ret == 0);
while (1) {
taosSsleep(1);
}
return 0;
}

View File

@ -81,7 +81,10 @@ SSyncNode* syncNodeInit() {
SSyncNode* syncInitTest() { return syncNodeInit(); }
void logStoreTest() {
logStorePrint(pSyncNode->pLogStore);
logStorePrint2((char*)"logStoreTest2", pSyncNode->pLogStore);
assert(pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore) == SYNC_INDEX_INVALID);
for (int i = 0; i < 5; ++i) {
int32_t dataLen = 10;
SSyncRaftEntry* pEntry = syncEntryBuild(dataLen);
@ -97,6 +100,10 @@ void logStoreTest() {
// syncEntryPrint2((char*)"write entry:", pEntry);
pSyncNode->pLogStore->appendEntry(pSyncNode->pLogStore, pEntry);
syncEntryDestory(pEntry);
if (i == 0) {
assert(pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore) == SYNC_INDEX_BEGIN);
}
}
logStorePrint(pSyncNode->pLogStore);
@ -129,6 +136,8 @@ int main(int argc, char** argv) {
ret = syncEnvStart();
assert(ret == 0);
taosRemoveDir("./wal_test");
pSyncNode = syncInitTest();
assert(pSyncNode != NULL);

View File

@ -20,7 +20,7 @@ SyncRequestVote *createMsg() {
pMsg->srcId.vgId = 100;
pMsg->destId.addr = syncUtilAddr2U64("127.0.0.1", 5678);
pMsg->destId.vgId = 100;
pMsg->currentTerm = 11;
pMsg->term = 11;
pMsg->lastLogIndex = 22;
pMsg->lastLogTerm = 33;
return pMsg;

View File

@ -57,7 +57,7 @@ SyncRequestVote *createSyncRequestVote() {
pMsg->srcId.vgId = 100;
pMsg->destId.addr = syncUtilAddr2U64("127.0.0.1", 5678);
pMsg->destId.vgId = 100;
pMsg->currentTerm = 11;
pMsg->term = 11;
pMsg->lastLogIndex = 22;
pMsg->lastLogTerm = 33;
return pMsg;
@ -100,7 +100,7 @@ SyncAppendEntriesReply *createSyncAppendEntriesReply() {
void test1() {
SyncTimeout *pMsg = createSyncTimeout();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncTimeout2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test1", &rpcMsg);
syncTimeoutDestroy(pMsg);
@ -108,7 +108,7 @@ void test1() {
void test2() {
SyncPing *pMsg = createSyncPing();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncPing2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test2", &rpcMsg);
syncPingDestroy(pMsg);
@ -116,7 +116,7 @@ void test2() {
void test3() {
SyncPingReply *pMsg = createSyncPingReply();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncPingReply2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test3", &rpcMsg);
syncPingReplyDestroy(pMsg);
@ -132,7 +132,7 @@ void test4() {
void test5() {
SyncRequestVoteReply *pMsg = createSyncRequestVoteReply();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncRequestVoteReply2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test5", &rpcMsg);
syncRequestVoteReplyDestroy(pMsg);
@ -140,7 +140,7 @@ void test5() {
void test6() {
SyncAppendEntries *pMsg = createSyncAppendEntries();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncAppendEntries2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test6", &rpcMsg);
syncAppendEntriesDestroy(pMsg);
@ -148,7 +148,7 @@ void test6() {
void test7() {
SyncAppendEntriesReply *pMsg = createSyncAppendEntriesReply();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test7", &rpcMsg);
syncAppendEntriesReplyDestroy(pMsg);
@ -156,7 +156,7 @@ void test7() {
void test8() {
SyncClientRequest *pMsg = createSyncClientRequest();
SRpcMsg rpcMsg;
SRpcMsg rpcMsg;
syncClientRequest2RpcMsg(pMsg, &rpcMsg);
syncRpcMsgPrint2((char *)"test8", &rpcMsg);
syncClientRequestDestroy(pMsg);

View File

@ -26,6 +26,7 @@ int main() {
tsAsyncLog = 0;
sDebugFlag = 143 + 64;
logTest();
electRandomMSTest();
return 0;

View File

@ -120,6 +120,10 @@ typedef struct {
// SEpSet* pSet; // for synchronous API
} SRpcReqContext;
typedef SRpcMsg STransMsg;
typedef SRpcInfo STrans;
typedef SRpcConnInfo STransHandleInfo;
typedef struct {
SEpSet epSet; // ip list provided by app
void* ahandle; // handle provided by app
@ -134,8 +138,8 @@ typedef struct {
int8_t connType; // connection type
int64_t rid; // refId returned by taosAddRef
SRpcMsg* pRsp; // for synchronous API
tsem_t* pSem; // for synchronous API
STransMsg* pRsp; // for synchronous API
tsem_t* pSem; // for synchronous API
int hThrdIdx;
char* ip;
@ -249,4 +253,15 @@ void transUnrefSrvHandle(void* handle);
void transRefCliHandle(void* handle);
void transUnrefCliHandle(void* handle);
void transSendRequest(void* shandle, const char* ip, uint32_t port, STransMsg* pMsg);
void transSendRecv(void* shandle, const char* ip, uint32_t port, STransMsg* pMsg, STransMsg* pRsp);
void transSendResponse(const STransMsg* pMsg);
int transGetConnInfo(void* thandle, STransHandleInfo* pInfo);
void* transInitServer(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle);
void* transInitClient(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle);
void transCloseClient(void* arg);
void transCloseServer(void* arg);
#endif

View File

@ -64,6 +64,7 @@ typedef struct {
void (*cfp)(void* parent, SRpcMsg*, SEpSet*);
int (*afp)(void* parent, char* user, char* spi, char* encrypt, char* secret, char* ckey);
bool (*pfp)(void* parent, tmsg_t msgType);
void* (*mfp)(void* parent, tmsg_t msgType);
int32_t refCount;
void* parent;

View File

@ -18,8 +18,9 @@
#include "transComm.h"
void* (*taosInitHandle[])(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle) = {
taosInitServer, taosInitClient};
void (*taosCloseHandle[])(void* arg) = {taosCloseServer, taosCloseClient};
transInitServer, transInitClient};
void (*taosCloseHandle[])(void* arg) = {transCloseServer, transCloseClient};
void* rpcOpen(const SRpcInit* pInit) {
SRpcInfo* pRpc = calloc(1, sizeof(SRpcInfo));
@ -34,11 +35,12 @@ void* rpcOpen(const SRpcInit* pInit) {
pRpc->cfp = pInit->cfp;
pRpc->afp = pInit->afp;
pRpc->pfp = pInit->pfp;
pRpc->mfp = pInit->mfp;
if (pInit->connType == TAOS_CONN_SERVER) {
pRpc->numOfThreads = pInit->numOfThreads > TSDB_MAX_RPC_THREADS ? TSDB_MAX_RPC_THREADS : pInit->numOfThreads;
} else {
pRpc->numOfThreads = pInit->numOfThreads;
pRpc->numOfThreads = pInit->numOfThreads > TSDB_MAX_RPC_THREADS ? TSDB_MAX_RPC_THREADS : pInit->numOfThreads;
}
pRpc->connType = pInit->connType;
@ -51,7 +53,6 @@ void* rpcOpen(const SRpcInit* pInit) {
if (pInit->secret) {
memcpy(pRpc->secret, pInit->secret, strlen(pInit->secret));
}
return pRpc;
}
void rpcClose(void* arg) {
@ -111,16 +112,19 @@ void rpcSendRedirectRsp(void* thandle, const SEpSet* pEpSet) {
int rpcReportProgress(void* pConn, char* pCont, int contLen) { return -1; }
void rpcCancelRequest(int64_t rid) { return; }
int32_t rpcInit() {
// impl later
return 0;
void rpcSendRequest(void* shandle, const SEpSet* pEpSet, SRpcMsg* pMsg, int64_t* pRid) {
char* ip = (char*)(pEpSet->eps[pEpSet->inUse].fqdn);
uint32_t port = pEpSet->eps[pEpSet->inUse].port;
transSendRequest(shandle, ip, port, pMsg);
}
void rpcSendRecv(void* shandle, SEpSet* pEpSet, SRpcMsg* pMsg, SRpcMsg* pRsp) {
char* ip = (char*)(pEpSet->eps[pEpSet->inUse].fqdn);
uint32_t port = pEpSet->eps[pEpSet->inUse].port;
transSendRecv(shandle, ip, port, pMsg, pRsp);
}
void rpcCleanup(void) {
// impl later
//
return;
}
void rpcSendResponse(const SRpcMsg* pMsg) { transSendResponse(pMsg); }
int rpcGetConnInfo(void* thandle, SRpcConnInfo* pInfo) { return transGetConnInfo((void*)thandle, pInfo); }
void (*taosRefHandle[])(void* handle) = {transRefSrvHandle, transRefCliHandle};
void (*taosUnRefHandle[])(void* handle) = {transUnrefSrvHandle, transUnrefCliHandle};
@ -129,9 +133,19 @@ void rpcRefHandle(void* handle, int8_t type) {
assert(type == TAOS_CONN_SERVER || type == TAOS_CONN_CLIENT);
(*taosRefHandle[type])(handle);
}
void rpcUnrefHandle(void* handle, int8_t type) {
assert(type == TAOS_CONN_SERVER || type == TAOS_CONN_CLIENT);
(*taosUnRefHandle[type])(handle);
}
int32_t rpcInit() {
// impl later
return 0;
}
void rpcCleanup(void) {
// impl later
return;
}
#endif

View File

@ -42,7 +42,7 @@ typedef struct SCliConn {
typedef struct SCliMsg {
STransConnCtx* ctx;
SRpcMsg msg;
STransMsg msg;
queue q;
uint64_t st;
} SCliMsg;
@ -84,7 +84,7 @@ static void addConnToPool(void* pool, char* ip, uint32_t port, SCliConn* co
// register timer in each thread to clear expire conn
static void cliTimeoutCb(uv_timer_t* handle);
// alloc buf for recv
static void cliAllocBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf);
static void cliAllocRecvBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf);
// callback after read nbytes from socket
static void cliRecvCb(uv_stream_t* cli, ssize_t nread, const uv_buf_t* buf);
// callback after write data to socket
@ -105,9 +105,9 @@ static void cliHandleExcept(SCliConn* conn);
static void cliHandleReq(SCliMsg* pMsg, SCliThrdObj* pThrd);
static void cliHandleQuit(SCliMsg* pMsg, SCliThrdObj* pThrd);
static void cliSendQuit(SCliThrdObj* thrd);
static void destroyUserdata(SRpcMsg* userdata);
static void destroyUserdata(STransMsg* userdata);
static int cliRBChoseIdx(SRpcInfo* pTransInst);
static int cliRBChoseIdx(STrans* pTransInst);
static void destroyCmsg(SCliMsg* cmsg);
static void transDestroyConnCtx(STransConnCtx* ctx);
@ -118,11 +118,11 @@ static void destroyThrdObj(SCliThrdObj* pThrd);
#define CONN_HOST_THREAD_INDEX(conn) (conn ? ((SCliConn*)conn)->hThrdIdx : -1)
#define CONN_PERSIST_TIME(para) (para * 1000 * 10)
#define CONN_GET_INST_LABEL(conn) (((SRpcInfo*)(((SCliThrdObj*)conn->hostThrd)->pTransInst))->label)
#define CONN_GET_INST_LABEL(conn) (((STrans*)(((SCliThrdObj*)conn->hostThrd)->pTransInst))->label)
#define CONN_HANDLE_THREAD_QUIT(conn, thrd) \
do { \
if (thrd->quit) { \
cliHandleExcept(conn); \
cliHandleExcept(conn); \
goto _RETURE; \
} \
} while (0)
@ -130,62 +130,76 @@ static void destroyThrdObj(SCliThrdObj* pThrd);
#define CONN_HANDLE_BROKEN(conn) \
do { \
if (conn->broken) { \
cliHandleExcept(conn); \
cliHandleExcept(conn); \
goto _RETURE; \
} \
} while (0);
#define CONN_SET_PERSIST_BY_APP(conn) \
do { \
if (conn->persist == false) { \
conn->persist = true; \
transRefCliHandle(conn); \
} \
} while (0)
#define CONN_NO_PERSIST_BY_APP(conn) ((conn)->persist == false)
static void* cliWorkThread(void* arg);
static void* cliNotifyApp() {}
static void cliHandleResp(SCliConn* conn) {
SCliMsg* pMsg = conn->data;
STransConnCtx* pCtx = pMsg->ctx;
void cliHandleResp(SCliConn* conn) {
SCliThrdObj* pThrd = conn->hostThrd;
SRpcInfo* pTransInst = pThrd->pTransInst;
STrans* pTransInst = pThrd->pTransInst;
STransMsgHead* pHead = (STransMsgHead*)(conn->readBuf.buf);
pHead->code = htonl(pHead->code);
pHead->msgLen = htonl(pHead->msgLen);
// buf's mem alread translated to rpcMsg.pCont
STransMsg transMsg = {0};
transMsg.contLen = transContLenFromMsg(pHead->msgLen);
transMsg.pCont = transContFromHead((char*)pHead);
transMsg.code = pHead->code;
transMsg.msgType = pHead->msgType;
transMsg.ahandle = NULL;
SCliMsg* pMsg = conn->data;
STransConnCtx* pCtx = pMsg ? pMsg->ctx : NULL;
if (pMsg == NULL && !CONN_NO_PERSIST_BY_APP(conn)) {
transMsg.ahandle = pTransInst->mfp ? (*pTransInst->mfp)(pTransInst->parent, transMsg.msgType) : NULL;
} else {
transMsg.ahandle = pCtx ? pCtx->ahandle : NULL;
}
// if (rpcMsg.ahandle == NULL) {
// tDebug("%s cli conn %p handle except", CONN_GET_INST_LABEL(conn), conn);
// return;
//}
// buf's mem alread translated to transMsg.pCont
transClearBuffer(&conn->readBuf);
SRpcMsg rpcMsg = {0};
rpcMsg.contLen = transContLenFromMsg(pHead->msgLen);
rpcMsg.pCont = transContFromHead((char*)pHead);
rpcMsg.code = pHead->code;
rpcMsg.msgType = pHead->msgType;
rpcMsg.ahandle = pCtx->ahandle;
if (pTransInst->pfp != NULL && (pTransInst->pfp)(pTransInst->parent, rpcMsg.msgType)) {
rpcMsg.handle = conn;
transRefCliHandle(conn);
conn->persist = 1;
tDebug("cli conn %p persist by app", conn);
if (pTransInst->pfp != NULL && (*pTransInst->pfp)(pTransInst->parent, transMsg.msgType)) {
transMsg.handle = conn;
CONN_SET_PERSIST_BY_APP(conn);
tDebug("%s cli conn %p ref by app", CONN_GET_INST_LABEL(conn), conn);
}
tDebug("%s cli conn %p %s received from %s:%d, local info: %s:%d, msg size: %d", pTransInst->label, conn,
TMSG_INFO(pHead->msgType), inet_ntoa(conn->addr.sin_addr), ntohs(conn->addr.sin_port),
inet_ntoa(conn->locaddr.sin_addr), ntohs(conn->locaddr.sin_port), rpcMsg.contLen);
inet_ntoa(conn->locaddr.sin_addr), ntohs(conn->locaddr.sin_port), transMsg.contLen);
conn->secured = pHead->secured;
if (pCtx->pSem == NULL) {
if (pCtx == NULL || pCtx->pSem == NULL) {
tTrace("%s cli conn %p handle resp", pTransInst->label, conn);
(pTransInst->cfp)(pTransInst->parent, &rpcMsg, NULL);
(pTransInst->cfp)(pTransInst->parent, &transMsg, NULL);
} else {
tTrace("%s cli conn(sync) %p handle resp", pTransInst->label, conn);
memcpy((char*)pCtx->pRsp, (char*)&rpcMsg, sizeof(rpcMsg));
memcpy((char*)pCtx->pRsp, (char*)&transMsg, sizeof(transMsg));
tsem_post(pCtx->pSem);
}
uv_read_start((uv_stream_t*)conn->stream, cliAllocBufferCb, cliRecvCb);
uv_read_start((uv_stream_t*)conn->stream, cliAllocRecvBufferCb, cliRecvCb);
// user owns conn->persist = 1
if (conn->persist == 0) {
if (CONN_NO_PERSIST_BY_APP(conn)) {
addConnToPool(pThrd->pool, pCtx->ip, pCtx->port, conn);
}
destroyCmsg(conn->data);
@ -196,29 +210,37 @@ static void cliHandleResp(SCliConn* conn) {
// uv_timer_start((uv_timer_t*)&pThrd->timer, cliTimeoutCb, CONN_PERSIST_TIME(pRpc->idleTime) / 2, 0);
}
}
static void cliHandleExcept(SCliConn* pConn) {
void cliHandleExcept(SCliConn* pConn) {
if (pConn->data == NULL) {
// handle conn except in conn pool
transUnrefCliHandle(pConn);
return;
if (pConn->broken == true || CONN_NO_PERSIST_BY_APP(pConn)) {
transUnrefCliHandle(pConn);
return;
}
}
SCliThrdObj* pThrd = pConn->hostThrd;
SRpcInfo* pTransInst = pThrd->pTransInst;
STrans* pTransInst = pThrd->pTransInst;
SCliMsg* pMsg = pConn->data;
STransConnCtx* pCtx = pMsg->ctx;
STransConnCtx* pCtx = pMsg ? pMsg->ctx : NULL;
SRpcMsg rpcMsg = {0};
rpcMsg.ahandle = pCtx->ahandle;
rpcMsg.code = TSDB_CODE_RPC_NETWORK_UNAVAIL;
rpcMsg.msgType = pMsg->msg.msgType + 1;
STransMsg transMsg = {0};
transMsg.code = TSDB_CODE_RPC_NETWORK_UNAVAIL;
transMsg.msgType = pMsg ? pMsg->msg.msgType + 1 : 0;
transMsg.ahandle = NULL;
if (pCtx->pSem == NULL) {
if (pMsg == NULL && !CONN_NO_PERSIST_BY_APP(pConn)) {
transMsg.ahandle = pTransInst->mfp ? (*pTransInst->mfp)(pTransInst->parent, transMsg.msgType) : NULL;
} else {
transMsg.ahandle = pCtx ? pCtx->ahandle : NULL;
}
if (pCtx == NULL || pCtx->pSem == NULL) {
tTrace("%s cli conn %p handle resp", pTransInst->label, pConn);
(pTransInst->cfp)(pTransInst->parent, &rpcMsg, NULL);
(pTransInst->cfp)(pTransInst->parent, &transMsg, NULL);
} else {
tTrace("%s cli conn(sync) %p handle resp", pTransInst->label, pConn);
memcpy((char*)(pCtx->pRsp), (char*)(&rpcMsg), sizeof(rpcMsg));
memcpy((char*)(pCtx->pRsp), (char*)(&transMsg), sizeof(transMsg));
tsem_post(pCtx->pSem);
}
destroyCmsg(pConn->data);
@ -228,11 +250,11 @@ static void cliHandleExcept(SCliConn* pConn) {
transUnrefCliHandle(pConn);
}
static void cliTimeoutCb(uv_timer_t* handle) {
void cliTimeoutCb(uv_timer_t* handle) {
SCliThrdObj* pThrd = handle->data;
SRpcInfo* pRpc = pThrd->pTransInst;
STrans* pTransInst = pThrd->pTransInst;
int64_t currentTime = pThrd->nextTimeout;
tTrace("%s, cli conn timeout, try to remove expire conn from conn pool", pRpc->label);
tTrace("%s, cli conn timeout, try to remove expire conn from conn pool", pTransInst->label);
SConnList* p = taosHashIterate((SHashObj*)pThrd->pool, NULL);
while (p != NULL) {
@ -249,14 +271,15 @@ static void cliTimeoutCb(uv_timer_t* handle) {
p = taosHashIterate((SHashObj*)pThrd->pool, p);
}
pThrd->nextTimeout = taosGetTimestampMs() + CONN_PERSIST_TIME(pRpc->idleTime);
uv_timer_start(handle, cliTimeoutCb, CONN_PERSIST_TIME(pRpc->idleTime) / 2, 0);
pThrd->nextTimeout = taosGetTimestampMs() + CONN_PERSIST_TIME(pTransInst->idleTime);
uv_timer_start(handle, cliTimeoutCb, CONN_PERSIST_TIME(pTransInst->idleTime) / 2, 0);
}
static void* createConnPool(int size) {
void* createConnPool(int size) {
// thread local, no lock
return taosHashInit(size, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
}
static void* destroyConnPool(void* pool) {
void* destroyConnPool(void* pool) {
SConnList* connList = taosHashIterate((SHashObj*)pool, NULL);
while (connList != NULL) {
while (!QUEUE_IS_EMPTY(&connList->conn)) {
@ -301,15 +324,15 @@ static void addConnToPool(void* pool, char* ip, uint32_t port, SCliConn* conn) {
tstrncpy(key + strlen(key), (char*)(&port), sizeof(port));
tTrace("cli conn %p added to conn pool, read buf cap: %d", conn, conn->readBuf.cap);
SRpcInfo* pRpc = ((SCliThrdObj*)conn->hostThrd)->pTransInst;
STrans* pTransInst = ((SCliThrdObj*)conn->hostThrd)->pTransInst;
conn->expireTime = taosGetTimestampMs() + CONN_PERSIST_TIME(pRpc->idleTime);
conn->expireTime = taosGetTimestampMs() + CONN_PERSIST_TIME(pTransInst->idleTime);
SConnList* plist = taosHashGet((SHashObj*)pool, key, strlen(key));
// list already create before
assert(plist != NULL);
QUEUE_PUSH(&plist->conn, &conn->conn);
}
static void cliAllocBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf) {
static void cliAllocRecvBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf) {
SCliConn* conn = handle->data;
SConnBuffer* pBuf = &conn->readBuf;
transAllocBuffer(pBuf, buf);
@ -358,6 +381,7 @@ static SCliConn* cliCreateConn(SCliThrdObj* pThrd) {
QUEUE_INIT(&conn->conn);
conn->hostThrd = pThrd;
conn->persist = false;
conn->broken = false;
transRefCliHandle(conn);
return conn;
@ -392,19 +416,19 @@ static void cliSendCb(uv_write_t* req, int status) {
cliHandleExcept(pConn);
return;
}
uv_read_start((uv_stream_t*)pConn->stream, cliAllocBufferCb, cliRecvCb);
uv_read_start((uv_stream_t*)pConn->stream, cliAllocRecvBufferCb, cliRecvCb);
}
static void cliSend(SCliConn* pConn) {
void cliSend(SCliConn* pConn) {
CONN_HANDLE_BROKEN(pConn);
SCliMsg* pCliMsg = pConn->data;
STransConnCtx* pCtx = pCliMsg->ctx;
SCliThrdObj* pThrd = pConn->hostThrd;
SRpcInfo* pTransInst = pThrd->pTransInst;
STrans* pTransInst = pThrd->pTransInst;
SRpcMsg* pMsg = (SRpcMsg*)(&pCliMsg->msg);
STransMsg* pMsg = (STransMsg*)(&pCliMsg->msg);
STransMsgHead* pHead = transHeadFromCont(pMsg->pCont);
int msgLen = transMsgLenFromCont(pMsg->contLen);
@ -442,7 +466,8 @@ static void cliSend(SCliConn* pConn) {
_RETURE:
return;
}
static void cliConnCb(uv_connect_t* req, int status) {
void cliConnCb(uv_connect_t* req, int status) {
// impl later
SCliConn* pConn = req->data;
if (status != 0) {
@ -472,11 +497,11 @@ static void cliHandleQuit(SCliMsg* pMsg, SCliThrdObj* pThrd) {
pThrd->quit = true;
uv_stop(pThrd->loop);
}
static SCliConn* cliGetConn(SCliMsg* pMsg, SCliThrdObj* pThrd) {
SCliConn* cliGetConn(SCliMsg* pMsg, SCliThrdObj* pThrd) {
SCliConn* conn = NULL;
if (pMsg->msg.handle != NULL) {
conn = (SCliConn*)(pMsg->msg.handle);
transUnrefCliHandle(conn);
if (conn != NULL) {
tTrace("%s cli conn %p reused", CONN_GET_INST_LABEL(conn), conn);
}
@ -487,13 +512,14 @@ static SCliConn* cliGetConn(SCliMsg* pMsg, SCliThrdObj* pThrd) {
}
return conn;
}
static void cliHandleReq(SCliMsg* pMsg, SCliThrdObj* pThrd) {
void cliHandleReq(SCliMsg* pMsg, SCliThrdObj* pThrd) {
uint64_t et = taosGetTimestampUs();
uint64_t el = et - pMsg->st;
tTrace("%s cli msg tran time cost: %" PRIu64 "us", ((SRpcInfo*)pThrd->pTransInst)->label, el);
tTrace("%s cli msg tran time cost: %" PRIu64 "us", ((STrans*)pThrd->pTransInst)->label, el);
STransConnCtx* pCtx = pMsg->ctx;
SRpcInfo* pTransInst = pThrd->pTransInst;
STrans* pTransInst = pThrd->pTransInst;
SCliConn* conn = cliGetConn(pMsg, pThrd);
if (conn != NULL) {
@ -514,6 +540,7 @@ static void cliHandleReq(SCliMsg* pMsg, SCliThrdObj* pThrd) {
tTrace("%s cli conn %p try to connect to %s:%d", pTransInst->label, conn, pMsg->ctx->ip, pMsg->ctx->port);
uv_tcp_connect(&conn->connReq, (uv_tcp_t*)(conn->stream), (const struct sockaddr*)&addr, cliConnCb);
}
conn->hThrdIdx = pCtx->hThrdIdx;
}
static void cliAsyncCb(uv_async_t* handle) {
@ -522,7 +549,7 @@ static void cliAsyncCb(uv_async_t* handle) {
SCliMsg* pMsg = NULL;
// batch process to avoid to lock/unlock frequently
queue wq;
queue wq;
pthread_mutex_lock(&item->mtx);
QUEUE_MOVE(&item->qmsg, &wq);
pthread_mutex_unlock(&item->mtx);
@ -551,17 +578,17 @@ static void* cliWorkThread(void* arg) {
uv_run(pThrd->loop, UV_RUN_DEFAULT);
}
void* taosInitClient(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle) {
void* transInitClient(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle) {
SCliObj* cli = calloc(1, sizeof(SCliObj));
SRpcInfo* pRpc = shandle;
STrans* pTransInst = shandle;
memcpy(cli->label, label, strlen(label));
cli->numOfThreads = numOfThreads;
cli->pThreadObj = (SCliThrdObj**)calloc(cli->numOfThreads, sizeof(SCliThrdObj*));
for (int i = 0; i < cli->numOfThreads; i++) {
SCliThrdObj* pThrd = createThrdObj();
pThrd->nextTimeout = taosGetTimestampMs() + CONN_PERSIST_TIME(pRpc->idleTime);
pThrd->nextTimeout = taosGetTimestampMs() + CONN_PERSIST_TIME(pTransInst->idleTime);
pThrd->pTransInst = shandle;
int err = pthread_create(&pThrd->thread, NULL, cliWorkThread, (void*)(pThrd));
@ -573,7 +600,7 @@ void* taosInitClient(uint32_t ip, uint32_t port, char* label, int numOfThreads,
return cli;
}
static void destroyUserdata(SRpcMsg* userdata) {
static void destroyUserdata(STransMsg* userdata) {
if (userdata->pCont == NULL) {
return;
}
@ -629,12 +656,20 @@ static void transDestroyConnCtx(STransConnCtx* ctx) {
free(ctx);
}
//
static void cliSendQuit(SCliThrdObj* thrd) {
void cliSendQuit(SCliThrdObj* thrd) {
// cli can stop gracefully
SCliMsg* msg = calloc(1, sizeof(SCliMsg));
transSendAsync(thrd->asyncPool, &msg->q);
}
void taosCloseClient(void* arg) {
int cliRBChoseIdx(STrans* pTransInst) {
int64_t index = pTransInst->index;
if (pTransInst->index++ >= pTransInst->numOfThreads) {
pTransInst->index = 0;
}
return index % pTransInst->numOfThreads;
}
void transCloseClient(void* arg) {
SCliObj* cli = arg;
for (int i = 0; i < cli->numOfThreads; i++) {
cliSendQuit(cli->pThreadObj[i]);
@ -643,13 +678,6 @@ void taosCloseClient(void* arg) {
free(cli->pThreadObj);
free(cli);
}
static int cliRBChoseIdx(SRpcInfo* pTransInst) {
int64_t index = pTransInst->index;
if (pTransInst->index++ >= pTransInst->numOfThreads) {
pTransInst->index = 0;
}
return index % pTransInst->numOfThreads;
}
void transRefCliHandle(void* handle) {
if (handle == NULL) {
return;
@ -665,17 +693,11 @@ void transUnrefCliHandle(void* handle) {
if (ref == 0) {
cliDestroyConn((SCliConn*)handle, true);
}
// unref cli handle
}
void rpcSendRequest(void* shandle, const SEpSet* pEpSet, SRpcMsg* pMsg, int64_t* pRid) {
// impl later
char* ip = (char*)(pEpSet->eps[pEpSet->inUse].fqdn);
uint32_t port = pEpSet->eps[pEpSet->inUse].port;
SRpcInfo* pTransInst = (SRpcInfo*)shandle;
int index = CONN_HOST_THREAD_INDEX(pMsg->handle);
void transSendRequest(void* shandle, const char* ip, uint32_t port, STransMsg* pMsg) {
STrans* pTransInst = (STrans*)shandle;
int index = CONN_HOST_THREAD_INDEX((SCliConn*)pMsg->handle);
if (index == -1) {
index = cliRBChoseIdx(pTransInst);
}
@ -683,6 +705,7 @@ void rpcSendRequest(void* shandle, const SEpSet* pEpSet, SRpcMsg* pMsg, int64_t*
if (transCompressMsg(pMsg->pCont, pMsg->contLen, &flen)) {
// imp later
}
tDebug("send request at thread:%d %p", index, pMsg);
STransConnCtx* pCtx = calloc(1, sizeof(STransConnCtx));
pCtx->ahandle = pMsg->ahandle;
pCtx->msgType = pMsg->msgType;
@ -701,14 +724,9 @@ void rpcSendRequest(void* shandle, const SEpSet* pEpSet, SRpcMsg* pMsg, int64_t*
SCliThrdObj* thrd = ((SCliObj*)pTransInst->tcphandle)->pThreadObj[index];
transSendAsync(thrd->asyncPool, &(cliMsg->q));
}
void rpcSendRecv(void* shandle, SEpSet* pEpSet, SRpcMsg* pReq, SRpcMsg* pRsp) {
char* ip = (char*)(pEpSet->eps[pEpSet->inUse].fqdn);
uint32_t port = pEpSet->eps[pEpSet->inUse].port;
SRpcInfo* pTransInst = (SRpcInfo*)shandle;
int index = CONN_HOST_THREAD_INDEX(pReq->handle);
void transSendRecv(void* shandle, const char* ip, uint32_t port, STransMsg* pReq, STransMsg* pRsp) {
STrans* pTransInst = (STrans*)shandle;
int index = CONN_HOST_THREAD_INDEX(pReq->handle);
if (index == -1) {
index = cliRBChoseIdx(pTransInst);
}
@ -734,7 +752,6 @@ void rpcSendRecv(void* shandle, SEpSet* pEpSet, SRpcMsg* pReq, SRpcMsg* pRsp) {
tsem_wait(pSem);
tsem_destroy(pSem);
free(pSem);
return;
}
#endif

View File

@ -16,20 +16,6 @@
#include "transComm.h"
int rpcAuthenticateMsg(void* pMsg, int msgLen, void* pAuth, void* pKey) {
T_MD5_CTX context;
int ret = -1;
tMD5Init(&context);
tMD5Update(&context, (uint8_t*)pKey, TSDB_PASSWORD_LEN);
tMD5Update(&context, (uint8_t*)pMsg, msgLen);
tMD5Update(&context, (uint8_t*)pKey, TSDB_PASSWORD_LEN);
tMD5Final(&context);
if (memcmp(context.digest, pAuth, sizeof(context.digest)) == 0) ret = 0;
return ret;
}
int transAuthenticateMsg(void* pMsg, int msgLen, void* pAuth, void* pKey) {
T_MD5_CTX context;
int ret = -1;
@ -44,17 +30,7 @@ int transAuthenticateMsg(void* pMsg, int msgLen, void* pAuth, void* pKey) {
return ret;
}
void rpcBuildAuthHead(void* pMsg, int msgLen, void* pAuth, void* pKey) {
T_MD5_CTX context;
tMD5Init(&context);
tMD5Update(&context, (uint8_t*)pKey, TSDB_PASSWORD_LEN);
tMD5Update(&context, (uint8_t*)pMsg, msgLen);
tMD5Update(&context, (uint8_t*)pKey, TSDB_PASSWORD_LEN);
tMD5Final(&context);
memcpy(pAuth, context.digest, sizeof(context.digest));
}
void transBuildAuthHead(void* pMsg, int msgLen, void* pAuth, void* pKey) {
T_MD5_CTX context;
@ -67,45 +43,6 @@ void transBuildAuthHead(void* pMsg, int msgLen, void* pAuth, void* pKey) {
memcpy(pAuth, context.digest, sizeof(context.digest));
}
int32_t rpcCompressRpcMsg(char* pCont, int32_t contLen) {
SRpcHead* pHead = rpcHeadFromCont(pCont);
int32_t finalLen = 0;
int overhead = sizeof(SRpcComp);
if (!NEEDTO_COMPRESSS_MSG(contLen)) {
return contLen;
}
char* buf = malloc(contLen + overhead + 8); // 8 extra bytes
if (buf == NULL) {
tError("failed to allocate memory for rpc msg compression, contLen:%d", contLen);
return contLen;
}
int32_t compLen = LZ4_compress_default(pCont, buf, contLen, contLen + overhead);
tDebug("compress rpc msg, before:%d, after:%d, overhead:%d", contLen, compLen, overhead);
/*
* only the compressed size is less than the value of contLen - overhead, the compression is applied
* The first four bytes is set to 0, the second four bytes are utilized to keep the original length of message
*/
if (compLen > 0 && compLen < contLen - overhead) {
SRpcComp* pComp = (SRpcComp*)pCont;
pComp->reserved = 0;
pComp->contLen = htonl(contLen);
memcpy(pCont + overhead, buf, compLen);
pHead->comp = 1;
tDebug("compress rpc msg, before:%d, after:%d", contLen, compLen);
finalLen = compLen + overhead;
} else {
finalLen = contLen;
}
free(buf);
return finalLen;
}
bool transCompressMsg(char* msg, int32_t len, int32_t* flen) {
return false;
// SRpcHead* pHead = rpcHeadFromCont(pCont);
@ -154,39 +91,6 @@ bool transDecompressMsg(char* msg, int32_t len, int32_t* flen) {
return false;
}
SRpcHead* rpcDecompressRpcMsg(SRpcHead* pHead) {
int overhead = sizeof(SRpcComp);
SRpcHead* pNewHead = NULL;
uint8_t* pCont = pHead->content;
SRpcComp* pComp = (SRpcComp*)pHead->content;
if (pHead->comp) {
// decompress the content
assert(pComp->reserved == 0);
int contLen = htonl(pComp->contLen);
// prepare the temporary buffer to decompress message
char* temp = (char*)malloc(contLen + RPC_MSG_OVERHEAD);
pNewHead = (SRpcHead*)(temp + sizeof(SRpcReqContext)); // reserve SRpcReqContext
if (pNewHead) {
int compLen = rpcContLenFromMsg(pHead->msgLen) - overhead;
int origLen = LZ4_decompress_safe((char*)(pCont + overhead), (char*)pNewHead->content, compLen, contLen);
assert(origLen == contLen);
memcpy(pNewHead, pHead, sizeof(SRpcHead));
pNewHead->msgLen = rpcMsgLenFromCont(origLen);
/// rpcFreeMsg(pHead); // free the compressed message buffer
pHead = pNewHead;
tTrace("decomp malloc mem:%p", temp);
} else {
tError("failed to allocate memory to decompress msg, contLen:%d", contLen);
}
}
return pHead;
}
void transConnCtxDestroy(STransConnCtx* ctx) {
free(ctx->ip);
free(ctx);
@ -315,7 +219,7 @@ int transSendAsync(SAsyncPool* pool, queue* q) {
if (el > 50) {
// tInfo("lock and unlock cost: %d", (int)el);
}
return uv_async_send(async);
}
#endif

View File

@ -37,8 +37,7 @@ typedef struct SSrvConn {
struct sockaddr_in addr;
struct sockaddr_in locaddr;
// SRpcMsg sendMsg;
// del later
char secured;
int spi;
char info[64];
@ -49,7 +48,7 @@ typedef struct SSrvConn {
typedef struct SSrvMsg {
SSrvConn* pConn;
SRpcMsg msg;
STransMsg msg;
queue q;
} SSrvMsg;
@ -59,12 +58,13 @@ typedef struct SWorkThrdObj {
uv_os_fd_t fd;
uv_loop_t* loop;
SAsyncPool* asyncPool;
// uv_async_t* workerAsync; //
queue msg;
queue conn;
pthread_mutex_t msgMtx;
void* pTransInst;
bool quit;
queue conn;
void* pTransInst;
bool quit;
} SWorkThrdObj;
typedef struct SServerObj {
@ -91,7 +91,7 @@ static int transAddAuthPart(SSrvConn* pConn, char* msg, int msgLen);
static int uvAuthMsg(SSrvConn* pConn, char* msg, int msgLen);
static void uvAllocConnBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf);
static void uvAllocReadBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf);
static void uvAllocRecvBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf);
static void uvOnRecvCb(uv_stream_t* cli, ssize_t nread, const uv_buf_t* buf);
static void uvOnTimeoutCb(uv_timer_t* handle);
static void uvOnSendCb(uv_write_t* req, int status);
@ -121,7 +121,7 @@ static void* acceptThread(void* arg);
static bool addHandleToWorkloop(void* arg);
static bool addHandleToAcceptloop(void* arg);
void uvAllocReadBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf) {
void uvAllocRecvBufferCb(uv_handle_t* handle, size_t suggested_size, uv_buf_t* buf) {
SSrvConn* conn = handle->data;
SConnBuffer* pBuf = &conn->readBuf;
transAllocBuffer(pBuf, buf);
@ -163,7 +163,7 @@ static int uvAuthMsg(SSrvConn* pConn, char* msg, int len) {
tWarn("%s, time diff:%d is too big, msg discarded", pConn->info, delta);
code = TSDB_CODE_RPC_INVALID_TIME_STAMP;
} else {
if (rpcAuthenticateMsg(pHead, len - TSDB_AUTH_LEN, pDigest->auth, pConn->secret) < 0) {
if (transAuthenticateMsg(pHead, len - TSDB_AUTH_LEN, pDigest->auth, pConn->secret) < 0) {
// tDebug("%s, authentication failed, msg discarded", pConn->info);
code = TSDB_CODE_RPC_AUTH_FAILURE;
} else {
@ -204,37 +204,36 @@ static void uvHandleReq(SSrvConn* pConn) {
memcpy(pConn->user, uMsg->user, tListLen(uMsg->user));
memcpy(pConn->secret, uMsg->secret, tListLen(uMsg->secret));
}
pConn->inType = pHead->msgType;
SRpcInfo* pRpc = (SRpcInfo*)p->shandle;
pHead->code = htonl(pHead->code);
int32_t dlen = 0;
if (transDecompressMsg(NULL, 0, NULL)) {
// add compress later
// pHead = rpcDecompressRpcMsg(pHead);
// pHead = rpcDecompresSTransMsg(pHead);
} else {
pHead->msgLen = htonl(pHead->msgLen);
// impl later
//
}
SRpcMsg rpcMsg;
rpcMsg.contLen = transContLenFromMsg(pHead->msgLen);
rpcMsg.pCont = pHead->content;
rpcMsg.msgType = pHead->msgType;
rpcMsg.code = pHead->code;
rpcMsg.ahandle = NULL;
rpcMsg.handle = pConn;
STransMsg transMsg;
transMsg.contLen = transContLenFromMsg(pHead->msgLen);
transMsg.pCont = pHead->content;
transMsg.msgType = pHead->msgType;
transMsg.code = pHead->code;
transMsg.ahandle = NULL;
transMsg.handle = pConn;
transClearBuffer(&pConn->readBuf);
pConn->inType = pHead->msgType;
transRefSrvHandle(pConn);
tDebug("server conn %p %s received from %s:%d, local info: %s:%d, msg size: %d", pConn, TMSG_INFO(rpcMsg.msgType),
tDebug("server conn %p %s received from %s:%d, local info: %s:%d, msg size: %d", pConn, TMSG_INFO(transMsg.msgType),
inet_ntoa(pConn->addr.sin_addr), ntohs(pConn->addr.sin_port), inet_ntoa(pConn->locaddr.sin_addr),
ntohs(pConn->locaddr.sin_port), rpcMsg.contLen);
(*(pRpc->cfp))(pRpc->parent, &rpcMsg, NULL);
ntohs(pConn->locaddr.sin_port), transMsg.contLen);
STrans* pTransInst = (STrans*)p->shandle;
(*((STrans*)p->shandle)->cfp)(pTransInst->parent, &transMsg, NULL);
// uv_timer_start(&pConn->pTimer, uvHandleActivityTimeout, pRpc->idleTime * 10000, 0);
// auth
// validate msg type
@ -260,7 +259,7 @@ void uvOnRecvCb(uv_stream_t* cli, ssize_t nread, const uv_buf_t* buf) {
}
tError("server conn %p read error: %s", conn, uv_err_name(nread));
if (nread < 0 || nread == UV_EOF) {
if (nread < 0) {
conn->broken = true;
transUnrefSrvHandle(conn);
@ -318,8 +317,8 @@ static void uvPrepareSendData(SSrvMsg* smsg, uv_buf_t* wb) {
// impl later;
tTrace("server conn %p prepare to send resp", smsg->pConn);
SSrvConn* pConn = smsg->pConn;
SRpcMsg* pMsg = &smsg->msg;
SSrvConn* pConn = smsg->pConn;
STransMsg* pMsg = &smsg->msg;
if (pMsg->pCont == 0) {
pMsg->pCont = (void*)rpcMallocCont(0);
pMsg->contLen = 0;
@ -526,7 +525,7 @@ void uvOnConnectionCb(uv_stream_t* q, ssize_t nread, const uv_buf_t* buf) {
return;
}
uv_read_start((uv_stream_t*)(pConn->pTcp), uvAllocReadBufferCb, uvOnRecvCb);
uv_read_start((uv_stream_t*)(pConn->pTcp), uvAllocRecvBufferCb, uvOnRecvCb);
} else {
tDebug("failed to create new connection");
@ -547,7 +546,6 @@ static bool addHandleToWorkloop(void* arg) {
return false;
}
// SRpcInfo* pRpc = pThrd->shandle;
uv_pipe_init(pThrd->loop, pThrd->pipe, 1);
uv_pipe_open(pThrd->pipe, pThrd->fd);
@ -668,7 +666,7 @@ static int transAddAuthPart(SSrvConn* pConn, char* msg, int msgLen) {
return msgLen;
}
void* taosInitServer(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle) {
void* transInitServer(uint32_t ip, uint32_t port, char* label, int numOfThreads, void* fp, void* shandle) {
SServerObj* srv = calloc(1, sizeof(SServerObj));
srv->loop = (uv_loop_t*)malloc(sizeof(uv_loop_t));
srv->numOfThreads = numOfThreads;
@ -720,7 +718,7 @@ void* taosInitServer(uint32_t ip, uint32_t port, char* label, int numOfThreads,
return srv;
End:
taosCloseServer(srv);
transCloseServer(srv);
return NULL;
}
@ -740,7 +738,7 @@ void sendQuitToWorkThrd(SWorkThrdObj* pThrd) {
transSendAsync(pThrd->asyncPool, &srvMsg->q);
}
void taosCloseServer(void* arg) {
void transCloseServer(void* arg) {
// impl later
SServerObj* srv = arg;
for (int i = 0; i < srv->numOfThreads; i++) {
@ -786,7 +784,7 @@ void transUnrefSrvHandle(void* handle) {
}
// unref srv handle
}
void rpcSendResponse(const SRpcMsg* pMsg) {
void transSendResponse(const STransMsg* pMsg) {
if (pMsg->handle == NULL) {
return;
}
@ -799,14 +797,12 @@ void rpcSendResponse(const SRpcMsg* pMsg) {
tTrace("server conn %p start to send resp", pConn);
transSendAsync(pThrd->asyncPool, &srvMsg->q);
}
int rpcGetConnInfo(void* thandle, SRpcConnInfo* pInfo) {
SSrvConn* pConn = thandle;
int transGetConnInfo(void* thandle, STransHandleInfo* pInfo) {
SSrvConn* pConn = thandle;
struct sockaddr_in addr = pConn->addr;
pInfo->clientIp = (uint32_t)(addr.sin_addr.s_addr);
pInfo->clientPort = ntohs(addr.sin_port);
tstrncpy(pInfo->user, pConn->user, sizeof(pInfo->user));
return 0;
}

View File

@ -29,24 +29,25 @@ const char *ckey = "ckey";
class Server;
int port = 7000;
// server process
typedef void (*CB)(void *parent, SRpcMsg *pMsg, SEpSet *pEpSet);
static void processReq(void *parent, SRpcMsg *pMsg, SEpSet *pEpSet);
// client process;
static void processResp(void *parent, SRpcMsg *pMsg, SEpSet *pEpSet);
class Client {
public:
void Init(int nThread) {
memset(&rpcInit, 0, sizeof(rpcInit));
rpcInit.localPort = 0;
rpcInit.label = (char *)label;
rpcInit.numOfThreads = nThread;
rpcInit.cfp = processResp;
rpcInit.user = (char *)user;
rpcInit.secret = (char *)secret;
rpcInit.ckey = (char *)ckey;
rpcInit.spi = 1;
rpcInit.parent = this;
rpcInit.connType = TAOS_CONN_CLIENT;
this->transCli = rpcOpen(&rpcInit);
memset(&rpcInit_, 0, sizeof(rpcInit_));
rpcInit_.localPort = 0;
rpcInit_.label = (char *)label;
rpcInit_.numOfThreads = nThread;
rpcInit_.cfp = processResp;
rpcInit_.user = (char *)user;
rpcInit_.secret = (char *)secret;
rpcInit_.ckey = (char *)ckey;
rpcInit_.spi = 1;
rpcInit_.parent = this;
rpcInit_.connType = TAOS_CONN_CLIENT;
this->transCli = rpcOpen(&rpcInit_);
tsem_init(&this->sem, 0, 0);
}
void SetResp(SRpcMsg *pMsg) {
@ -55,9 +56,27 @@ class Client {
}
SRpcMsg *Resp() { return &this->resp; }
void Restart() {
void Restart(CB cb) {
rpcClose(this->transCli);
this->transCli = rpcOpen(&rpcInit);
rpcInit_.cfp = cb;
this->transCli = rpcOpen(&rpcInit_);
}
void setPersistFP(bool (*pfp)(void *parent, tmsg_t msgType)) {
rpcClose(this->transCli);
rpcInit_.pfp = pfp;
this->transCli = rpcOpen(&rpcInit_);
}
void setConstructFP(void *(*mfp)(void *parent, tmsg_t msgType)) {
rpcClose(this->transCli);
rpcInit_.mfp = mfp;
this->transCli = rpcOpen(&rpcInit_);
}
void setPAndMFp(bool (*pfp)(void *parent, tmsg_t msgType), void *(*mfp)(void *parent, tmsg_t msgType)) {
rpcClose(this->transCli);
rpcInit_.pfp = pfp;
rpcInit_.mfp = mfp;
this->transCli = rpcOpen(&rpcInit_);
}
void SendAndRecv(SRpcMsg *req, SRpcMsg *resp) {
@ -79,7 +98,7 @@ class Client {
private:
tsem_t sem;
SRpcInit rpcInit;
SRpcInit rpcInit_;
void * transCli;
SRpcMsg resp;
};
@ -133,39 +152,56 @@ static void processResp(void *parent, SRpcMsg *pMsg, SEpSet *pEpSet) {
client->SetResp(pMsg);
client->SemPost();
}
static void initEnv() {
dDebugFlag = 143;
vDebugFlag = 0;
mDebugFlag = 143;
cDebugFlag = 0;
jniDebugFlag = 0;
tmrDebugFlag = 143;
uDebugFlag = 143;
rpcDebugFlag = 143;
qDebugFlag = 0;
wDebugFlag = 0;
sDebugFlag = 0;
tsdbDebugFlag = 0;
tsLogEmbedded = 1;
tsAsyncLog = 0;
std::string path = "/tmp/transport";
taosRemoveDir(path.c_str());
taosMkDir(path.c_str());
tstrncpy(tsLogDir, path.c_str(), PATH_MAX);
if (taosInitLog("taosdlog", 1) != 0) {
printf("failed to init log file\n");
}
}
class TransObj {
public:
TransObj() {
dDebugFlag = 143;
vDebugFlag = 0;
mDebugFlag = 143;
cDebugFlag = 0;
jniDebugFlag = 0;
tmrDebugFlag = 143;
uDebugFlag = 143;
rpcDebugFlag = 143;
qDebugFlag = 0;
wDebugFlag = 0;
sDebugFlag = 0;
tsdbDebugFlag = 0;
tsLogEmbedded = 1;
tsAsyncLog = 0;
std::string path = "/tmp/transport";
taosRemoveDir(path.c_str());
taosMkDir(path.c_str());
tstrncpy(tsLogDir, path.c_str(), PATH_MAX);
if (taosInitLog("taosdlog", 1) != 0) {
printf("failed to init log file\n");
}
initEnv();
cli = new Client;
cli->Init(1);
srv = new Server;
srv->Start();
}
void RestartCli() { cli->Restart(); }
void RestartCli(CB cb) { cli->Restart(cb); }
void StopSrv() { srv->Stop(); }
void SetCliPersistFp(bool (*pfp)(void *parent, tmsg_t msgType)) {
// do nothing
cli->setPersistFP(pfp);
}
void SetCliMFp(void *(*mfp)(void *parent, tmsg_t msgType)) {
// do nothing
cli->setConstructFP(mfp);
}
void SetMAndPFp(bool (*pfp)(void *parent, tmsg_t msgType), void *(*mfp)(void *parent, tmsg_t msgType)) {
// do nothing
cli->setPAndMFp(pfp, mfp);
}
void RestartSrv() { srv->Restart(); }
void cliSendAndRecv(SRpcMsg *req, SRpcMsg *resp) { cli->SendAndRecv(req, resp); }
~TransObj() {
@ -191,16 +227,16 @@ class TransEnv : public ::testing::Test {
TransObj *tr = NULL;
};
// TEST_F(TransEnv, 01sendAndRec) {
// for (int i = 0; i < 1; i++) {
// SRpcMsg req = {0}, resp = {0};
// req.msgType = 0;
// req.pCont = rpcMallocCont(10);
// req.contLen = 10;
// tr->cliSendAndRecv(&req, &resp);
// assert(resp.code == 0);
// }
//}
TEST_F(TransEnv, 01sendAndRec) {
for (int i = 0; i < 1; i++) {
SRpcMsg req = {0}, resp = {0};
req.msgType = 0;
req.pCont = rpcMallocCont(10);
req.contLen = 10;
tr->cliSendAndRecv(&req, &resp);
assert(resp.code == 0);
}
}
TEST_F(TransEnv, 02StopServer) {
for (int i = 0; i < 1; i++) {
@ -218,6 +254,31 @@ TEST_F(TransEnv, 02StopServer) {
tr->StopSrv();
// tr->RestartSrv();
tr->cliSendAndRecv(&req, &resp);
assert(resp.code != 0);
}
TEST_F(TransEnv, clientUserDefined) {}
TEST_F(TransEnv, cliPersistHandle) {
// impl late
}
TEST_F(TransEnv, srvPersistHandle) {
// impl later
}
TEST_F(TransEnv, srvPersisHandleExcept) {
// conn breken
//
}
TEST_F(TransEnv, cliPersisHandleExcept) {
// conn breken
}
TEST_F(TransEnv, multiCliPersisHandleExcept) {
// conn breken
}
TEST_F(TransEnv, multiSrvPersisHandleExcept) {
// conn breken
}
TEST_F(TransEnv, queryExcept) {
// query and conn is broken
}

View File

@ -13,6 +13,7 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define ALLOW_FORBID_FUNC
#define _DEFAULT_SOURCE
#include "osLocale.h"

View File

@ -349,6 +349,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_TDB_NO_AVAIL_DISK, "No available disk")
TAOS_DEFINE_ERROR(TSDB_CODE_TDB_MESSED_MSG, "TSDB messed message")
TAOS_DEFINE_ERROR(TSDB_CODE_TDB_IVLD_TAG_VAL, "TSDB invalid tag value")
TAOS_DEFINE_ERROR(TSDB_CODE_TDB_NO_CACHE_LAST_ROW, "TSDB no cache last row data")
TAOS_DEFINE_ERROR(TSDB_CODE_TDB_NO_SMA_INDEX_IN_META, "No sma index in meta")
// query
TAOS_DEFINE_ERROR(TSDB_CODE_QRY_INVALID_QHANDLE, "Invalid handle")

59
source/util/src/tuuid.c Normal file
View File

@ -0,0 +1,59 @@
/*
* 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/>.
*/
#include "tuuid.h"
static int64_t hashId = 0;
static int32_t SerialNo = 0;
int32_t tGenIdPI32(void) {
if (hashId == 0) {
char uid[64];
int32_t code = taosGetSystemUUID(uid, tListLen(uid));
if (code != TSDB_CODE_SUCCESS) {
terrno = TAOS_SYSTEM_ERROR(errno);
return -1;
} else {
hashId = MurmurHash3_32(uid, strlen(uid));
}
}
int64_t ts = taosGetTimestampMs();
uint64_t pid = taosGetPId();
int32_t val = atomic_add_fetch_32(&SerialNo, 1);
int32_t id = ((hashId & 0x1F) << 26) | ((pid & 0x3F) << 20) | ((ts & 0xFFF) << 8) | (val & 0xFF);
return id;
}
int64_t tGenIdPI64(void) {
if (hashId == 0) {
char uid[64];
int32_t code = taosGetSystemUUID(uid, tListLen(uid));
if (code != TSDB_CODE_SUCCESS) {
terrno = TAOS_SYSTEM_ERROR(errno);
return -1;
} else {
hashId = MurmurHash3_32(uid, strlen(uid));
}
}
int64_t ts = taosGetTimestampMs();
uint64_t pid = taosGetPId();
int32_t val = atomic_add_fetch_32(&SerialNo, 1);
int64_t id = ((hashId & 0x07FF) << 52) | ((pid & 0x0FFF) << 40) | ((ts & 0xFFFFFF) << 16) | (val & 0xFFFF);
return id;
}

View File

@ -188,7 +188,7 @@ void tFWorkerFreeQueue(SFWorkerPool *pool, STaosQueue *queue) { tQWorkerFreeQueu
int32_t tWWorkerInit(SWWorkerPool *pool) {
pool->nextId = 0;
pool->workers = calloc(sizeof(SWWorker), pool->max);
pool->workers = calloc(pool->max, sizeof(SWWorker));
if (pool->workers == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;

View File

@ -16,4 +16,6 @@
# ---- dnode
./test.sh -f tsim/dnode/basic1.sim
# ---- insert
./test.sh -f tsim/insert/basic0.sim
#======================b1-end===============

View File

@ -14,13 +14,19 @@ $st = $stPrefix . $i
$tb = $tbPrefix . $i
print =============== step1
sql create database $db replica 1 days 20 keep 2000 cache 16 vgroups 4
# quorum presicion
sql create database $db vgroups 8 replica 1 days 20 keep 3650 cache 32 blocks 12 minrows 80 maxrows 10000 wal 2 fsync 1000 comp 0 cachelast 2 precision 'us'
sql show databases
print $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07
print $data0_1 $data1_1 $data2_1 $data3_1 $data4_1 $data5_1 $data6_1 $data7_1 $data8_1 $data9_1
print $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
if $rows != 1 then
return -1
endi
if $data00 != $db then
return -1
endi
if $data02 != 4 then
if $data02 != 8 then
return -1
endi
if $data03 != 0 then
@ -32,16 +38,23 @@ endi
if $data06 != 20 then
return -1
endi
if $data08 != 16 then
if $data07 != 3650,3650,3650 then
return -1
endi
if $data08 != 32 then
return -1
endi
if $data09 != 12 then
return -1
endi
print =============== step2
sql create database $db
sql_error create database $db
sql create database if not exists $db
sql show databases
if $rows != 1 then
return -1
endi
endi
print =============== step3
sql drop database $db

View File

@ -0,0 +1,334 @@
system sh/stop_dnodes.sh
system sh/deploy.sh -n dnode1 -i 1
system sh/exec.sh -n dnode1 -s start
sleep 50
sql connect
print =============== create database
sql create database d0
sql show databases
if $rows != 1 then
return -1
endi
print $data00 $data01 $data02
sql use d0
print =============== create super table, include column type for count/sum/min/max/first
sql create table if not exists stb (ts timestamp, c1 int, c2 float, c3 double) tags (t1 int unsigned)
sql show stables
if $rows != 1 then
return -1
endi
print =============== create child table
sql create table ct1 using stb tags(1000)
sql create table ct2 using stb tags(2000)
sql show tables
if $rows != 2 then
return -1
endi
print =============== insert data, mode1: one row one table in sql
print =============== insert data, mode1: mulit rows one table in sql
print =============== insert data, mode1: one rows mulit table in sql
print =============== insert data, mode1: mulit rows mulit table in sql
sql insert into ct1 values(now+0s, 10, 2.0, 3.0)
sql insert into ct1 values(now+1s, 11, 2.1, 3.1)(now+2s, 12, 2.2, 3.2)(now+3s, 13, 2.3, 3.3)
sql insert into ct2 values(now+0s, 10, 2.0, 3.0)
sql insert into ct2 values(now+1s, 11, 2.1, 3.1)(now+2s, 12, 2.2, 3.2)(now+3s, 13, 2.3, 3.3)
# after fix bug, modify sql_error to sql
sql_error insert into ct1 values(now+4s, -14, -2.4, -3.4) ct2 values(now+4s, -14, -2.4, -3.4)
sql_error insert into ct1 values(now+5s, -15, -2.5, -3.5)(now+6s, -16, -2.6, -3.6) ct2 values(now+5s, -15, -2.5, -3.5)(now+6s, -16, -2.6, -3.6)
#===================================================================
#===================================================================
print =============== query data from child table
sql select * from ct1
if $rows != 4 then # after fix bug, modify 4 to 7
return -1
endi
if $data01 != 10 then
return -1
endi
if $data02 != 2.00000 then
return -1
endi
if $data03 != 3.000000000 then
return -1
endi
#if $data41 != -14 then
# return -1
#endi
#if $data42 != -2.40000 then
# return -1
#endi
#if $data43 != -3.400000000 then
# return -1
#endi
print =============== select count(*) from child table
sql select count(*) from ct1
if $rows != 1 then
return -1
endi
print $data00 $data01 $data02
if $data00 != 4 then
return -1
endi
print =============== select count(column) from child table
sql select count(ts), count(c1), count(c2), count(c3) from ct1
print $data00 $data01 $data02 $data03
if $data00 != 4 then
return -1
endi
if $data01 != 4 then
return -1
endi
if $data02 != 4 then
return -1
endi
if $data03 != 4 then
return -1
endi
#print =============== select first(*)/first(column) from child table
#sql select first(*) from ct1
#sql select first(ts), first(c1), first(c2), first(c3) from ct1
print =============== select min(column) from child table
sql select min(c1), min(c2), min(c3) from ct1
print $data00 $data01 $data02 $data03
if $rows != 1 then
return -1
endi
if $data00 != 10 then
return -1
endi
if $data01 != 2.00000 then
return -1
endi
if $data02 != 3.000000000 then
return -1
endi
print =============== select max(column) from child table
sql select max(c1), max(c2), max(c3) from ct1
print $data00 $data01 $data02 $data03
if $rows != 1 then
return -1
endi
if $data00 != 13 then
return -1
endi
if $data01 != 2.30000 then
return -1
endi
if $data02 != 3.300000000 then
return -1
endi
print =============== select sum(column) from child table
sql select sum(c1), sum(c2), sum(c3) from ct1
print $data00 $data01 $data02 $data03
if $rows != 1 then
return -1
endi
if $data00 != 46 then
return -1
endi
if $data01 != 8.599999905 then
return -1
endi
if $data02 != 12.600000000 then
return -1
endi
print =============== select column, from child table
sql select c1, c2, c3 from ct1
print $data00 $data01 $data02
#if $rows != 4 then
# return -1
#endi
#if $data00 != 10 then
# return -1
#endi
#if $data01 != 2.00000 then
# return -1
#endi
#if $data02 != 3.000000000 then
# return -1
#endi
#if $data10 != 11 then
# return -1
#endi
#if $data11 != 2.10000 then
# return -1
#endi
#if $data12 != 3.100000000 then
# return -1
#endi
#if $data30 != 13 then
# return -1
#endi
#if $data31 != 2.30000 then
# return -1
#endi
#if $data32 != 3.300000000 then
# return -1
#endi
#===================================================================
#===================================================================
#print =============== query data from stb
#sql select * from stb
#if $rows != 4 then
# return -1
#endi
#print =============== select count(*) from supter table
#sql select count(*) from stb
#if $rows != 1 then
# return -1
#endi
#
#print $data00 $data01 $data02
#if $data00 != 8 then
# return -1
#endi
#
#print =============== select count(column) from supter table
#sql select count(ts), count(c1), count(c2), count(c3) from stb
#print $data00 $data01 $data02 $data03
#if $data00 != 8 then
# return -1
#endi
#if $data01 != 8 then
# return -1
#endi
#if $data02 != 8 then
# return -1
#endi
#if $data03 != 8 then
# return -1
#endi
#===================================================================
#===================================================================
print =============== stop and restart taosd, then again do query above
system sh/exec.sh -n dnode1 -s stop -x SIGINT
system sh/exec.sh -n dnode1 -s start
sleep 2000
sql select * from ct1
if $rows != 4 then # after fix bug, modify 4 to 7
return -1
endi
if $data01 != 10 then
return -1
endi
if $data02 != 2.00000 then
return -1
endi
if $data03 != 3.000000000 then
return -1
endi
#if $data41 != -14 then
# return -1
#endi
#if $data42 != -2.40000 then
# return -1
#endi
#if $data43 != -3.400000000 then
# return -1
#endi
print =============== select count(*) from child table
sql select count(*) from ct1
if $rows != 1 then
return -1
endi
print $data00 $data01 $data02
if $data00 != 4 then
return -1
endi
print =============== select count(column) from child table
sql select count(ts), count(c1), count(c2), count(c3) from ct1
print $data00 $data01 $data02 $data03
if $data00 != 4 then
return -1
endi
if $data01 != 4 then
return -1
endi
if $data02 != 4 then
return -1
endi
if $data03 != 4 then
return -1
endi
#print =============== select first(*)/first(column) from child table
#sql select first(*) from ct1
#sql select first(ts), first(c1), first(c2), first(c3) from ct1
print =============== select min(column) from child table
sql select min(c1), min(c2), min(c3) from ct1
print $data00 $data01 $data02 $data03
if $rows != 1 then
return -1
endi
if $data00 != 10 then
return -1
endi
if $data01 != 2.00000 then
return -1
endi
if $data02 != 3.000000000 then
return -1
endi
print =============== select max(column) from child table
sql select max(c1), max(c2), max(c3) from ct1
print $data00 $data01 $data02 $data03
if $rows != 1 then
return -1
endi
if $data00 != 13 then
return -1
endi
if $data01 != 2.30000 then
return -1
endi
if $data02 != 3.300000000 then
return -1
endi
print =============== select sum(column) from child table
sql select sum(c1), sum(c2), sum(c3) from ct1
print $data00 $data01 $data02 $data03
if $rows != 1 then
return -1
endi
if $data00 != 46 then
return -1
endi
if $data01 != 8.599999905 then
return -1
endi
if $data02 != 12.600000000 then
return -1
endi
#system sh/exec.sh -n dnode1 -s stop -x SIGINT

View File

@ -61,10 +61,50 @@ if $rows != 7 then
endi
print $data00 $data01 $data02
print $data10 $data11 $data22
print $data20 $data11 $data22
print $data10 $data11 $data12
print $data20 $data21 $data22
print =============== create normal table
sql create database ndb
sql use ndb
sql create table nt0 (ts timestamp, i int)
sql create table if not exists nt0 (ts timestamp, i int)
sql create table nt1 (ts timestamp, i int)
sql create table if not exists nt1 (ts timestamp, i int)
sql create table if not exists nt3 (ts timestamp, i int)
sql show tables
if $rows != 3 then
return -1
endi
sql insert into nt0 values(now+1s, 1)(now+2s, 2)(now+3s, 3)
sql insert into nt1 values(now+1s, 1)(now+2s, 2)(now+3s, 3)
sql select * from nt1
if $rows != 3 then
return -1
endi
print $data00 $data01
print $data10 $data11
print $data20 $data21
if $data01 != 1 then
return -1
endi
if $data11 != 2 then
return -1
endi
if $data21 != 3 then
return -1
endi
print =============== insert data
sql use d1
sql insert into c1 values(now+1s, 1)
sql insert into c1 values(now+2s, 2)
sql insert into c1 values(now+3s, 3)
@ -95,7 +135,7 @@ endi
print $data00 $data01
print $data10 $data11
print $data20 $data11
print $data20 $data21
if $data01 != 1 then
return -1
@ -160,7 +200,7 @@ endi
print $data00 $data01
print $data10 $data11
print $data20 $data11
print $data20 $data21
if $data01 != 1 then
return -1
@ -210,4 +250,27 @@ if $rows != 21 then
return -1
endi
print =============== query data from normal table after restart dnode
sql use ndb
sql select * from nt1
if $rows != 3 then
return -1
endi
print $data00 $data01
print $data10 $data11
print $data20 $data21
if $data01 != 1 then
return -1
endi
if $data11 != 2 then
return -1
endi
if $data21 != 3 then
return -1
endi
system sh/exec.sh -n dnode1 -s stop -x SIGINT