提交 9a792404 编写于 作者: M Minghao Li

sync integration

上级 01008939
......@@ -84,6 +84,9 @@ typedef struct SSyncFSM {
struct SSyncRaftEntry;
typedef struct SSyncRaftEntry SSyncRaftEntry;
#define SYNC_INDEX_BEGIN 0
#define SYNC_INDEX_INVALID -1
// abstract definition of log store in raft
// SWal implements it
typedef struct SSyncLogStore {
......
......@@ -109,6 +109,7 @@ void vmStopWorker(SVnodesMgmt *pMgmt);
int32_t vmAllocQueue(SVnodesMgmt *pMgmt, SVnodeObj *pVnode);
void vmFreeQueue(SVnodesMgmt *pMgmt, SVnodeObj *pVnode);
int32_t vmPutMsgToSyncQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc); // sync integration
int32_t vmPutMsgToQueryQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
int32_t vmPutMsgToFetchQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
int32_t vmPutMsgToApplyQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
......
......@@ -106,6 +106,16 @@ static void vmGenerateVnodeCfg(SCreateVnodeReq *pCreate, SVnodeCfg *pCfg) {
pCfg->hashBegin = pCreate->hashBegin;
pCfg->hashEnd = pCreate->hashEnd;
pCfg->hashMethod = pCreate->hashMethod;
// sync integration
pCfg->syncCfg.myIndex = pCreate->selfIndex;
pCfg->syncCfg.replicaNum = pCreate->replica;
memset(&(pCfg->syncCfg.nodeInfo), 0, sizeof(pCfg->syncCfg.nodeInfo));
for (int i = 0; i < pCreate->replica; ++i) {
(pCfg->syncCfg.nodeInfo)[i].nodePort = (pCreate->replicas)[i].port;
snprintf((pCfg->syncCfg.nodeInfo)[i].nodeFqdn, sizeof((pCfg->syncCfg.nodeInfo)[i].nodeFqdn), "%s",
(pCreate->replicas)[i].fqdn);
}
}
static void vmGenerateWrapperCfg(SVnodesMgmt *pMgmt, SCreateVnodeReq *pCreate, SWrapperCfg *pCfg) {
......@@ -157,6 +167,7 @@ int32_t vmProcessCreateVnodeReq(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
msgCb.queueFps[QUERY_QUEUE] = vmPutMsgToQueryQueue;
msgCb.queueFps[FETCH_QUEUE] = vmPutMsgToFetchQueue;
msgCb.queueFps[APPLY_QUEUE] = vmPutMsgToApplyQueue;
msgCb.queueFps[SYNC_QUEUE] = vmPutMsgToSyncQueue; // sync integration
msgCb.qsizeFp = vmGetQueueSize;
SVnode *pImpl = vnodeOpen(path, pMgmt->pTfs, msgCb);
......@@ -357,4 +368,15 @@ void vmInitMsgHandle(SMgmtWrapper *pWrapper) {
dmSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
// sync integration
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_TIMEOUT, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_PING, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_PING_REPLY, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_CLIENT_REQUEST, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_CLIENT_REQUEST_REPLY, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_REQUEST_VOTE, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_REQUEST_VOTE_REPLY, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_APPEND_ENTRIES, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_APPEND_ENTRIES_REPLY, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);
}
......@@ -137,6 +137,7 @@ static void *vmOpenVnodeFunc(void *param) {
msgCb.queueFps[QUERY_QUEUE] = vmPutMsgToQueryQueue;
msgCb.queueFps[FETCH_QUEUE] = vmPutMsgToFetchQueue;
msgCb.queueFps[APPLY_QUEUE] = vmPutMsgToApplyQueue;
msgCb.queueFps[SYNC_QUEUE] = vmPutMsgToSyncQueue; // sync integration
msgCb.qsizeFp = vmGetQueueSize;
snprintf(path, TSDB_FILENAME_LEN, "vnode%svnode%d", TD_DIRSEP, pCfg->vgId);
SVnode *pImpl = vnodeOpen(path, pMgmt->pTfs, msgCb);
......
......@@ -306,6 +306,10 @@ static int32_t vmPutRpcMsgToQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc, EQueueT
dTrace("msg:%p, will be put into vnode-merge queue", pMsg);
taosWriteQitem(pVnode->pMergeQ, pMsg);
break;
case SYNC_QUEUE: // sync integration
dTrace("msg:%p, will be put into vnode-sync queue", pMsg);
taosWriteQitem(pVnode->pSyncQ, pMsg);
break;
default:
code = -1;
terrno = TSDB_CODE_INVALID_PARA;
......@@ -332,6 +336,11 @@ int32_t vmPutMsgToMergeQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
return vmPutRpcMsgToQueue(pWrapper, pRpc, MERGE_QUEUE);
}
// sync integration
int32_t vmPutMsgToSyncQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
return vmPutRpcMsgToQueue(pWrapper, pRpc, SYNC_QUEUE);
}
int32_t vmGetQueueSize(SMgmtWrapper *pWrapper, int32_t vgId, EQueueType qtype) {
int32_t size = -1;
SVnodeObj *pVnode = vmAcquireVnode(pWrapper->pMgmt, vgId);
......
......@@ -16,6 +16,7 @@ target_sources(
"src/vnd/vnodeWrite.c"
"src/vnd/vnodeModule.c"
"src/vnd/vnodeSvr.c"
"src/vnd/vnodeSync.c"
# meta
"src/meta/metaOpen.c"
......
......@@ -24,6 +24,7 @@
#include "tarray.h"
#include "tfs.h"
#include "wal.h"
#include "sync.h"
#include "tcommon.h"
#include "tfs.h"
......@@ -144,6 +145,7 @@ struct SVnodeCfg {
bool isWeak;
STsdbCfg tsdbCfg;
SWalCfg walCfg;
SSyncCfg syncCfg; // sync integration
uint32_t hashBegin;
uint32_t hashEnd;
int8_t hashMethod;
......
......@@ -54,8 +54,8 @@ typedef struct SQWorkerMgmt SQHandle;
#define VNODE_META_DIR "meta"
#define VNODE_TSDB_DIR "tsdb"
#define VNODE_TQ_DIR "tq"
#define VNODE_WAL_DIR "wal"
#define VNODE_TQ_DIR "tq"
#define VNODE_WAL_DIR "wal"
typedef struct {
int8_t streamType; // sma or other
......@@ -68,8 +68,8 @@ typedef struct {
} SStreamSinkInfo;
typedef struct {
SVnode* pVnode;
SHashObj* pHash; // streamId -> SStreamSinkInfo
SVnode *pVnode;
SHashObj *pHash; // streamId -> SStreamSinkInfo
} SSink;
// SVState
......@@ -85,25 +85,26 @@ struct SVnodeInfo {
};
struct SVnode {
char* path;
char *path;
SVnodeCfg config;
SVState state;
STfs* pTfs;
STfs *pTfs;
SMsgCb msgCb;
SVBufPool* pBufPool;
SMeta* pMeta;
STsdb* pTsdb;
SWal* pWal;
STQ* pTq;
SSink* pSink;
SVBufPool *pBufPool;
SMeta *pMeta;
STsdb *pTsdb;
SWal *pWal;
STQ *pTq;
SSink *pSink;
int64_t sync; // sync integration
tsem_t canCommit;
SQHandle* pQuery;
SQHandle *pQuery;
};
#define TD_VID(PVNODE) (PVNODE)->config.vgId
// sma
void smaHandleRes(void* pVnode, int64_t smaId, const SArray* data);
void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data);
#include "vnd.h"
......@@ -113,6 +114,8 @@ void smaHandleRes(void* pVnode, int64_t smaId, const SArray* data);
#include "tq.h"
#include "vnodeSync.h"
#ifdef __cplusplus
}
#endif
......
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _TD_VNODE_SYNC_H_
#define _TD_VNODE_SYNC_H_
#ifdef __cplusplus
extern "C" {
#endif
int32_t vnodeSyncOpen(SVnode *pVnode);
int32_t vnodeSyncStart(SVnode *pVnode);
void vnodeSyncClose(SVnode *pVnode);
int32_t vnodeSyncEqMsg(void *qHandle, SRpcMsg *pMsg);
int32_t vnodeSendMsg(void *rpcHandle, const SEpSet *pEpSet, SRpcMsg *pMsg);
void vnodeSyncCommitCb(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta);
void vnodeSyncPreCommitCb(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta);
void vnodeSyncRollBackCb(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta);
int32_t vnodeSyncGetSnapshotCb(struct SSyncFSM *pFsm, SSnapshot *pSnapshot);
SSyncFSM *syncVnodeMakeFsm();
#ifdef __cplusplus
}
#endif
#endif /*_TD_VNODE_SYNC_H_*/
......@@ -124,6 +124,12 @@ SVnode *vnodeOpen(const char *path, STfs *pTfs, SMsgCb msgCb) {
goto _err;
}
// sync integration
// open sync
if (vnodeSyncOpen(pVnode)) {
goto _err;
}
#if 0
if (vnodeBegin() < 0) {
goto _err;
......
......@@ -13,6 +13,8 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "sync.h"
#include "syncTools.h"
#include "vnodeInt.h"
static int vnodeProcessCreateStbReq(SVnode *pVnode, void *pReq);
......@@ -198,8 +200,86 @@ void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data) {
tsdbInsertTSmaData(((SVnode *)pVnode)->pTsdb, smaId, (const char *)data);
}
// sync integration
int vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
/*vInfo("sync message is processed");*/
SSyncNode *pSyncNode = syncNodeAcquire(pVnode->sync);
assert(pSyncNode != NULL);
ESyncState state = syncGetMyRole(pVnode->sync);
SyncTerm currentTerm = syncGetMyTerm(pVnode->sync);
SMsgHead *pHead = pMsg->pCont;
char logBuf[512];
char *syncNodeStr = sync2SimpleStr(pVnode->sync);
snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
syncRpcMsgLog2(logBuf, pMsg);
taosMemoryFree(syncNodeStr);
SRpcMsg *pRpcMsg = pMsg;
if (pRpcMsg->msgType == TDMT_VND_SYNC_TIMEOUT) {
SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
syncTimeoutDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_PING) {
SyncPing *pSyncMsg = syncPingFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnPingCb(pSyncNode, pSyncMsg);
syncPingDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_PING_REPLY) {
SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
syncPingReplyDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_CLIENT_REQUEST) {
SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
syncClientRequestDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_REQUEST_VOTE) {
SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
syncRequestVoteDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_REQUEST_VOTE_REPLY) {
SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
syncRequestVoteReplyDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_APPEND_ENTRIES) {
SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
syncAppendEntriesDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_VND_SYNC_APPEND_ENTRIES_REPLY) {
SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
} else {
vError("==vnodeProcessSyncReq== error msg type:%d", pRpcMsg->msgType);
}
syncNodeRelease(pSyncNode);
return 0;
}
......
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "sync.h"
#include "syncTools.h"
#include "tmsgcb.h"
#include "vnodeInt.h"
// sync integration
int32_t vnodeSyncOpen(SVnode *pVnode) {
SSyncInfo syncInfo;
syncInfo.vgId = pVnode->config.vgId;
SSyncCfg *pCfg = &(syncInfo.syncCfg);
pCfg->replicaNum = pVnode->config.syncCfg.replicaNum;
pCfg->myIndex = pVnode->config.syncCfg.myIndex;
memcpy(pCfg->nodeInfo, pVnode->config.syncCfg.nodeInfo, sizeof(pCfg->nodeInfo));
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s/sync", pVnode->path);
syncInfo.pWal = pVnode->pWal;
syncInfo.pFsm = syncVnodeMakeFsm(pVnode);
syncInfo.rpcClient = NULL;
syncInfo.FpSendMsg = vnodeSendMsg;
syncInfo.queue = NULL;
syncInfo.FpEqMsg = vnodeSyncEqMsg;
pVnode->sync = syncOpen(&syncInfo);
assert(pVnode->sync > 0);
// for test
setPingTimerMS(pVnode->sync, 3000);
setElectTimerMS(pVnode->sync, 500);
setHeartbeatTimerMS(pVnode->sync, 100);
return 0;
}
int32_t vnodeSyncStart(SVnode *pVnode) {
syncStart(pVnode->sync);
return 0;
}
void vnodeSyncClose(SVnode *pVnode) {
// stop by ref id
syncStop(pVnode->sync);
}
void vnodeSyncSetQ(SVnode *pVnode, void *qHandle) { syncSetQ(pVnode->sync, (void *)(&(pVnode->msgCb))); }
void vnodeSyncSetRpc(SVnode *pVnode, void *rpcHandle) { syncSetRpc(pVnode->sync, (void *)(&(pVnode->msgCb))); }
int32_t vnodeSyncEqMsg(void *qHandle, SRpcMsg *pMsg) {
int32_t ret = 0;
SMsgCb *pMsgCb = qHandle;
if (pMsgCb->queueFps[SYNC_QUEUE] != NULL) {
tmsgPutToQueue(qHandle, SYNC_QUEUE, pMsg);
} else {
vError("vnodeSyncEqMsg queue is NULL, SYNC_QUEUE:%d", SYNC_QUEUE);
}
return ret;
}
int32_t vnodeSendMsg(void *rpcHandle, const SEpSet *pEpSet, SRpcMsg *pMsg) {
int32_t ret = 0;
SMsgCb *pMsgCb = rpcHandle;
if (pMsgCb->queueFps[SYNC_QUEUE] != NULL) {
tmsgSendReq(rpcHandle, pEpSet, pMsg);
} else {
vError("vnodeSendMsg queue is NULL, SYNC_QUEUE:%d", SYNC_QUEUE);
}
return ret;
}
int32_t vnodeSyncGetSnapshotCb(struct SSyncFSM *pFsm, SSnapshot *pSnapshot) {
pSnapshot->data = NULL;
pSnapshot->lastApplyIndex = 0;
pSnapshot->lastApplyTerm = 0;
return 0;
}
void vnodeSyncCommitCb(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta) {
SyncIndex beginIndex = SYNC_INDEX_INVALID;
if (pFsm->FpGetSnapshot != NULL) {
SSnapshot snapshot;
pFsm->FpGetSnapshot(pFsm, &snapshot);
beginIndex = snapshot.lastApplyIndex;
}
if (cbMeta.index > beginIndex) {
char logBuf[256];
snprintf(
logBuf, sizeof(logBuf),
"==callback== ==CommitCb== execute, pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s, beginIndex :%ld\n",
pFsm, cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state), beginIndex);
syncRpcMsgLog2(logBuf, (SRpcMsg *)pMsg);
SRpcMsg applyMsg;
applyMsg = *pMsg;
applyMsg.pCont = rpcMallocCont(applyMsg.contLen);
assert(applyMsg.contLen == pMsg->contLen);
memcpy(applyMsg.pCont, pMsg->pCont, applyMsg.contLen);
// recover handle for response
SVnode *pVnode = (SVnode *)(pFsm->data);
SRpcMsg saveRpcMsg;
int32_t ret = syncGetAndDelRespRpc(pVnode->sync, cbMeta.seqNum, &saveRpcMsg);
if (ret == 1) {
applyMsg.handle = saveRpcMsg.handle;
applyMsg.ahandle = saveRpcMsg.ahandle;
}
// put to applyQ
tmsgPutToQueue(&(pVnode->msgCb), APPLY_QUEUE, &applyMsg);
} else {
char logBuf[256];
snprintf(logBuf, sizeof(logBuf),
"==callback== ==CommitCb== do not execute, pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s, "
"beginIndex :%ld\n",
pFsm, cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state),
beginIndex);
syncRpcMsgLog2(logBuf, (SRpcMsg *)pMsg);
}
}
void vnodeSyncPreCommitCb(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta) {
char logBuf[256];
snprintf(logBuf, sizeof(logBuf),
"==callback== ==PreCommitCb== pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s \n", pFsm, cbMeta.index,
cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state));
syncRpcMsgLog2(logBuf, (SRpcMsg *)pMsg);
}
void vnodeSyncRollBackCb(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta) {
char logBuf[256];
snprintf(logBuf, sizeof(logBuf), "==callback== ==RollBackCb== pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s \n",
pFsm, cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state));
syncRpcMsgLog2(logBuf, (SRpcMsg *)pMsg);
}
SSyncFSM *syncVnodeMakeFsm(SVnode *pVnode) {
SSyncFSM *pFsm = (SSyncFSM *)taosMemoryMalloc(sizeof(SSyncFSM));
pFsm->data = pVnode;
pFsm->FpCommitCb = vnodeSyncCommitCb;
pFsm->FpPreCommitCb = vnodeSyncPreCommitCb;
pFsm->FpRollBackCb = vnodeSyncRollBackCb;
pFsm->FpGetSnapshot = vnodeSyncGetSnapshotCb;
return pFsm;
}
......@@ -27,9 +27,6 @@ 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;
......
......@@ -112,7 +112,7 @@ SSyncNode *syncNodeInit() {
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);
taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
}
SSyncNode *pSyncNode = syncNodeOpen(&syncInfo);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册