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

fix: crash whle reconfig mnode will cause taosd reboot failure

This commit is contained in:
Shengliang Guan 2022-10-17 14:34:51 +08:00
parent eb26fbeaf2
commit 8fe672feab
2 changed files with 34 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
source/dnode/mnode/impl/src/mndMnode.c
View File

@ -394,7 +394,6 @@ static int32_t mndCreateMnode(SMnode *pMnode, SRpcMsg *pReq, SDnodeObj *pDnode,
if (mndSetCreateMnodeRedoActions(pMnode, pTrans, pDnode, &mnodeObj) != 0) goto _OVER; if (mndSetCreateMnodeRedoActions(pMnode, pTrans, pDnode, &mnodeObj) != 0) goto _OVER;
if (mndSetCreateMnodeRedoLogs(pMnode, pTrans, &mnodeObj) != 0) goto _OVER; if (mndSetCreateMnodeRedoLogs(pMnode, pTrans, &mnodeObj) != 0) goto _OVER;
if (mndSetCreateMnodeCommitLogs(pMnode, pTrans, &mnodeObj) != 0) goto _OVER; if (mndSetCreateMnodeCommitLogs(pMnode, pTrans, &mnodeObj) != 0) goto _OVER;
if (mndTransAppendNullLog(pTrans) != 0) goto _OVER;
if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER; if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
code = 0; code = 0;
@ -478,7 +477,6 @@ static int32_t mndSetDropMnodeCommitLogs(SMnode *pMnode, STrans *pTrans, SMnodeO
static int32_t mndSetDropMnodeRedoActions(SMnode *pMnode, STrans *pTrans, SDnodeObj *pDnode, SMnodeObj *pObj) { static int32_t mndSetDropMnodeRedoActions(SMnode *pMnode, STrans *pTrans, SDnodeObj *pDnode, SMnodeObj *pObj) {
SSdb *pSdb = pMnode->pSdb; SSdb *pSdb = pMnode->pSdb;
void *pIter = NULL; void *pIter = NULL;
int32_t numOfReplicas = 0;
SDDropMnodeReq dropReq = {0}; SDDropMnodeReq dropReq = {0};
SEpSet dropEpSet = {0}; SEpSet dropEpSet = {0};
@ -505,9 +503,8 @@ static int32_t mndSetDropMnodeRedoActions(SMnode *pMnode, STrans *pTrans, SDnode
int32_t mndSetDropMnodeInfoToTrans(SMnode *pMnode, STrans *pTrans, SMnodeObj *pObj) { int32_t mndSetDropMnodeInfoToTrans(SMnode *pMnode, STrans *pTrans, SMnodeObj *pObj) {
if (pObj == NULL) return 0; if (pObj == NULL) return 0;
if (mndSetDropMnodeCommitLogs(pMnode, pTrans, pObj) != 0) return -1;
if (mndSetDropMnodeRedoActions(pMnode, pTrans, pObj->pDnode, pObj) != 0) return -1; if (mndSetDropMnodeRedoActions(pMnode, pTrans, pObj->pDnode, pObj) != 0) return -1;
if (mndTransAppendNullLog(pTrans) != 0) return -1; if (mndSetDropMnodeCommitLogs(pMnode, pTrans, pObj) != 0) return -1;
return 0; return 0;
} }
@ -715,7 +712,8 @@ static void mndReloadSyncConfig(SMnode *pMnode) {
SMnodeObj *pObj = NULL; SMnodeObj *pObj = NULL;
ESdbStatus objStatus = 0; ESdbStatus objStatus = 0;
void *pIter = NULL; void *pIter = NULL;
bool hasUpdatingMnode = false; int32_t updatingMnodes = 0;
int32_t readyMnodes = 0;
SSyncCfg cfg = {.myIndex = -1}; SSyncCfg cfg = {.myIndex = -1};
while (1) { while (1) {
@ -723,7 +721,11 @@ static void mndReloadSyncConfig(SMnode *pMnode) {
if (pIter == NULL) break; if (pIter == NULL) break;
if (objStatus == SDB_STATUS_CREATING || objStatus == SDB_STATUS_DROPPING) { if (objStatus == SDB_STATUS_CREATING || objStatus == SDB_STATUS_DROPPING) {
mInfo("vgId:1, has updating mnode:%d, status:%s", pObj->id, sdbStatusName(objStatus)); mInfo("vgId:1, has updating mnode:%d, status:%s", pObj->id, sdbStatusName(objStatus));
hasUpdatingMnode = true; updatingMnodes++;
}
if (objStatus == SDB_STATUS_READY) {
mInfo("vgId:1, has ready mnode:%d, status:%s", pObj->id, sdbStatusName(objStatus));
readyMnodes++;
} }
if (objStatus == SDB_STATUS_READY || objStatus == SDB_STATUS_CREATING) { if (objStatus == SDB_STATUS_READY || objStatus == SDB_STATUS_CREATING) {
@ -739,18 +741,18 @@ static void mndReloadSyncConfig(SMnode *pMnode) {
sdbReleaseLock(pSdb, pObj, false); sdbReleaseLock(pSdb, pObj, false);
} }
if (readyMnodes <= 0 || updatingMnodes <= 0) {
mInfo("vgId:1, mnode sync not reconfig since readyMnodes:%d updatingMnodes:%d", readyMnodes, updatingMnodes);
return;
}
if (cfg.myIndex == -1) { if (cfg.myIndex == -1) {
mInfo("vgId:1, mnode not reload since selfIndex is -1"); mInfo("vgId:1, mnode sync not reconfig since selfIndex is -1");
return; return;
} }
if (!mndGetRestored(pMnode)) { if (updatingMnodes > 0) {
mInfo("vgId:1, mnode not reload since restore not finished"); mInfo("vgId:1, mnode sync reconfig, replica:%d myIndex:%d", cfg.replicaNum, cfg.myIndex);
return;
}
if (hasUpdatingMnode) {
mInfo("vgId:1, start to reload mnode sync, replica:%d myIndex:%d", cfg.replicaNum, cfg.myIndex);
for (int32_t i = 0; i < cfg.replicaNum; ++i) { for (int32_t i = 0; i < cfg.replicaNum; ++i) {
SNodeInfo *pNode = &cfg.nodeInfo[i]; SNodeInfo *pNode = &cfg.nodeInfo[i];
mInfo("vgId:1, index:%d, fqdn:%s port:%d", i, pNode->nodeFqdn, pNode->nodePort); mInfo("vgId:1, index:%d, fqdn:%s port:%d", i, pNode->nodeFqdn, pNode->nodePort);

18
source/libs/sync/src/syncMain.c
View File

@ -1891,8 +1891,26 @@ inline bool syncNodeInConfig(SSyncNode* pSyncNode, const SSyncCfg* config) {
return b1; return b1;
} }
static bool syncIsConfigChanged(const SSyncCfg* pOldCfg, const SSyncCfg* pNewCfg) {
if (pOldCfg->replicaNum != pNewCfg->replicaNum) return true;
if (pOldCfg->myIndex != pNewCfg->myIndex) return true;
for (int32_t i = 0; i < pOldCfg->replicaNum; ++i) {
const SNodeInfo* pOldInfo = &pOldCfg->nodeInfo[i];
const SNodeInfo* pNewInfo = &pNewCfg->nodeInfo[i];
if (strcmp(pOldInfo->nodeFqdn, pNewInfo->nodeFqdn) != 0) return true;
if (pOldInfo->nodePort != pNewInfo->nodePort) return true;
}
return false;
}
void syncNodeDoConfigChange(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex lastConfigChangeIndex) { void syncNodeDoConfigChange(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex lastConfigChangeIndex) {
SSyncCfg oldConfig = pSyncNode->pRaftCfg->cfg; SSyncCfg oldConfig = pSyncNode->pRaftCfg->cfg;
if (!syncIsConfigChanged(&oldConfig, pNewConfig)) {
sInfo("vgId:1, sync not reconfig since not changed");
return;
}
pSyncNode->pRaftCfg->cfg = *pNewConfig; pSyncNode->pRaftCfg->cfg = *pNewConfig;
pSyncNode->pRaftCfg->lastConfigIndex = lastConfigChangeIndex; pSyncNode->pRaftCfg->lastConfigIndex = lastConfigChangeIndex;