Skip to content

M
MariaDB
Commit 2aefa6cf authored by jonas@perch.ndb.mysql.com's avatar jonas@perch.ndb.mysql.com
Browse files
Options

Merge joreland@bk-internal.mysql.com:/home/bk/mysql-5.1-new-ndb 

into  perch.ndb.mysql.com:/home/jonas/src/51-ndb
parents b303fc39 49a15472
Hide whitespace changes
Inline Side-by-side
Showing with 296 additions and 236 deletions
storage/ndb/src/kernel/blocks/cmvmi/Cmvmi.cpp
View file @ 2aefa6cf
......@@ -421,9 +421,10 @@ void Cmvmi::execCLOSE_COMREQ(Signal* signal)
// Uint32 noOfNodes = closeCom->noOfNodes;
jamEntry();
for (unsigned i = 0; i < MAX_NODES; i++){
if(NodeBitmask::get(closeCom->theNodes, i)){
for (unsigned i = 0; i < MAX_NODES; i++)
{
if(NodeBitmask::get(closeCom->theNodes, i))
{
jam();
//-----------------------------------------------------
......@@ -437,7 +438,9 @@ void Cmvmi::execCLOSE_COMREQ(Signal* signal)
globalTransporterRegistry.do_disconnect(i);
}
}
if (failNo != 0) {
if (failNo != 0)
{
jam();
signal->theData[0] = userRef;
signal->theData[1] = failNo;
......@@ -456,13 +459,21 @@ void Cmvmi::execOPEN_COMREQ(Signal* signal)
jamEntry();
const Uint32 len = signal->getLength();
if(len == 2){
if(len == 2)
{
#ifdef ERROR_INSERT
if (! ((ERROR_INSERTED(9000) || ERROR_INSERTED(9002))
&& c_error_9000_nodes_mask.get(tStartingNode)))
#endif
{
if (globalData.theStartLevel != NodeState::SL_STARTED &&
(getNodeInfo(tStartingNode).m_type != NodeInfo::DB &&
getNodeInfo(tStartingNode).m_type != NodeInfo::MGM))
{
jam();
goto done;
}
globalTransporterRegistry.do_connect(tStartingNode);
globalTransporterRegistry.setIOState(tStartingNode, HaltIO);
......@@ -475,9 +486,11 @@ void Cmvmi::execOPEN_COMREQ(Signal* signal)
//-----------------------------------------------------
}
} else {
for(unsigned int i = 1; i < MAX_NODES; i++ ) {
for(unsigned int i = 1; i < MAX_NODES; i++ )
{
jam();
if (i != getOwnNodeId() && getNodeInfo(i).m_type == tData2){
if (i != getOwnNodeId() && getNodeInfo(i).m_type == tData2)
{
jam();
#ifdef ERROR_INSERT
......@@ -496,6 +509,7 @@ void Cmvmi::execOPEN_COMREQ(Signal* signal)
}
}
done:
if (userRef != 0) {
jam();
signal->theData[0] = tStartingNode;
......@@ -536,24 +550,10 @@ void Cmvmi::execDISCONNECT_REP(Signal *signal)
setNodeInfo(hostId).m_connectCount++;
const NodeInfo::NodeType type = getNodeInfo(hostId).getType();
ndbrequire(type != NodeInfo::INVALID);
if(type == NodeInfo::DB || globalData.theStartLevel == NodeState::SL_STARTED){
jam();
DisconnectRep * const rep = (DisconnectRep *)&signal->theData[0];
rep->nodeId = hostId;
rep->err = errNo;
sendSignal(QMGR_REF, GSN_DISCONNECT_REP, signal,
DisconnectRep::SignalLength, JBA);
} else if((globalData.theStartLevel == NodeState::SL_CMVMI ||
globalData.theStartLevel == NodeState::SL_STARTING)
&& type == NodeInfo::MGM) {
/**
* Someone disconnected during cmvmi period
*/
jam();
globalTransporterRegistry.do_connect(hostId);
}
sendSignal(QMGR_REF, GSN_DISCONNECT_REP, signal,
DisconnectRep::SignalLength, JBA);
cancelSubscription(hostId);
signal->theData[0] = NDB_LE_Disconnected;
......@@ -587,6 +587,8 @@ void Cmvmi::execCONNECT_REP(Signal *signal){
*/
if(type == NodeInfo::MGM){
jam();
signal->theData[0] = hostId;
sendSignal(QMGR_REF, GSN_CONNECT_REP, signal, 1, JBA);
} else {
/**
* Dont allow api nodes to connect
......@@ -802,6 +804,8 @@ Cmvmi::execSTART_ORD(Signal* signal) {
}
}
}
EXECUTE_DIRECT(QMGR, GSN_START_ORD, signal, 1);
return ;
}
......@@ -829,9 +833,6 @@ Cmvmi::execSTART_ORD(Signal* signal) {
*
* Do Restart
*/
globalScheduler.clear();
globalTimeQueue.clear();
// Disconnect all nodes as part of the system restart.
// We need to ensure that we are starting up
......
storage/ndb/src/kernel/blocks/qmgr/Qmgr.hpp
View file @ 2aefa6cf
......@@ -265,6 +265,8 @@ private:
void execALLOC_NODEID_CONF(Signal *);
void execALLOC_NODEID_REF(Signal *);
void completeAllocNodeIdReq(Signal *);
void execSTART_ORD(Signal*);
// Arbitration signals
void execARBIT_CFG(Signal* signal);
......@@ -281,6 +283,7 @@ private:
void check_readnodes_reply(Signal* signal, Uint32 nodeId, Uint32 gsn);
Uint32 check_startup(Signal* signal);
void api_failed(Signal* signal, Uint32 aFailedNode);
void node_failed(Signal* signal, Uint16 aFailedNode);
void checkStartInterface(Signal* signal);
void failReport(Signal* signal,
......
storage/ndb/src/kernel/blocks/qmgr/QmgrInit.cpp
View file @ 2aefa6cf
......@@ -31,10 +31,6 @@ void Qmgr::initData()
cnoCommitFailedNodes = 0;
c_maxDynamicId = 0;
c_clusterNodes.clear();
Uint32 hbDBAPI = 500;
setHbApiDelay(hbDBAPI);
c_connectedNodes.set(getOwnNodeId());
c_stopReq.senderRef = 0;
/**
......@@ -43,6 +39,27 @@ void Qmgr::initData()
ndbrequire((Uint32)NodeInfo::DB == 0);
ndbrequire((Uint32)NodeInfo::API == 1);
ndbrequire((Uint32)NodeInfo::MGM == 2);
NodeRecPtr nodePtr;
nodePtr.i = getOwnNodeId();
ptrAss(nodePtr, nodeRec);
nodePtr.p->blockRef = reference();
c_connectedNodes.set(getOwnNodeId());
setNodeInfo(getOwnNodeId()).m_version = NDB_VERSION;
/**
* Timeouts
*/
const ndb_mgm_configuration_iterator * p =
m_ctx.m_config.getOwnConfigIterator();
ndbrequire(p != 0);
Uint32 hbDBAPI = 1500;
ndb_mgm_get_int_parameter(p, CFG_DB_API_HEARTBEAT_INTERVAL, &hbDBAPI);
setHbApiDelay(hbDBAPI);
}//Qmgr::initData()
void Qmgr::initRecords()
......@@ -113,6 +130,7 @@ Qmgr::Qmgr(Block_context& ctx)
addRecSignal(GSN_DIH_RESTARTREF, &Qmgr::execDIH_RESTARTREF);
addRecSignal(GSN_DIH_RESTARTCONF, &Qmgr::execDIH_RESTARTCONF);
addRecSignal(GSN_NODE_VERSION_REP, &Qmgr::execNODE_VERSION_REP);
addRecSignal(GSN_START_ORD, &Qmgr::execSTART_ORD);
initData();
}//Qmgr::Qmgr()
......
storage/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp
View file @ 2aefa6cf
......@@ -238,6 +238,38 @@ Qmgr::execREAD_CONFIG_REQ(Signal* signal)
ReadConfigConf::SignalLength, JBB);
}
void
Qmgr::execSTART_ORD(Signal* signal)
{
/**
* Start timer handling
*/
signal->theData[0] = ZTIMER_HANDLING;
sendSignal(QMGR_REF, GSN_CONTINUEB, signal, 1, JBB);
NodeRecPtr nodePtr;
for (nodePtr.i = 1; nodePtr.i < MAX_NODES; nodePtr.i++)
{
ptrAss(nodePtr, nodeRec);
nodePtr.p->ndynamicId = 0;
if(getNodeInfo(nodePtr.i).m_type == NodeInfo::DB)
{
nodePtr.p->phase = ZINIT;
c_definedNodes.set(nodePtr.i);
} else {
nodePtr.p->phase = ZAPI_INACTIVE;
}
setNodeInfo(nodePtr.i).m_heartbeat_cnt= 0;
nodePtr.p->sendPrepFailReqStatus = Q_NOT_ACTIVE;
nodePtr.p->sendCommitFailReqStatus = Q_NOT_ACTIVE;
nodePtr.p->sendPresToStatus = Q_NOT_ACTIVE;
nodePtr.p->failState = NORMAL;
nodePtr.p->rcv[0] = 0;
nodePtr.p->rcv[1] = 0;
}//for
}
/*
4.2 ADD NODE MODULE*/
/*##########################################################################*/
......@@ -298,8 +330,6 @@ void Qmgr::startphase1(Signal* signal)
nodePtr.i = getOwnNodeId();
ptrAss(nodePtr, nodeRec);
nodePtr.p->phase = ZSTARTING;
nodePtr.p->blockRef = reference();
c_connectedNodes.set(nodePtr.i);
signal->theData[0] = reference();
sendSignal(DBDIH_REF, GSN_DIH_RESTARTREQ, signal, 1, JBB);
......@@ -371,11 +401,14 @@ void Qmgr::execCONNECT_REP(Signal* signal)
case ZFAIL_CLOSING:
jam();
return;
case ZINIT:
ndbrequire(false);
case ZAPI_ACTIVE:
case ZAPI_INACTIVE:
return;
case ZINIT:
ndbrequire(getNodeInfo(nodeId).m_type == NodeInfo::MGM);
break;
default:
ndbrequire(false);
}
if (getNodeInfo(nodeId).getType() != NodeInfo::DB)
......@@ -1212,12 +1245,6 @@ void Qmgr::execCM_REGREF(Signal* signal)
{
jam();
electionWon(signal);
/**
* Start timer handling
*/
signal->theData[0] = ZTIMER_HANDLING;
sendSignal(QMGR_REF, GSN_CONTINUEB, signal, 10, JBB);
}
return;
......@@ -1855,12 +1882,6 @@ Qmgr::joinedCluster(Signal* signal, NodeRecPtr nodePtr){
sendSttorryLab(signal);
/**
* Start timer handling
*/
signal->theData[0] = ZTIMER_HANDLING;
sendSignal(QMGR_REF, GSN_CONTINUEB, signal, 10, JBB);
sendCmAckAdd(signal, getOwnNodeId(), CmAdd::CommitNew);
}
......@@ -2094,25 +2115,6 @@ void Qmgr::findNeighbours(Signal* signal)
/*---------------------------------------------------------------------------*/
void Qmgr::initData(Signal* signal)
{
NodeRecPtr nodePtr;
for (nodePtr.i = 1; nodePtr.i < MAX_NODES; nodePtr.i++) {
ptrAss(nodePtr, nodeRec);
nodePtr.p->ndynamicId = 0;
if(getNodeInfo(nodePtr.i).m_type == NodeInfo::DB){
nodePtr.p->phase = ZINIT;
c_definedNodes.set(nodePtr.i);
} else {
nodePtr.p->phase = ZAPI_INACTIVE;
}
setNodeInfo(nodePtr.i).m_heartbeat_cnt= 0;
nodePtr.p->sendPrepFailReqStatus = Q_NOT_ACTIVE;
nodePtr.p->sendCommitFailReqStatus = Q_NOT_ACTIVE;
nodePtr.p->sendPresToStatus = Q_NOT_ACTIVE;
nodePtr.p->failState = NORMAL;
nodePtr.p->rcv[0] = 0;
nodePtr.p->rcv[1] = 0;
}//for
cfailureNr = 1;
ccommitFailureNr = 1;
cprepareFailureNr = 1;
......@@ -2146,13 +2148,11 @@ void Qmgr::initData(Signal* signal)
ndbrequire(p != 0);
Uint32 hbDBDB = 1500;
Uint32 hbDBAPI = 1500;
Uint32 arbitTimeout = 1000;
c_restartPartialTimeout = 30000;
c_restartPartionedTimeout = 60000;
c_restartFailureTimeout = ~0;
ndb_mgm_get_int_parameter(p, CFG_DB_HEARTBEAT_INTERVAL, &hbDBDB);
ndb_mgm_get_int_parameter(p, CFG_DB_API_HEARTBEAT_INTERVAL, &hbDBAPI);
ndb_mgm_get_int_parameter(p, CFG_DB_ARBIT_TIMEOUT, &arbitTimeout);
ndb_mgm_get_int_parameter(p, CFG_DB_START_PARTIAL_TIMEOUT,
&c_restartPartialTimeout);
......@@ -2177,7 +2177,6 @@ void Qmgr::initData(Signal* signal)
}
setHbDelay(hbDBDB);
setHbApiDelay(hbDBAPI);
setArbitTimeout(arbitTimeout);
arbitRec.state = ARBIT_NULL; // start state for all nodes
......@@ -2204,7 +2203,6 @@ void Qmgr::initData(Signal* signal)
execARBIT_CFG(signal);
}
setNodeInfo(getOwnNodeId()).m_version = NDB_VERSION;
}//Qmgr::initData()
......@@ -2237,20 +2235,22 @@ void Qmgr::timerHandlingLab(Signal* signal)
hb_check_timer.reset();
}
}
if (interface_check_timer.check(TcurrentTime)) {
jam();
interface_check_timer.reset();
checkStartInterface(signal);
}
if (hb_api_timer.check(TcurrentTime))
{
jam();
hb_api_timer.reset();
apiHbHandlingLab(signal);
}
if (cactivateApiCheck != 0) {
jam();
if (hb_api_timer.check(TcurrentTime)) {
jam();
hb_api_timer.reset();
apiHbHandlingLab(signal);
}//if
if (clatestTransactionCheck == 0) {
//-------------------------------------------------------------
// Initialise the Transaction check timer.
......@@ -2367,18 +2367,21 @@ void Qmgr::apiHbHandlingLab(Signal* signal)
if(type == NodeInfo::INVALID)
continue;
if (TnodePtr.p->phase == ZAPI_ACTIVE){
if (c_connectedNodes.get(nodeId))
{
jam();
setNodeInfo(TnodePtr.i).m_heartbeat_cnt++;
if(getNodeInfo(TnodePtr.i).m_heartbeat_cnt > 2){
if(getNodeInfo(TnodePtr.i).m_heartbeat_cnt > 2)
{
signal->theData[0] = NDB_LE_MissedHeartbeat;
signal->theData[1] = nodeId;
signal->theData[2] = getNodeInfo(TnodePtr.i).m_heartbeat_cnt - 1;
sendSignal(CMVMI_REF, GSN_EVENT_REP, signal, 3, JBB);
}
if (getNodeInfo(TnodePtr.i).m_heartbeat_cnt > 4) {
if (getNodeInfo(TnodePtr.i).m_heartbeat_cnt > 4)
{
jam();
/*------------------------------------------------------------------*/
/* THE API NODE HAS NOT SENT ANY HEARTBEAT FOR THREE SECONDS.
......@@ -2390,8 +2393,8 @@ void Qmgr::apiHbHandlingLab(Signal* signal)
signal->theData[0] = NDB_LE_DeadDueToHeartbeat;
signal->theData[1] = nodeId;
sendSignal(CMVMI_REF, GSN_EVENT_REP, signal, 2, JBB);
node_failed(signal, nodeId);
api_failed(signal, nodeId);
}//if
}//if
}//for
......@@ -2480,26 +2483,6 @@ void Qmgr::sendApiFailReq(Signal* signal, Uint16 failedNodeNo)
sendSignal(DBTC_REF, GSN_API_FAILREQ, signal, 2, JBA);
sendSignal(DBDICT_REF, GSN_API_FAILREQ, signal, 2, JBA);
sendSignal(SUMA_REF, GSN_API_FAILREQ, signal, 2, JBA);
/**-------------------------------------------------------------------------
* THE OTHER NODE WAS AN API NODE. THE COMMUNICATION LINK IS ALREADY
* BROKEN AND THUS NO ACTION IS NEEDED TO BREAK THE CONNECTION.
* WE ONLY NEED TO SET PARAMETERS TO ENABLE A NEW CONNECTION IN A FEW
* SECONDS.
*-------------------------------------------------------------------------*/
setNodeInfo(failedNodePtr.i).m_heartbeat_cnt= 0;
setNodeInfo(failedNodePtr.i).m_version = 0;
recompute_version_info(getNodeInfo(failedNodePtr.i).m_type);
CloseComReqConf * const closeCom = (CloseComReqConf *)&signal->theData[0];
closeCom->xxxBlockRef = reference();
closeCom->failNo = 0;
closeCom->noOfNodes = 1;
NodeBitmask::clear(closeCom->theNodes);
NodeBitmask::set(closeCom->theNodes, failedNodePtr.i);
sendSignal(CMVMI_REF, GSN_CLOSE_COMREQ, signal,
CloseComReqConf::SignalLength, JBA);
}//Qmgr::sendApiFailReq()
void Qmgr::execAPI_FAILREQ(Signal* signal)
......@@ -2512,20 +2495,7 @@ void Qmgr::execAPI_FAILREQ(Signal* signal)
ndbrequire(getNodeInfo(failedNodePtr.i).getType() != NodeInfo::DB);
// ignore if api not active
if (failedNodePtr.p->phase != ZAPI_ACTIVE)
{
jam();
// But send to SUMA anyway...
sendSignal(SUMA_REF, GSN_API_FAILREQ, signal, 2, JBA);
return;
}
signal->theData[0] = NDB_LE_Disconnected;
signal->theData[1] = failedNodePtr.i;
sendSignal(CMVMI_REF, GSN_EVENT_REP, signal, 2, JBB);
node_failed(signal, failedNodePtr.i);
api_failed(signal, signal->theData[0]);
}
void Qmgr::execAPI_FAILCONF(Signal* signal)
......@@ -2649,6 +2619,13 @@ void Qmgr::execDISCONNECT_REP(Signal* signal)
ndbrequire(false);
}
if (getNodeInfo(nodeId).getType() != NodeInfo::DB)
{
jam();
api_failed(signal, nodeId);
return;
}
switch(nodePtr.p->phase){
case ZRUNNING:
jam();
......@@ -2685,66 +2662,109 @@ void Qmgr::node_failed(Signal* signal, Uint16 aFailedNode)
failedNodePtr.i = aFailedNode;
ptrCheckGuard(failedNodePtr, MAX_NODES, nodeRec);
if (getNodeInfo(failedNodePtr.i).getType() == NodeInfo::DB){
ndbrequire(getNodeInfo(failedNodePtr.i).getType() == NodeInfo::DB);
/**---------------------------------------------------------------------
* THE OTHER NODE IS AN NDB NODE, WE HANDLE IT AS IF A HEARTBEAT
* FAILURE WAS DISCOVERED.
*---------------------------------------------------------------------*/
switch(failedNodePtr.p->phase){
case ZRUNNING:
jam();
/**---------------------------------------------------------------------
* THE OTHER NODE IS AN NDB NODE, WE HANDLE IT AS IF A HEARTBEAT
* FAILURE WAS DISCOVERED.
*---------------------------------------------------------------------*/
switch(failedNodePtr.p->phase){
case ZRUNNING:
jam();
failReportLab(signal, aFailedNode, FailRep::ZLINK_FAILURE);
return;
case ZFAIL_CLOSING:
jam();
return;
case ZSTARTING:
c_start.reset();
// Fall-through
default:
jam();
/*---------------------------------------------------------------------*/
// The other node is still not in the cluster but disconnected.
// We must restart communication in three seconds.
/*---------------------------------------------------------------------*/
failedNodePtr.p->failState = NORMAL;
failedNodePtr.p->phase = ZFAIL_CLOSING;
setNodeInfo(failedNodePtr.i).m_heartbeat_cnt= 0;
CloseComReqConf * const closeCom =
(CloseComReqConf *)&signal->theData[0];
closeCom->xxxBlockRef = reference();
closeCom->failNo = 0;
closeCom->noOfNodes = 1;
NodeBitmask::clear(closeCom->theNodes);
NodeBitmask::set(closeCom->theNodes, failedNodePtr.i);
sendSignal(CMVMI_REF, GSN_CLOSE_COMREQ, signal,
CloseComReqConf::SignalLength, JBA);
}//if
failReportLab(signal, aFailedNode, FailRep::ZLINK_FAILURE);
return;
}
/**
* API code
*/
jam();
if (failedNodePtr.p->phase != ZFAIL_CLOSING){
case ZFAIL_CLOSING:
jam();
return;
case ZSTARTING:
c_start.reset();
// Fall-through
default:
jam();
//-------------------------------------------------------------------------
// The API was active and has now failed. We need to initiate API failure
// handling. If the API had already failed then we can ignore this
// discovery.
//-------------------------------------------------------------------------
/*---------------------------------------------------------------------*/
// The other node is still not in the cluster but disconnected.
// We must restart communication in three seconds.
/*---------------------------------------------------------------------*/
failedNodePtr.p->failState = NORMAL;
failedNodePtr.p->phase = ZFAIL_CLOSING;
sendApiFailReq(signal, aFailedNode);
arbitRec.code = ArbitCode::ApiFail;
handleArbitApiFail(signal, aFailedNode);
setNodeInfo(failedNodePtr.i).m_heartbeat_cnt= 0;
CloseComReqConf * const closeCom =
(CloseComReqConf *)&signal->theData[0];
closeCom->xxxBlockRef = reference();
closeCom->failNo = 0;
closeCom->noOfNodes = 1;
NodeBitmask::clear(closeCom->theNodes);
NodeBitmask::set(closeCom->theNodes, failedNodePtr.i);
sendSignal(CMVMI_REF, GSN_CLOSE_COMREQ, signal,
CloseComReqConf::SignalLength, JBA);
}//if
return;
}//Qmgr::node_failed()
}
void
Qmgr::api_failed(Signal* signal, Uint32 nodeId)
{
NodeRecPtr failedNodePtr;
/**------------------------------------------------------------------------
* A COMMUNICATION LINK HAS BEEN DISCONNECTED. WE MUST TAKE SOME ACTION
* DUE TO THIS.
*-----------------------------------------------------------------------*/
failedNodePtr.i = nodeId;
ptrCheckGuard(failedNodePtr, MAX_NODES, nodeRec);
if (failedNodePtr.p->phase == ZFAIL_CLOSING)
{
/**
* Failure handling already in progress
*/
jam();
return;
}
if (failedNodePtr.p->phase == ZAPI_ACTIVE)
{
jam();
sendApiFailReq(signal, nodeId);
arbitRec.code = ArbitCode::ApiFail;
handleArbitApiFail(signal, nodeId);
}
else
{
/**
* Always inform SUMA
*/
jam();
signal->theData[0] = nodeId;
signal->theData[1] = QMGR_REF;
sendSignal(SUMA_REF, GSN_API_FAILREQ, signal, 2, JBA);
failedNodePtr.p->failState = NORMAL;
}
failedNodePtr.p->phase = ZFAIL_CLOSING;
setNodeInfo(failedNodePtr.i).m_heartbeat_cnt= 0;
setNodeInfo(failedNodePtr.i).m_version = 0;
recompute_version_info(getNodeInfo(failedNodePtr.i).m_type);
CloseComReqConf * const closeCom = (CloseComReqConf *)&signal->theData[0];
closeCom->xxxBlockRef = reference();
closeCom->failNo = 0;
closeCom->noOfNodes = 1;
NodeBitmask::clear(closeCom->theNodes);
NodeBitmask::set(closeCom->theNodes, failedNodePtr.i);
sendSignal(CMVMI_REF, GSN_CLOSE_COMREQ, signal,
CloseComReqConf::SignalLength, JBA);
if (getNodeInfo(failedNodePtr.i).getType() == NodeInfo::MGM)
{
/**
* Allow MGM do reconnect "directly"
*/
jam();
setNodeInfo(failedNodePtr.i).m_heartbeat_cnt = 3;
}
}
/**--------------------------------------------------------------------------
* AN API NODE IS REGISTERING. IF FOR THE FIRST TIME WE WILL ENABLE
......@@ -4963,43 +4983,39 @@ Qmgr::execDUMP_STATE_ORD(Signal* signal)
c_start.m_president_candidate_gci);
infoEvent("ctoStatus = %d\n", ctoStatus);
for(Uint32 i = 1; i<MAX_NDB_NODES; i++){
if(getNodeInfo(i).getType() == NodeInfo::DB){
NodeRecPtr nodePtr;
nodePtr.i = i;
ptrCheckGuard(nodePtr, MAX_NDB_NODES, nodeRec);
char buf[100];
switch(nodePtr.p->phase){
case ZINIT:
sprintf(buf, "Node %d: ZINIT(%d)", i, nodePtr.p->phase);
break;
case ZSTARTING:
sprintf(buf, "Node %d: ZSTARTING(%d)", i, nodePtr.p->phase);
break;
case ZRUNNING:
sprintf(buf, "Node %d: ZRUNNING(%d)", i, nodePtr.p->phase);
break;
case ZPREPARE_FAIL:
sprintf(buf, "Node %d: ZPREPARE_FAIL(%d)", i, nodePtr.p->phase);
break;
case ZFAIL_CLOSING:
sprintf(buf, "Node %d: ZFAIL_CLOSING(%d)", i, nodePtr.p->phase);
break;
case ZAPI_INACTIVE:
sprintf(buf, "Node %d: ZAPI_INACTIVE(%d)", i, nodePtr.p->phase);
break;
case ZAPI_ACTIVE:
sprintf(buf, "Node %d: ZAPI_ACTIVE(%d)", i, nodePtr.p->phase);
break;
default:
sprintf(buf, "Node %d: <UNKNOWN>(%d)", i, nodePtr.p->phase);
break;
}
infoEvent(buf);
NodeRecPtr nodePtr;
nodePtr.i = i;
ptrCheckGuard(nodePtr, MAX_NDB_NODES, nodeRec);
char buf[100];
switch(nodePtr.p->phase){
case ZINIT:
sprintf(buf, "Node %d: ZINIT(%d)", i, nodePtr.p->phase);
break;
case ZSTARTING:
sprintf(buf, "Node %d: ZSTARTING(%d)", i, nodePtr.p->phase);
break;
case ZRUNNING:
sprintf(buf, "Node %d: ZRUNNING(%d)", i, nodePtr.p->phase);
break;
case ZPREPARE_FAIL:
sprintf(buf, "Node %d: ZPREPARE_FAIL(%d)", i, nodePtr.p->phase);
break;
case ZFAIL_CLOSING:
sprintf(buf, "Node %d: ZFAIL_CLOSING(%d)", i, nodePtr.p->phase);
break;
case ZAPI_INACTIVE:
sprintf(buf, "Node %d: ZAPI_INACTIVE(%d)", i, nodePtr.p->phase);
break;
case ZAPI_ACTIVE:
sprintf(buf, "Node %d: ZAPI_ACTIVE(%d)", i, nodePtr.p->phase);
break;
default:
sprintf(buf, "Node %d: <UNKNOWN>(%d)", i, nodePtr.p->phase);
break;
}
infoEvent(buf);
}
default:
;
}//switch
}
#ifdef ERROR_INSERT
if (signal->theData[0] == 935 && signal->getLength() == 2)
......
storage/ndb/src/mgmsrv/MgmtSrvr.cpp
View file @ 2aefa6cf
......@@ -119,7 +119,11 @@ operator<<(NdbOut& out, const LogLevel & ll)
void
MgmtSrvr::logLevelThreadRun()
{
while (!_isStopThread) {
while (!_isStopThread)
{
Vector<NodeId> failed_started_nodes;
Vector<EventSubscribeReq> failed_log_level_requests;
/**
* Handle started nodes
*/
......@@ -144,14 +148,15 @@ MgmtSrvr::logLevelThreadRun()
m_started_nodes.unlock();
if (setEventReportingLevelImpl(node, req))
{
ndbout_c("setEventReportingLevelImpl(%d): failed", node);
}
SetLogLevelOrd ord;
ord = m_nodeLogLevel[node];
setNodeLogLevelImpl(node, ord);
{
failed_started_nodes.push_back(node);
}
else
{
SetLogLevelOrd ord;
ord = m_nodeLogLevel[node];
setNodeLogLevelImpl(node, ord);
}
m_started_nodes.lock();
}
}
......@@ -166,17 +171,20 @@ MgmtSrvr::logLevelThreadRun()
if(req.blockRef == 0)
{
req.blockRef = _ownReference;
if (setEventReportingLevelImpl(0, req))
{
ndbout_c("setEventReportingLevelImpl: failed 2!");
}
req.blockRef = _ownReference;
if (setEventReportingLevelImpl(0, req))
{
failed_log_level_requests.push_back(req);
}
}
else
{
SetLogLevelOrd ord;
ord = req;
setNodeLogLevelImpl(req.blockRef, ord);
if (setNodeLogLevelImpl(req.blockRef, ord))
{
failed_log_level_requests.push_back(req);
}
}
m_log_level_requests.lock();
}
......@@ -185,7 +193,28 @@ MgmtSrvr::logLevelThreadRun()
if(!ERROR_INSERTED(10000))
m_event_listner.check_listeners();
NdbSleep_MilliSleep(_logLevelThreadSleep);
Uint32 sleeptime = _logLevelThreadSleep;
if (failed_started_nodes.size())
{
m_started_nodes.lock();
for (Uint32 i = 0; i<failed_started_nodes.size(); i++)
m_started_nodes.push_back(failed_started_nodes[i], false);
m_started_nodes.unlock();
failed_started_nodes.clear();
sleeptime = 100;
}
if (failed_log_level_requests.size())
{
m_log_level_requests.lock();
for (Uint32 i = 0; i<failed_log_level_requests.size(); i++)
m_log_level_requests.push_back(failed_log_level_requests[i], false);
m_log_level_requests.unlock();
failed_log_level_requests.clear();
sleeptime = 100;
}
NdbSleep_MilliSleep(sleeptime);
}
}
......@@ -1535,7 +1564,6 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg,
{
SignalSender ss(theFacade);
NdbNodeBitmask nodes;
int retries = 30;
nodes.clear();
while (1)
{
......@@ -1572,18 +1600,8 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg,
continue;
}
// api_reg_conf not recevied yet, need to retry
break;
}
}
if (nodeId <= max)
{
if (--retries)
{
ss.unlock();
NdbSleep_MilliSleep(100);
continue;
return SEND_OR_RECEIVE_FAILED;
}
return SEND_OR_RECEIVE_FAILED;
}
if (nodeId_arg == 0)
......@@ -1607,6 +1625,10 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg,
continue; // node is not connected, skip
if (ss.sendSignal(nodeId, &ssig) == SEND_OK)
nodes.set(nodeId);
else if (max == nodeId)
{
return SEND_OR_RECEIVE_FAILED;
}
}
break;
}
......@@ -2988,8 +3010,8 @@ int MgmtSrvr::connect_to_self(void)
return 0;
}
template class Vector<unsigned short>;
template class MutexVector<unsigned short>;
template class MutexVector<Ndb_mgmd_event_service::Event_listener>;
template class Vector<EventSubscribeReq>;
template class MutexVector<EventSubscribeReq>;
storage/ndb/src/ndbapi/ndberror.c
View file @ 2aefa6cf
......@@ -169,7 +169,7 @@ ErrorBundle ErrorCodes[] = {
{ 219, DMEC, TR, "219" },
{ 233, DMEC, TR,
"Out of operation records in transaction coordinator (increase MaxNoOfConcurrentOperations)" },
{ 275, DMEC, TR, "275" },
{ 275, DMEC, TR, "Out of transaction records for complete phase (increase MaxNoOfConcurrentTransactions)" },
{ 279, DMEC, TR, "Out of transaction markers in transaction coordinator" },
{ 414, DMEC, TR, "414" },
{ 418, DMEC, TR, "Out of transaction buffers in LQH" },
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7