summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorunknown <jonas@perch.ndb.mysql.com>2006-03-20 18:30:29 +0100
committerunknown <jonas@perch.ndb.mysql.com>2006-03-20 18:30:29 +0100
commit591aedaa2b594fdccca79a09c846cf4b4490f884 (patch)
tree8b6aa7a12a834356ce3c8db74ddafbda1713c74f
parent367442f754a97d87077c99bc5805b41da5ac7119 (diff)
parentad6dcfb1277b3b0a8692c3bfd802ba48cc3fe537 (diff)
downloadmariadb-git-591aedaa2b594fdccca79a09c846cf4b4490f884.tar.gz
Merge joreland@bk-internal.mysql.com:/home/bk/mysql-4.1-wl2610
into perch.ndb.mysql.com:/home/jonas/src/41-work
-rw-r--r--ndb/include/kernel/signaldata/TcContinueB.hpp3
-rw-r--r--ndb/src/kernel/blocks/dbdih/Dbdih.hpp3
-rw-r--r--ndb/src/kernel/blocks/dbdih/DbdihMain.cpp16
-rw-r--r--ndb/src/kernel/blocks/dblqh/DblqhMain.cpp166
-rw-r--r--ndb/src/kernel/blocks/dbtc/Dbtc.hpp27
-rw-r--r--ndb/src/kernel/blocks/dbtc/DbtcMain.cpp368
-rw-r--r--ndb/src/kernel/blocks/qmgr/QmgrMain.cpp101
-rw-r--r--ndb/src/ndbapi/NdbConnection.cpp4
-rw-r--r--ndb/src/ndbapi/Ndbif.cpp12
-rw-r--r--ndb/src/ndbapi/TransporterFacade.cpp13
-rw-r--r--ndb/src/ndbapi/TransporterFacade.hpp1
-rw-r--r--ndb/test/ndbapi/testNodeRestart.cpp50
-rw-r--r--ndb/test/ndbapi/testTimeout.cpp101
-rw-r--r--ndb/test/run-test/daily-basic-tests.txt8
14 files changed, 651 insertions, 222 deletions
diff --git a/ndb/include/kernel/signaldata/TcContinueB.hpp b/ndb/include/kernel/signaldata/TcContinueB.hpp
index 85213791b2a..b87b982e49b 100644
--- a/ndb/include/kernel/signaldata/TcContinueB.hpp
+++ b/ndb/include/kernel/signaldata/TcContinueB.hpp
@@ -44,7 +44,8 @@ private:
CHECK_WAIT_DROP_TAB_FAILED_LQH = 16,
TRIGGER_PENDING = 17,
- DelayTCKEYCONF = 18
+ DelayTCKEYCONF = 18,
+ ZNF_CHECK_TRANSACTIONS = 19
};
};
diff --git a/ndb/src/kernel/blocks/dbdih/Dbdih.hpp b/ndb/src/kernel/blocks/dbdih/Dbdih.hpp
index 0c107e35603..f74c0f36c4d 100644
--- a/ndb/src/kernel/blocks/dbdih/Dbdih.hpp
+++ b/ndb/src/kernel/blocks/dbdih/Dbdih.hpp
@@ -1038,7 +1038,8 @@ private:
void prepareReplicas(FragmentstorePtr regFragptr);
void removeNodeFromStored(Uint32 nodeId,
FragmentstorePtr regFragptr,
- ReplicaRecordPtr replicaPtr);
+ ReplicaRecordPtr replicaPtr,
+ bool temporary);
void removeOldStoredReplica(FragmentstorePtr regFragptr,
ReplicaRecordPtr replicaPtr);
void removeStoredReplica(FragmentstorePtr regFragptr,
diff --git a/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp b/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp
index 776e59ea495..fab428aadef 100644
--- a/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp
+++ b/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp
@@ -5212,6 +5212,7 @@ void Dbdih::removeNodeFromTable(Signal* signal,
//const Uint32 lcpId = SYSFILE->latestLCP_ID;
const bool lcpOngoingFlag = (tabPtr.p->tabLcpStatus== TabRecord::TLS_ACTIVE);
+ const bool temporary = !tabPtr.p->storedTable;
FragmentstorePtr fragPtr;
for(Uint32 fragNo = 0; fragNo < tabPtr.p->totalfragments; fragNo++){
@@ -5232,7 +5233,7 @@ void Dbdih::removeNodeFromTable(Signal* signal,
jam();
found = true;
noOfRemovedReplicas++;
- removeNodeFromStored(nodeId, fragPtr, replicaPtr);
+ removeNodeFromStored(nodeId, fragPtr, replicaPtr, temporary);
if(replicaPtr.p->lcpOngoingFlag){
jam();
/**
@@ -12051,9 +12052,18 @@ void Dbdih::removeDeadNode(NodeRecordPtr removeNodePtr)
/*---------------------------------------------------------------*/
void Dbdih::removeNodeFromStored(Uint32 nodeId,
FragmentstorePtr fragPtr,
- ReplicaRecordPtr replicatePtr)
+ ReplicaRecordPtr replicatePtr,
+ bool temporary)
{
- newCrashedReplica(nodeId, replicatePtr);
+ if (!temporary)
+ {
+ jam();
+ newCrashedReplica(nodeId, replicatePtr);
+ }
+ else
+ {
+ jam();
+ }
removeStoredReplica(fragPtr, replicatePtr);
linkOldStoredReplica(fragPtr, replicatePtr);
ndbrequire(fragPtr.p->storedReplicas != RNIL);
diff --git a/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp b/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp
index ff7e3c32924..0aeeaccd55e 100644
--- a/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp
+++ b/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp
@@ -18448,6 +18448,172 @@ Dblqh::execDUMP_STATE_ORD(Signal* signal)
c_error_insert_table_id = dumpState->args[1];
SET_ERROR_INSERT_VALUE(5042);
}
+
+ TcConnectionrec *regTcConnectionrec = tcConnectionrec;
+ Uint32 ttcConnectrecFileSize = ctcConnectrecFileSize;
+ Uint32 arg = dumpState->args[0];
+ if(arg == 2306)
+ {
+ for(Uint32 i = 0; i<1024; i++)
+ {
+ TcConnectionrecPtr tcRec;
+ tcRec.i = ctransidHash[i];
+ while(tcRec.i != RNIL)
+ {
+ ptrCheckGuard(tcRec, ttcConnectrecFileSize, regTcConnectionrec);
+ ndbout << "TcConnectionrec " << tcRec.i;
+ signal->theData[0] = 2307;
+ signal->theData[1] = tcRec.i;
+ execDUMP_STATE_ORD(signal);
+ tcRec.i = tcRec.p->nextHashRec;
+ }
+ }
+ }
+
+ if(arg == 2307 || arg == 2308)
+ {
+ TcConnectionrecPtr tcRec;
+ tcRec.i = signal->theData[1];
+ ptrCheckGuard(tcRec, ttcConnectrecFileSize, regTcConnectionrec);
+
+ ndbout << " transactionState = " << tcRec.p->transactionState<<endl;
+ ndbout << " operation = " << tcRec.p->operation<<endl;
+ ndbout << " tcNodeFailrec = " << tcRec.p->tcNodeFailrec
+ << " seqNoReplica = " << tcRec.p->seqNoReplica
+ << " simpleRead = " << tcRec.p->simpleRead
+ << endl;
+ ndbout << " replicaType = " << tcRec.p->replicaType
+ << " reclenAiLqhkey = " << tcRec.p->reclenAiLqhkey
+ << " opExec = " << tcRec.p->opExec
+ << endl;
+ ndbout << " opSimple = " << tcRec.p->opSimple
+ << " nextSeqNoReplica = " << tcRec.p->nextSeqNoReplica
+ << " lockType = " << tcRec.p->lockType
+ << endl;
+ ndbout << " lastReplicaNo = " << tcRec.p->lastReplicaNo
+ << " indTakeOver = " << tcRec.p->indTakeOver
+ << " dirtyOp = " << tcRec.p->dirtyOp
+ << endl;
+ ndbout << " activeCreat = " << tcRec.p->activeCreat
+ << " tcBlockref = " << hex << tcRec.p->tcBlockref
+ << " reqBlockref = " << hex << tcRec.p->reqBlockref
+ << " primKeyLen = " << tcRec.p->primKeyLen
+ << endl;
+ ndbout << " nextReplica = " << tcRec.p->nextReplica
+ << " tcBlockref = " << hex << tcRec.p->tcBlockref
+ << " reqBlockref = " << hex << tcRec.p->reqBlockref
+ << " primKeyLen = " << tcRec.p->primKeyLen
+ << endl;
+ ndbout << " logStopPageNo = " << tcRec.p->logStopPageNo
+ << " logStartPageNo = " << tcRec.p->logStartPageNo
+ << " logStartPageIndex = " << tcRec.p->logStartPageIndex
+ << endl;
+ ndbout << " errorCode = " << tcRec.p->errorCode
+ << " clientBlockref = " << hex << tcRec.p->clientBlockref
+ << " applRef = " << hex << tcRec.p->applRef
+ << " totSendlenAi = " << tcRec.p->totSendlenAi
+ << endl;
+ ndbout << " totReclenAi = " << tcRec.p->totReclenAi
+ << " tcScanRec = " << tcRec.p->tcScanRec
+ << " tcScanInfo = " << tcRec.p->tcScanInfo
+ << " tcOprec = " << hex << tcRec.p->tcOprec
+ << endl;
+ ndbout << " tableref = " << tcRec.p->tableref
+ << " simpleTcConnect = " << tcRec.p->simpleTcConnect
+ << " storedProcId = " << tcRec.p->storedProcId
+ << " schemaVersion = " << tcRec.p->schemaVersion
+ << endl;
+ ndbout << " reqinfo = " << tcRec.p->reqinfo
+ << " reqRef = " << tcRec.p->reqRef
+ << " readlenAi = " << tcRec.p->readlenAi
+ << " prevTc = " << tcRec.p->prevTc
+ << endl;
+ ndbout << " prevLogTcrec = " << tcRec.p->prevLogTcrec
+ << " prevHashRec = " << tcRec.p->prevHashRec
+ << " nodeAfterNext0 = " << tcRec.p->nodeAfterNext[0]
+ << " nodeAfterNext1 = " << tcRec.p->nodeAfterNext[1]
+ << endl;
+ ndbout << " nextTcConnectrec = " << tcRec.p->nextTcConnectrec
+ << " nextTc = " << tcRec.p->nextTc
+ << " nextTcLogQueue = " << tcRec.p->nextTcLogQueue
+ << " nextLogTcrec = " << tcRec.p->nextLogTcrec
+ << endl;
+ ndbout << " nextHashRec = " << tcRec.p->nextHashRec
+ << " logWriteState = " << tcRec.p->logWriteState
+ << " logStartFileNo = " << tcRec.p->logStartFileNo
+ << " listState = " << tcRec.p->listState
+ << endl;
+ ndbout << " lastAttrinbuf = " << tcRec.p->lastAttrinbuf
+ << " lastTupkeybuf = " << tcRec.p->lastTupkeybuf
+ << " hashValue = " << tcRec.p->hashValue
+ << endl;
+ ndbout << " gci = " << tcRec.p->gci
+ << " fragmentptr = " << tcRec.p->fragmentptr
+ << " fragmentid = " << tcRec.p->fragmentid
+ << " firstTupkeybuf = " << tcRec.p->firstTupkeybuf
+ << endl;
+ ndbout << " firstAttrinbuf = " << tcRec.p->firstAttrinbuf
+ << " currTupAiLen = " << tcRec.p->currTupAiLen
+ << " currReclenAi = " << tcRec.p->currReclenAi
+ << endl;
+ ndbout << " tcTimer = " << tcRec.p->tcTimer
+ << " clientConnectrec = " << tcRec.p->clientConnectrec
+ << " applOprec = " << hex << tcRec.p->applOprec
+ << " abortState = " << tcRec.p->abortState
+ << endl;
+ ndbout << " transid0 = " << hex << tcRec.p->transid[0]
+ << " transid1 = " << hex << tcRec.p->transid[1]
+ << " tupkeyData0 = " << tcRec.p->tupkeyData[0]
+ << " tupkeyData1 = " << tcRec.p->tupkeyData[1]
+ << endl;
+ ndbout << " tupkeyData2 = " << tcRec.p->tupkeyData[2]
+ << " tupkeyData3 = " << tcRec.p->tupkeyData[3]
+ << endl;
+ switch (tcRec.p->transactionState) {
+
+ case TcConnectionrec::SCAN_STATE_USED:
+ if (tcRec.p->tcScanRec < cscanrecFileSize){
+ ScanRecordPtr TscanPtr;
+ c_scanRecordPool.getPtr(TscanPtr, tcRec.p->tcScanRec);
+ ndbout << " scanState = " << TscanPtr.p->scanState << endl;
+ //TscanPtr.p->scanLocalref[2];
+ ndbout << " copyPtr="<<TscanPtr.p->copyPtr
+ << " scanAccPtr="<<TscanPtr.p->scanAccPtr
+ << " scanAiLength="<<TscanPtr.p->scanAiLength
+ << endl;
+ ndbout << " m_curr_batch_size_rows="<<
+ TscanPtr.p->m_curr_batch_size_rows
+ << " m_max_batch_size_rows="<<
+ TscanPtr.p->m_max_batch_size_rows
+ << " scanErrorCounter="<<TscanPtr.p->scanErrorCounter
+ << endl;
+ ndbout << " scanSchemaVersion="<<TscanPtr.p->scanSchemaVersion
+ << " scanStoredProcId="<<TscanPtr.p->scanStoredProcId
+ << " scanTcrec="<<TscanPtr.p->scanTcrec
+ << endl;
+ ndbout << " scanType="<<TscanPtr.p->scanType
+ << " scanApiBlockref="<<TscanPtr.p->scanApiBlockref
+ << " scanNodeId="<<TscanPtr.p->scanNodeId
+ << " scanCompletedStatus="<<TscanPtr.p->scanCompletedStatus
+ << endl;
+ ndbout << " scanFlag="<<TscanPtr.p->scanFlag
+ << " scanLockHold="<<TscanPtr.p->scanLockHold
+ << " scanLockMode="<<TscanPtr.p->scanLockMode
+ << " scanNumber="<<TscanPtr.p->scanNumber
+ << endl;
+ ndbout << " scanReleaseCounter="<<TscanPtr.p->scanReleaseCounter
+ << " scanTcWaiting="<<TscanPtr.p->scanTcWaiting
+ << " scanKeyinfoFlag="<<TscanPtr.p->scanKeyinfoFlag
+ << endl;
+ } else{
+ ndbout << "No connected scan record found" << endl;
+ }
+ break;
+ default:
+ break;
+ }
+ ndbrequire(arg != 2308);
+ }
}//Dblqh::execDUMP_STATE_ORD()
diff --git a/ndb/src/kernel/blocks/dbtc/Dbtc.hpp b/ndb/src/kernel/blocks/dbtc/Dbtc.hpp
index 61afef30b43..b1332a4fd0b 100644
--- a/ndb/src/kernel/blocks/dbtc/Dbtc.hpp
+++ b/ndb/src/kernel/blocks/dbtc/Dbtc.hpp
@@ -211,14 +211,6 @@ public:
LTS_ACTIVE = 1
};
- enum TakeOverState {
- TOS_NOT_DEFINED = 0,
- TOS_IDLE = 1,
- TOS_ACTIVE = 2,
- TOS_COMPLETED = 3,
- TOS_NODE_FAILED = 4
- };
-
enum FailState {
FS_IDLE = 0,
FS_LISTENING = 1,
@@ -636,6 +628,7 @@ public:
ConnectionState apiConnectstate;
UintR transid[2];
UintR firstTcConnect;
+ NdbNodeBitmask m_transaction_nodes;
//---------------------------------------------------
// Second 16 byte cache line. Hot variables.
@@ -932,7 +925,6 @@ public:
struct HostRecord {
HostState hostStatus;
LqhTransState lqhTransStatus;
- TakeOverState takeOverStatus;
bool inPackedList;
UintR noOfPackedWordsLqh;
UintR packedWordsLqh[26];
@@ -941,6 +933,17 @@ public:
UintR noOfWordsTCINDXCONF;
UintR packedWordsTCINDXCONF[30];
BlockReference hostLqhBlockRef;
+
+ enum NodeFailBits
+ {
+ NF_TAKEOVER = 0x1,
+ NF_CHECK_SCAN = 0x2,
+ NF_CHECK_TRANSACTION = 0x4,
+ NF_CHECK_DROP_TAB = 0x8,
+ NF_NODE_FAIL_BITS = 0xF // All bits...
+ };
+ Uint32 m_nf_bits;
+ NdbNodeBitmask m_lqh_trans_conf;
}; /* p2c: size = 128 bytes */
typedef Ptr<HostRecord> HostRecordPtr;
@@ -1578,7 +1581,7 @@ private:
void wrongSchemaVersionErrorLab(Signal* signal);
void noFreeConnectionErrorLab(Signal* signal);
void tckeyreq050Lab(Signal* signal);
- void timeOutFoundLab(Signal* signal, UintR anAdd);
+ void timeOutFoundLab(Signal* signal, UintR anAdd, Uint32 errCode);
void completeTransAtTakeOverLab(Signal* signal, UintR TtakeOverInd);
void completeTransAtTakeOverDoLast(Signal* signal, UintR TtakeOverInd);
void completeTransAtTakeOverDoOne(Signal* signal, UintR TtakeOverInd);
@@ -1600,6 +1603,9 @@ private:
void checkScanFragList(Signal*, Uint32 failedNodeId, ScanRecord * scanP,
LocalDLList<ScanFragRec>::Head&);
+ void nodeFailCheckTransactions(Signal*,Uint32 transPtrI,Uint32 failedNodeId);
+ void checkNodeFailComplete(Signal* signal, Uint32 failedNodeId, Uint32 bit);
+
// Initialisation
void initData();
void initRecords();
@@ -1626,6 +1632,7 @@ private:
HostRecord *hostRecord;
HostRecordPtr hostptr;
UintR chostFilesize;
+ NdbNodeBitmask c_alive_nodes;
GcpRecord *gcpRecord;
GcpRecordPtr gcpPtr;
diff --git a/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp b/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp
index d9d1f01b213..ff9b279592c 100644
--- a/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp
+++ b/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp
@@ -262,6 +262,10 @@ void Dbtc::execCONTINUEB(Signal* signal)
jam();
checkScanActiveInFailedLqh(signal, Tdata0, Tdata1);
return;
+ case TcContinueB::ZNF_CHECK_TRANSACTIONS:
+ jam();
+ nodeFailCheckTransactions(signal, Tdata0, Tdata1);
+ return;
case TcContinueB::CHECK_WAIT_DROP_TAB_FAILED_LQH:
jam();
checkWaitDropTabFailedLqh(signal, Tdata0, Tdata1);
@@ -299,8 +303,8 @@ void Dbtc::execINCL_NODEREQ(Signal* signal)
hostptr.i = signal->theData[1];
ptrCheckGuard(hostptr, chostFilesize, hostRecord);
hostptr.p->hostStatus = HS_ALIVE;
- hostptr.p->takeOverStatus = TOS_IDLE;
signal->theData[0] = cownref;
+ c_alive_nodes.set(hostptr.i);
sendSignal(tblockref, GSN_INCL_NODECONF, signal, 1, JBB);
}
@@ -487,6 +491,7 @@ Dbtc::checkWaitDropTabFailedLqh(Signal* signal, Uint32 nodeId, Uint32 tableId)
* Finished
*/
jam();
+ checkNodeFailComplete(signal, nodeId, HostRecord::NF_CHECK_DROP_TAB);
return;
}
@@ -850,8 +855,6 @@ void Dbtc::execREAD_NODESCONF(Signal* signal)
hostptr.i = i;
ptrCheckGuard(hostptr, chostFilesize, hostRecord);
- hostptr.p->takeOverStatus = TOS_IDLE;
-
if (NodeBitmask::get(readNodes->inactiveNodes, i)) {
jam();
hostptr.p->hostStatus = HS_DEAD;
@@ -859,6 +862,7 @@ void Dbtc::execREAD_NODESCONF(Signal* signal)
jam();
con_lineNodes++;
hostptr.p->hostStatus = HS_ALIVE;
+ c_alive_nodes.set(i);
}//if
}//if
}//for
@@ -2314,6 +2318,7 @@ void Dbtc::initApiConnectRec(Signal* signal,
regApiPtr->commitAckMarker = RNIL;
regApiPtr->buddyPtr = RNIL;
regApiPtr->currSavePointId = 0;
+ regApiPtr->m_transaction_nodes.clear();
// Trigger data
releaseFiredTriggerData(&regApiPtr->theFiredTriggers),
// Index data
@@ -2921,6 +2926,10 @@ void Dbtc::tckeyreq050Lab(Signal* signal)
signal->theData[0] = TdihConnectptr;
signal->theData[1] = Ttableref;
signal->theData[2] = TdistrHashValue;
+ signal->theData[3] = 0;
+ signal->theData[4] = 0;
+ signal->theData[5] = 0;
+ signal->theData[6] = 0;
/*-------------------------------------------------------------*/
/* FOR EFFICIENCY REASONS WE AVOID THE SIGNAL SENDING HERE AND */
@@ -3098,6 +3107,7 @@ void Dbtc::sendlqhkeyreq(Signal* signal,
TcConnectRecord * const regTcPtr = tcConnectptr.p;
ApiConnectRecord * const regApiPtr = apiConnectptr.p;
CacheRecord * const regCachePtr = cachePtr.p;
+ UintR sig0, sig1, sig2, sig3, sig4, sig5, sig6;
#ifdef ERROR_INSERT
if (ERROR_INSERTED(8002)) {
systemErrorLab(signal);
@@ -3135,6 +3145,9 @@ void Dbtc::sendlqhkeyreq(Signal* signal,
LqhKeyReq::setScanTakeOverFlag(tslrAttrLen, regCachePtr->scanTakeOverInd);
Tdata10 = 0;
+ sig0 = regCachePtr->opSimple;
+ sig1 = regTcPtr->operation;
+ bool simpleRead = (sig1 == ZREAD && sig0 == ZTRUE);
LqhKeyReq::setKeyLen(Tdata10, regCachePtr->keylen);
LqhKeyReq::setLastReplicaNo(Tdata10, regTcPtr->lastReplicaNo);
LqhKeyReq::setLockType(Tdata10, regCachePtr->opLock);
@@ -3144,8 +3157,8 @@ void Dbtc::sendlqhkeyreq(Signal* signal,
LqhKeyReq::setApplicationAddressFlag(Tdata10, 1);
LqhKeyReq::setDirtyFlag(Tdata10, regTcPtr->dirtyOp);
LqhKeyReq::setInterpretedFlag(Tdata10, regCachePtr->opExec);
- LqhKeyReq::setSimpleFlag(Tdata10, regCachePtr->opSimple);
- LqhKeyReq::setOperation(Tdata10, regTcPtr->operation);
+ LqhKeyReq::setSimpleFlag(Tdata10, sig0);
+ LqhKeyReq::setOperation(Tdata10, sig1);
/* -----------------------------------------------------------------------
* Sequential Number of first LQH = 0, bit 22-23
* IF ATTRIBUTE INFORMATION IS SENT IN TCKEYREQ,
@@ -3158,18 +3171,16 @@ void Dbtc::sendlqhkeyreq(Signal* signal,
* ----------------------------------------------------------------------- */
//LqhKeyReq::setAPIVersion(Tdata10, regCachePtr->apiVersionNo);
Uint32 commitAckMarker = regTcPtr->commitAckMarker;
+ const Uint32 noOfLqhs = regTcPtr->noOfNodes;
if(commitAckMarker != RNIL){
jam();
-
LqhKeyReq::setMarkerFlag(Tdata10, 1);
- CommitAckMarker * tmp;
- tmp = m_commitAckMarkerHash.getPtr(commitAckMarker);
+ CommitAckMarker * tmp = m_commitAckMarkerHash.getPtr(commitAckMarker);
/**
* Populate LQH array
*/
- const Uint32 noOfLqhs = regTcPtr->noOfNodes;
tmp->noOfLqhs = noOfLqhs;
for(Uint32 i = 0; i<noOfLqhs; i++){
tmp->lqhNodeId[i] = regTcPtr->tcNodedata[i];
@@ -3180,7 +3191,6 @@ void Dbtc::sendlqhkeyreq(Signal* signal,
/* NO READ LENGTH SENT FROM TC. SEQUENTIAL NUMBER IS 1 AND IT */
/* IS SENT TO A PRIMARY NODE. */
/* ************************************************************> */
- UintR sig0, sig1, sig2, sig3, sig4, sig5, sig6;
LqhKeyReq * const lqhKeyReq = (LqhKeyReq *)signal->getDataPtrSend();
@@ -3204,6 +3214,14 @@ void Dbtc::sendlqhkeyreq(Signal* signal,
sig5 = regTcPtr->clientData;
sig6 = regCachePtr->scanInfo;
+ if (! simpleRead)
+ {
+ regApiPtr->m_transaction_nodes.set(regTcPtr->tcNodedata[0]);
+ regApiPtr->m_transaction_nodes.set(regTcPtr->tcNodedata[1]);
+ regApiPtr->m_transaction_nodes.set(regTcPtr->tcNodedata[2]);
+ regApiPtr->m_transaction_nodes.set(regTcPtr->tcNodedata[3]);
+ }
+
lqhKeyReq->tableSchemaVersion = sig0;
lqhKeyReq->fragmentData = sig1;
lqhKeyReq->transId1 = sig2;
@@ -4587,6 +4605,7 @@ void Dbtc::copyApi(Signal* signal)
UintR TgcpPointer = regTmpApiPtr->gcpPointer;
UintR TgcpFilesize = cgcpFilesize;
UintR TcommitAckMarker = regTmpApiPtr->commitAckMarker;
+ NdbNodeBitmask Tnodes = regTmpApiPtr->m_transaction_nodes;
GcpRecord *localGcpRecord = gcpRecord;
regApiPtr->ndbapiBlockref = regTmpApiPtr->ndbapiBlockref;
@@ -4597,6 +4616,7 @@ void Dbtc::copyApi(Signal* signal)
regApiPtr->transid[1] = Ttransid2;
regApiPtr->lqhkeyconfrec = Tlqhkeyconfrec;
regApiPtr->commitAckMarker = TcommitAckMarker;
+ regApiPtr->m_transaction_nodes = Tnodes;
gcpPtr.i = TgcpPointer;
ptrCheckGuard(gcpPtr, TgcpFilesize, localGcpRecord);
@@ -4607,6 +4627,7 @@ void Dbtc::copyApi(Signal* signal)
regTmpApiPtr->commitAckMarker = RNIL;
regTmpApiPtr->firstTcConnect = RNIL;
regTmpApiPtr->lastTcConnect = RNIL;
+ regTmpApiPtr->m_transaction_nodes.clear();
releaseAllSeizedIndexOperations(regTmpApiPtr);
}//Dbtc::copyApi()
@@ -4865,7 +4886,7 @@ void Dbtc::releaseTransResources(Signal* signal)
TcConnectRecordPtr localTcConnectptr;
UintR TtcConnectFilesize = ctcConnectFilesize;
TcConnectRecord *localTcConnectRecord = tcConnectRecord;
-
+ apiConnectptr.p->m_transaction_nodes.clear();
localTcConnectptr.i = apiConnectptr.p->firstTcConnect;
do {
jam();
@@ -5269,7 +5290,8 @@ void Dbtc::execTC_COMMITREQ(Signal* signal)
break;
case CS_ABORTING:
jam();
- errorCode = ZABORTINPROGRESS;
+ errorCode = regApiPtr->returncode ?
+ regApiPtr->returncode : ZABORTINPROGRESS;
break;
case CS_START_SCAN:
jam();
@@ -5808,9 +5830,9 @@ void Dbtc::abort010Lab(Signal* signal)
if (transP->firstTcConnect == RNIL) {
jam();
- /*-----------------------------------------------------------------------*/
- /* WE HAVE NO PARTICIPANTS IN THE TRANSACTION. */
- /*-----------------------------------------------------------------------*/
+ /*--------------------------------------------------------------------*/
+ /* WE HAVE NO PARTICIPANTS IN THE TRANSACTION. */
+ /*--------------------------------------------------------------------*/
releaseAbortResources(signal);
return;
}//if
@@ -6087,10 +6109,12 @@ void Dbtc::timeOutLoopStartLab(Signal* signal, Uint32 api_con_ptr)
if (api_timer != 0) {
time_out_value= time_out_param + (api_con_ptr & mask_value);
time_passed= tc_timer - api_timer;
- if (time_passed > time_out_value) {
+ if (time_passed > time_out_value)
+ {
jam();
- timeOutFoundLab(signal, api_con_ptr);
- return;
+ timeOutFoundLab(signal, api_con_ptr, ZTIME_OUT_ERROR);
+ api_con_ptr++;
+ break;
}
}
}
@@ -6110,10 +6134,8 @@ void Dbtc::timeOutLoopStartLab(Signal* signal, Uint32 api_con_ptr)
return;
}//Dbtc::timeOutLoopStartLab()
-void Dbtc::timeOutFoundLab(Signal* signal, Uint32 TapiConPtr)
+void Dbtc::timeOutFoundLab(Signal* signal, Uint32 TapiConPtr, Uint32 errCode)
{
- sendContinueTimeOutControl(signal, TapiConPtr + 1);
-
apiConnectptr.i = TapiConPtr;
ptrCheckGuard(apiConnectptr, capiConnectFilesize, apiConnectRecord);
/*------------------------------------------------------------------*/
@@ -6126,7 +6148,8 @@ void Dbtc::timeOutFoundLab(Signal* signal, Uint32 TapiConPtr)
<< "Time-out in state = " << apiConnectptr.p->apiConnectstate
<< " apiConnectptr.i = " << apiConnectptr.i
<< " - exec: " << apiConnectptr.p->m_exec_flag
- << " - place: " << c_apiConTimer_line[apiConnectptr.i]);
+ << " - place: " << c_apiConTimer_line[apiConnectptr.i]
+ << " code: " << errCode);
switch (apiConnectptr.p->apiConnectstate) {
case CS_STARTED:
if(apiConnectptr.p->lqhkeyreqrec == apiConnectptr.p->lqhkeyconfrec){
@@ -6143,7 +6166,7 @@ void Dbtc::timeOutFoundLab(Signal* signal, Uint32 TapiConPtr)
}//if
}
apiConnectptr.p->returnsignal = RS_TCROLLBACKREP;
- apiConnectptr.p->returncode = ZTIME_OUT_ERROR;
+ apiConnectptr.p->returncode = errCode;
abort010Lab(signal);
return;
case CS_RECEIVING:
@@ -6156,7 +6179,7 @@ void Dbtc::timeOutFoundLab(Signal* signal, Uint32 TapiConPtr)
/* START ABORTING THE TRANSACTION. ALSO START CHECKING THE */
/* REMAINING TRANSACTIONS. */
/*------------------------------------------------------------------*/
- terrorCode = ZTIME_OUT_ERROR;
+ terrorCode = errCode;
abortErrorLab(signal);
return;
case CS_COMMITTING:
@@ -6800,58 +6823,44 @@ void Dbtc::execNODE_FAILREP(Signal* signal)
const Uint32 tnewMasterId = nodeFail->masterNodeId;
arrGuard(tnoOfNodes, MAX_NDB_NODES);
+ Uint32 i;
int index = 0;
- for (unsigned i = 1; i< MAX_NDB_NODES; i++) {
- if(NodeBitmask::get(nodeFail->theNodes, i)){
+ for (i = 1; i< MAX_NDB_NODES; i++)
+ {
+ if(NodeBitmask::get(nodeFail->theNodes, i))
+ {
cdata[index] = i;
index++;
}//if
}//for
+ cmasterNodeId = tnewMasterId;
+
tcNodeFailptr.i = 0;
ptrAss(tcNodeFailptr, tcFailRecord);
- Uint32 tindex;
- for (tindex = 0; tindex < tnoOfNodes; tindex++) {
+ for (i = 0; i < tnoOfNodes; i++)
+ {
jam();
- hostptr.i = cdata[tindex];
+ hostptr.i = cdata[i];
ptrCheckGuard(hostptr, chostFilesize, hostRecord);
+
/*------------------------------------------------------------*/
/* SET STATUS OF THE FAILED NODE TO DEAD SINCE IT HAS */
/* FAILED. */
/*------------------------------------------------------------*/
hostptr.p->hostStatus = HS_DEAD;
+ hostptr.p->m_nf_bits = HostRecord::NF_NODE_FAIL_BITS;
+ c_alive_nodes.clear(hostptr.i);
- if (hostptr.p->takeOverStatus == TOS_COMPLETED) {
- jam();
- /*------------------------------------------------------------*/
- /* A VERY UNUSUAL SITUATION. THE TAKE OVER WAS COMPLETED*/
- /* EVEN BEFORE WE HEARD ABOUT THE NODE FAILURE REPORT. */
- /* HOWEVER UNUSUAL THIS SITUATION IS POSSIBLE. */
- /*------------------------------------------------------------*/
- /* RELEASE THE CURRENTLY UNUSED LQH CONNECTIONS. THE */
- /* REMAINING WILL BE RELEASED WHEN THE TRANSACTION THAT */
- /* USED THEM IS COMPLETED. */
- /*------------------------------------------------------------*/
- {
- NFCompleteRep * const nfRep = (NFCompleteRep *)&signal->theData[0];
- nfRep->blockNo = DBTC;
- nfRep->nodeId = cownNodeid;
- nfRep->failedNodeId = hostptr.i;
- }
- sendSignal(cdihblockref, GSN_NF_COMPLETEREP, signal,
- NFCompleteRep::SignalLength, JBB);
- } else {
- ndbrequire(hostptr.p->takeOverStatus == TOS_IDLE);
- hostptr.p->takeOverStatus = TOS_NODE_FAILED;
- }//if
-
- if (tcNodeFailptr.p->failStatus == FS_LISTENING) {
+ if (tcNodeFailptr.p->failStatus == FS_LISTENING)
+ {
jam();
/*------------------------------------------------------------*/
/* THE CURRENT TAKE OVER CAN BE AFFECTED BY THIS NODE */
/* FAILURE. */
/*------------------------------------------------------------*/
- if (hostptr.p->lqhTransStatus == LTS_ACTIVE) {
+ if (hostptr.p->lqhTransStatus == LTS_ACTIVE)
+ {
jam();
/*------------------------------------------------------------*/
/* WE WERE WAITING FOR THE FAILED NODE IN THE TAKE OVER */
@@ -6863,86 +6872,46 @@ void Dbtc::execNODE_FAILREP(Signal* signal)
}//if
}//if
- }//for
-
- const bool masterFailed = (cmasterNodeId != tnewMasterId);
- cmasterNodeId = tnewMasterId;
-
- if(getOwnNodeId() == cmasterNodeId && masterFailed){
- /**
- * Master has failed and I'm the new master
- */
- jam();
-
- for (hostptr.i = 1; hostptr.i < MAX_NDB_NODES; hostptr.i++) {
+ if (getOwnNodeId() != tnewMasterId)
+ {
jam();
- ptrAss(hostptr, hostRecord);
- if (hostptr.p->hostStatus != HS_ALIVE) {
- jam();
- if (hostptr.p->takeOverStatus == TOS_COMPLETED) {
- jam();
- /*------------------------------------------------------------*/
- /* SEND TAKE OVER CONFIRMATION TO ALL ALIVE NODES IF */
- /* TAKE OVER IS COMPLETED. THIS IS PERFORMED TO ENSURE */
- /* THAT ALL NODES AGREE ON THE IDLE STATE OF THE TAKE */
- /* OVER. THIS MIGHT BE MISSED IN AN ERROR SITUATION IF */
- /* MASTER FAILS AFTER SENDING CONFIRMATION TO NEW */
- /* MASTER BUT FAILING BEFORE SENDING TO ANOTHER NODE */
- /* WHICH WAS NOT MASTER. IF THIS NODE LATER BECOMES */
- /* MASTER IT MIGHT START A NEW TAKE OVER EVEN AFTER THE */
- /* CRASHED NODE HAVE ALREADY RECOVERED. */
- /*------------------------------------------------------------*/
- for(tmpHostptr.i = 1; tmpHostptr.i < MAX_NDB_NODES;tmpHostptr.i++) {
- jam();
- ptrAss(tmpHostptr, hostRecord);
- if (tmpHostptr.p->hostStatus == HS_ALIVE) {
- jam();
- tblockref = calcTcBlockRef(tmpHostptr.i);
- signal->theData[0] = hostptr.i;
- sendSignal(tblockref, GSN_TAKE_OVERTCCONF, signal, 1, JBB);
- }//if
- }//for
- }//if
- }//if
- }//for
- }
-
- if(getOwnNodeId() == cmasterNodeId){
- jam();
- for (hostptr.i = 1; hostptr.i < MAX_NDB_NODES; hostptr.i++) {
+ /**
+ * Only master does takeover currently
+ */
+ hostptr.p->m_nf_bits &= ~HostRecord::NF_TAKEOVER;
+ }
+ else
+ {
jam();
- ptrAss(hostptr, hostRecord);
- if (hostptr.p->hostStatus != HS_ALIVE) {
- jam();
- if (hostptr.p->takeOverStatus == TOS_NODE_FAILED) {
- jam();
- /*------------------------------------------------------------*/
- /* CONCLUDE ALL ACTIVITIES THE FAILED TC DID CONTROL */
- /* SINCE WE ARE THE MASTER. THIS COULD HAVE BEEN STARTED*/
- /* BY A PREVIOUS MASTER BUT HAVE NOT BEEN CONCLUDED YET.*/
- /*------------------------------------------------------------*/
- hostptr.p->takeOverStatus = TOS_ACTIVE;
- signal->theData[0] = hostptr.i;
- sendSignal(cownref, GSN_TAKE_OVERTCREQ, signal, 1, JBB);
- }//if
- }//if
- }//for
- }//if
- for (tindex = 0; tindex < tnoOfNodes; tindex++) {
- jam();
- hostptr.i = cdata[tindex];
- ptrCheckGuard(hostptr, chostFilesize, hostRecord);
- /*------------------------------------------------------------*/
- /* LOOP THROUGH AND ABORT ALL SCANS THAT WHERE */
- /* CONTROLLED BY THIS TC AND ACTIVE IN THE FAILED */
- /* NODE'S LQH */
- /*------------------------------------------------------------*/
+ signal->theData[0] = hostptr.i;
+ sendSignal(cownref, GSN_TAKE_OVERTCREQ, signal, 1, JBB);
+ }
+
checkScanActiveInFailedLqh(signal, 0, hostptr.i);
checkWaitDropTabFailedLqh(signal, hostptr.i, 0); // nodeid, tableid
- }//for
-
+ nodeFailCheckTransactions(signal, 0, hostptr.i);
+ }
}//Dbtc::execNODE_FAILREP()
+void
+Dbtc::checkNodeFailComplete(Signal* signal,
+ Uint32 failedNodeId,
+ Uint32 bit)
+{
+ hostptr.i = failedNodeId;
+ ptrCheckGuard(hostptr, chostFilesize, hostRecord);
+ hostptr.p->m_nf_bits &= ~bit;
+ if (hostptr.p->m_nf_bits == 0)
+ {
+ NFCompleteRep * const nfRep = (NFCompleteRep *)&signal->theData[0];
+ nfRep->blockNo = DBTC;
+ nfRep->nodeId = cownNodeid;
+ nfRep->failedNodeId = hostptr.i;
+ sendSignal(cdihblockref, GSN_NF_COMPLETEREP, signal,
+ NFCompleteRep::SignalLength, JBB);
+ }
+}
+
void Dbtc::checkScanActiveInFailedLqh(Signal* signal,
Uint32 scanPtrI,
Uint32 failedNodeId){
@@ -6984,8 +6953,44 @@ void Dbtc::checkScanActiveInFailedLqh(Signal* signal,
sendSignal(cownref, GSN_CONTINUEB, signal, 3, JBB);
return;
}//for
+
+ checkNodeFailComplete(signal, failedNodeId, HostRecord::NF_CHECK_SCAN);
+}
+
+void
+Dbtc::nodeFailCheckTransactions(Signal* signal,
+ Uint32 transPtrI,
+ Uint32 failedNodeId)
+{
+ jam();
+ Ptr<ApiConnectRecord> transPtr;
+ for (transPtr.i = transPtrI; transPtr.i < capiConnectFilesize; transPtr.i++)
+ {
+ ptrCheckGuard(transPtr, capiConnectFilesize, apiConnectRecord);
+ if (transPtr.p->m_transaction_nodes.get(failedNodeId))
+ {
+ jam();
+ // Force timeout regardless of state
+ Uint32 save = c_appl_timeout_value;
+ c_appl_timeout_value = 1;
+ setApiConTimer(transPtr.i, 0, __LINE__);
+ timeOutFoundLab(signal, transPtr.i, ZNODEFAIL_BEFORE_COMMIT);
+ c_appl_timeout_value = save;
+ }
+
+ // Send CONTINUEB to continue later
+ signal->theData[0] = TcContinueB::ZNF_CHECK_TRANSACTIONS;
+ signal->theData[1] = transPtr.i + 1; // Check next
+ signal->theData[2] = failedNodeId;
+ sendSignal(cownref, GSN_CONTINUEB, signal, 3, JBB);
+ return;
+ }
+
+ checkNodeFailComplete(signal, failedNodeId,
+ HostRecord::NF_CHECK_TRANSACTION);
}
+
void
Dbtc::checkScanFragList(Signal* signal,
Uint32 failedNodeId,
@@ -7001,54 +7006,17 @@ void Dbtc::execTAKE_OVERTCCONF(Signal* signal)
tfailedNodeId = signal->theData[0];
hostptr.i = tfailedNodeId;
ptrCheckGuard(hostptr, chostFilesize, hostRecord);
- switch (hostptr.p->takeOverStatus) {
- case TOS_IDLE:
- jam();
- /*------------------------------------------------------------*/
- /* THIS MESSAGE ARRIVED EVEN BEFORE THE NODE_FAILREP */
- /* MESSAGE. THIS IS POSSIBLE IN EXTREME SITUATIONS. */
- /* WE SET THE STATE TO TAKE_OVER_COMPLETED AND WAIT */
- /* FOR THE NODE_FAILREP MESSAGE. */
- /*------------------------------------------------------------*/
- hostptr.p->takeOverStatus = TOS_COMPLETED;
- break;
- case TOS_NODE_FAILED:
- case TOS_ACTIVE:
- jam();
- /*------------------------------------------------------------*/
- /* WE ARE NOT MASTER AND THE TAKE OVER IS ACTIVE OR WE */
- /* ARE MASTER AND THE TAKE OVER IS ACTIVE. IN BOTH */
- /* WE SET THE STATE TO TAKE_OVER_COMPLETED. */
- /*------------------------------------------------------------*/
- /* RELEASE THE CURRENTLY UNUSED LQH CONNECTIONS. THE */
- /* REMAINING WILL BE RELEASED WHEN THE TRANSACTION THAT */
- /* USED THEM IS COMPLETED. */
- /*------------------------------------------------------------*/
- hostptr.p->takeOverStatus = TOS_COMPLETED;
- {
- NFCompleteRep * const nfRep = (NFCompleteRep *)&signal->theData[0];
- nfRep->blockNo = DBTC;
- nfRep->nodeId = cownNodeid;
- nfRep->failedNodeId = hostptr.i;
- }
- sendSignal(cdihblockref, GSN_NF_COMPLETEREP, signal,
- NFCompleteRep::SignalLength, JBB);
- break;
- case TOS_COMPLETED:
- jam();
- /*------------------------------------------------------------*/
- /* WE HAVE ALREADY RECEIVED THE CONF SIGNAL. IT IS MOST */
- /* LIKELY SENT FROM A NEW MASTER WHICH WASN'T SURE IF */
- /* THIS NODE HEARD THE CONF SIGNAL FROM THE OLD MASTER. */
- /* WE SIMPLY IGNORE THE MESSAGE. */
- /*------------------------------------------------------------*/
- /*empty*/;
- break;
- default:
+
+ ndbout_c("received execTAKE_OVERTCCONF(%d) from %x (%x)",
+ tfailedNodeId, signal->getSendersBlockRef(), reference());
+ if (signal->getSendersBlockRef() != reference())
+ {
jam();
- systemErrorLab(signal);
return;
- }//switch
+ }
+
+
+ checkNodeFailComplete(signal, hostptr.i, HostRecord::NF_TAKEOVER);
}//Dbtc::execTAKE_OVERTCCONF()
void Dbtc::execTAKE_OVERTCREQ(Signal* signal)
@@ -7288,16 +7256,10 @@ void Dbtc::completeTransAtTakeOverDoLast(Signal* signal, UintR TtakeOverInd)
/* TO REPORT THE COMPLETION OF THE TAKE OVER TO ALL */
/* NODES THAT ARE ALIVE. */
/*------------------------------------------------------------*/
- for (hostptr.i = 1; hostptr.i < MAX_NDB_NODES; hostptr.i++) {
- jam();
- ptrAss(hostptr, hostRecord);
- if (hostptr.p->hostStatus == HS_ALIVE) {
- jam();
- tblockref = calcTcBlockRef(hostptr.i);
- signal->theData[0] = tcNodeFailptr.p->takeOverNode;
- sendSignal(tblockref, GSN_TAKE_OVERTCCONF, signal, 1, JBB);
- }//if
- }//for
+ NodeReceiverGroup rg(DBTC, c_alive_nodes);
+ signal->theData[0] = tcNodeFailptr.p->takeOverNode;
+ sendSignal(rg, GSN_TAKE_OVERTCCONF, signal, 1, JBB);
+
if (tcNodeFailptr.p->queueIndex > 0) {
jam();
/*------------------------------------------------------------*/
@@ -7979,6 +7941,7 @@ void Dbtc::initApiConnectFail(Signal* signal)
apiConnectptr.p->ndbapiBlockref = 0;
apiConnectptr.p->ndbapiConnect = 0;
apiConnectptr.p->buddyPtr = RNIL;
+ apiConnectptr.p->m_transaction_nodes.clear();
setApiConTimer(apiConnectptr.i, 0, __LINE__);
switch(ttransStatus){
case LqhTransConf::Committed:
@@ -9756,6 +9719,7 @@ void Dbtc::initApiConnect(Signal* signal)
apiConnectptr.p->executingIndexOp = RNIL;
apiConnectptr.p->buddyPtr = RNIL;
apiConnectptr.p->currSavePointId = 0;
+ apiConnectptr.p->m_transaction_nodes.clear();
}//for
apiConnectptr.i = tiacTmp - 1;
ptrCheckGuard(apiConnectptr, capiConnectFilesize, apiConnectRecord);
@@ -9783,6 +9747,7 @@ void Dbtc::initApiConnect(Signal* signal)
apiConnectptr.p->executingIndexOp = RNIL;
apiConnectptr.p->buddyPtr = RNIL;
apiConnectptr.p->currSavePointId = 0;
+ apiConnectptr.p->m_transaction_nodes.clear();
}//for
apiConnectptr.i = (2 * tiacTmp) - 1;
ptrCheckGuard(apiConnectptr, capiConnectFilesize, apiConnectRecord);
@@ -9810,6 +9775,7 @@ void Dbtc::initApiConnect(Signal* signal)
apiConnectptr.p->executingIndexOp = RNIL;
apiConnectptr.p->buddyPtr = RNIL;
apiConnectptr.p->currSavePointId = 0;
+ apiConnectptr.p->m_transaction_nodes.clear();
}//for
apiConnectptr.i = (3 * tiacTmp) - 1;
ptrCheckGuard(apiConnectptr, capiConnectFilesize, apiConnectRecord);
@@ -9870,13 +9836,13 @@ void Dbtc::inithost(Signal* signal)
ptrAss(hostptr, hostRecord);
hostptr.p->hostStatus = HS_DEAD;
hostptr.p->inPackedList = false;
- hostptr.p->takeOverStatus = TOS_NOT_DEFINED;
hostptr.p->lqhTransStatus = LTS_IDLE;
hostptr.p->noOfWordsTCKEYCONF = 0;
hostptr.p->noOfWordsTCINDXCONF = 0;
hostptr.p->noOfPackedWordsLqh = 0;
hostptr.p->hostLqhBlockRef = calcLqhBlockRef(hostptr.i);
}//for
+ c_alive_nodes.clear();
}//Dbtc::inithost()
void Dbtc::initialiseRecordsLab(Signal* signal, UintR Tdata0,
@@ -10126,6 +10092,7 @@ void Dbtc::releaseAbortResources(Signal* signal)
}//while
apiConnectptr.p->firstTcConnect = RNIL;
apiConnectptr.p->lastTcConnect = RNIL;
+ apiConnectptr.p->m_transaction_nodes.clear();
// MASV let state be CS_ABORTING until all
// signals in the "air" have been received. Reset to CS_CONNECTED
@@ -10199,6 +10166,7 @@ void Dbtc::releaseApiCon(Signal* signal, UintR TapiConnectPtr)
cfirstfreeApiConnect = TlocalApiConnectptr.i;
setApiConTimer(TlocalApiConnectptr.i, 0, __LINE__);
TlocalApiConnectptr.p->apiConnectstate = CS_DISCONNECTED;
+ ndbassert(TlocalApiConnectptr.p->m_transaction_nodes.isclear());
ndbassert(TlocalApiConnectptr.p->apiScanRec == RNIL);
TlocalApiConnectptr.p->ndbapiBlockref = 0;
}//Dbtc::releaseApiCon()
@@ -10734,6 +10702,34 @@ Dbtc::execDUMP_STATE_ORD(Signal* signal)
c_theIndexOperationPool.getSize(),
c_theIndexOperationPool.getNoOfFree());
}
+
+ if (dumpState->args[0] == 2514)
+ {
+ if (signal->getLength() == 2)
+ {
+ dumpState->args[0] = DumpStateOrd::TcDumpOneApiConnectRec;
+ execDUMP_STATE_ORD(signal);
+ }
+
+ NodeReceiverGroup rg(CMVMI, c_alive_nodes);
+ dumpState->args[0] = 15;
+ sendSignal(rg, GSN_DUMP_STATE_ORD, signal, 1, JBB);
+
+ signal->theData[0] = 2515;
+ sendSignalWithDelay(cownref, GSN_DUMP_STATE_ORD, signal, 1000, 1);
+ return;
+ }
+
+ if (dumpState->args[0] == 2515)
+ {
+ NdbNodeBitmask mask = c_alive_nodes;
+ mask.clear(getOwnNodeId());
+ NodeReceiverGroup rg(NDBCNTR, mask);
+
+ sendSignal(rg, GSN_SYSTEM_ERROR, signal, 1, JBB);
+ sendSignalWithDelay(cownref, GSN_SYSTEM_ERROR, signal, 300, 1);
+ return;
+ }
}//Dbtc::execDUMP_STATE_ORD()
void Dbtc::execSET_VAR_REQ(Signal* signal)
diff --git a/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp b/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp
index 6095895e7c2..70084e6b171 100644
--- a/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp
+++ b/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp
@@ -257,6 +257,7 @@ void Qmgr::setArbitTimeout(UintR aArbitTimeout)
void Qmgr::execCONNECT_REP(Signal* signal)
{
+ jamEntry();
const Uint32 nodeId = signal->theData[0];
c_connectedNodes.set(nodeId);
NodeRecPtr nodePtr;
@@ -264,9 +265,13 @@ void Qmgr::execCONNECT_REP(Signal* signal)
ptrCheckGuard(nodePtr, MAX_NODES, nodeRec);
switch(nodePtr.p->phase){
case ZSTARTING:
+ case ZRUNNING:
jam();
+ if(!c_start.m_nodes.isWaitingFor(nodeId)){
+ jam();
+ return;
+ }
break;
- case ZRUNNING:
case ZPREPARE_FAIL:
case ZFAIL_CLOSING:
jam();
@@ -277,21 +282,28 @@ void Qmgr::execCONNECT_REP(Signal* signal)
case ZAPI_INACTIVE:
return;
}
-
- if(!c_start.m_nodes.isWaitingFor(nodeId)){
- jam();
- return;
- }
-
+
switch(c_start.m_gsn){
case GSN_CM_REGREQ:
jam();
sendCmRegReq(signal, nodeId);
return;
- case GSN_CM_NODEINFOREQ:{
+ case GSN_CM_NODEINFOREQ:
jam();
sendCmNodeInfoReq(signal, nodeId, nodePtr.p);
return;
+ case GSN_CM_ADD:{
+ jam();
+
+ ndbrequire(getOwnNodeId() != cpresident);
+ c_start.m_nodes.clearWaitingFor(nodeId);
+ c_start.m_gsn = RNIL;
+
+ NodeRecPtr addNodePtr;
+ addNodePtr.i = nodeId;
+ ptrCheckGuard(addNodePtr, MAX_NDB_NODES, nodeRec);
+ cmAddPrepare(signal, addNodePtr, nodePtr.p);
+ return;
}
default:
return;
@@ -924,15 +936,27 @@ Qmgr::cmAddPrepare(Signal* signal, NodeRecPtr nodePtr, const NodeRec * self){
return;
case ZFAIL_CLOSING:
jam();
-#ifdef VM_TRACE
- ndbout_c("Enabling communication to CM_ADD node state=%d",
- nodePtr.p->phase);
-#endif
+
+#if 1
+ warningEvent("Recieved request to incorperate node %u, "
+ "while error handling has not yet completed",
+ nodePtr.i);
+
+ ndbrequire(getOwnNodeId() != cpresident);
+ ndbrequire(signal->header.theVerId_signalNumber == GSN_CM_ADD);
+ c_start.m_nodes.clearWaitingFor();
+ c_start.m_nodes.setWaitingFor(nodePtr.i);
+ c_start.m_gsn = GSN_CM_ADD;
+#else
+ warningEvent("Enabling communication to CM_ADD node %u state=%d",
+ nodePtr.i,
+ nodePtr.p->phase);
nodePtr.p->phase = ZSTARTING;
nodePtr.p->failState = NORMAL;
signal->theData[0] = 0;
signal->theData[1] = nodePtr.i;
sendSignal(CMVMI_REF, GSN_OPEN_COMREQ, signal, 2, JBA);
+#endif
return;
case ZSTARTING:
break;
@@ -1766,11 +1790,27 @@ void Qmgr::execNDB_FAILCONF(Signal* signal)
jamEntry();
failedNodePtr.i = signal->theData[0];
+
+ if (ERROR_INSERTED(930))
+ {
+ CLEAR_ERROR_INSERT_VALUE;
+ infoEvent("Discarding NDB_FAILCONF for %u", failedNodePtr.i);
+ return;
+ }
+
ptrCheckGuard(failedNodePtr, MAX_NODES, nodeRec);
if (failedNodePtr.p->failState == WAITING_FOR_NDB_FAILCONF){
failedNodePtr.p->failState = NORMAL;
} else {
jam();
+
+ char buf[100];
+ BaseString::snprintf(buf, 100,
+ "Received NDB_FAILCONF for node %u with state: %d %d",
+ failedNodePtr.i,
+ failedNodePtr.p->phase,
+ failedNodePtr.p->failState);
+ progError(__LINE__, 0, buf);
systemErrorLab(signal, __LINE__);
}//if
if (cpresident == getOwnNodeId()) {
@@ -2077,10 +2117,42 @@ void Qmgr::failReportLab(Signal* signal, Uint16 aFailedNode,
ptrCheckGuard(failedNodePtr, MAX_NODES, nodeRec);
if (failedNodePtr.i == getOwnNodeId()) {
jam();
- systemErrorLab(signal, __LINE__);
+
+ const char * msg = 0;
+ switch(aFailCause){
+ case FailRep::ZOWN_FAILURE:
+ msg = "Own failure";
+ break;
+ case FailRep::ZOTHER_NODE_WHEN_WE_START:
+ case FailRep::ZOTHERNODE_FAILED_DURING_START:
+ msg = "Other node died during start";
+ break;
+ case FailRep::ZIN_PREP_FAIL_REQ:
+ msg = "Prep fail";
+ break;
+ case FailRep::ZSTART_IN_REGREQ:
+ msg = "Start timeout";
+ break;
+ case FailRep::ZHEARTBEAT_FAILURE:
+ msg = "Hearbeat failure";
+ break;
+ case FailRep::ZLINK_FAILURE:
+ msg = "Connection failure";
+ break;
+ }
+
+ char buf[100];
+ BaseString::snprintf(buf, 100,
+ "We(%u) have been declared dead by %u reason: %s(%u)",
+ getOwnNodeId(),
+ refToNode(signal->getSendersBlockRef()),
+ aFailCause,
+ msg ? msg : "<Unknown>");
+
+ progError(__LINE__, 0, buf);
return;
}//if
-
+
myNodePtr.i = getOwnNodeId();
ptrCheckGuard(myNodePtr, MAX_NDB_NODES, nodeRec);
if (myNodePtr.p->phase != ZRUNNING) {
@@ -2791,6 +2863,7 @@ void Qmgr::failReport(Signal* signal,
cfailureNr = cprepareFailureNr;
ctoFailureNr = 0;
ctoStatus = Q_ACTIVE;
+ c_start.reset(); // Don't take over nodes being started
if (cnoCommitFailedNodes > 0) {
jam();
/**-----------------------------------------------------------------
diff --git a/ndb/src/ndbapi/NdbConnection.cpp b/ndb/src/ndbapi/NdbConnection.cpp
index c9e26f8ccaf..9cd7d6ed42e 100644
--- a/ndb/src/ndbapi/NdbConnection.cpp
+++ b/ndb/src/ndbapi/NdbConnection.cpp
@@ -450,12 +450,12 @@ NdbConnection::executeNoBlobs(ExecType aTypeOfExec,
//------------------------------------------------------------------------
Ndb* tNdb = theNdb;
+ Uint32 timeout = TransporterFacade::instance()->m_waitfor_timeout;
m_waitForReply = false;
executeAsynchPrepare(aTypeOfExec, NULL, NULL, abortOption);
if (m_waitForReply){
while (1) {
- int noOfComp = tNdb->sendPollNdb((3 * WAITFOR_RESPONSE_TIMEOUT),
- 1, forceSend);
+ int noOfComp = tNdb->sendPollNdb(3 * timeout, 1, forceSend);
if (noOfComp == 0) {
/**
* This timeout situation can occur if NDB crashes.
diff --git a/ndb/src/ndbapi/Ndbif.cpp b/ndb/src/ndbapi/Ndbif.cpp
index 3ebba7e1c4a..d753117aa9a 100644
--- a/ndb/src/ndbapi/Ndbif.cpp
+++ b/ndb/src/ndbapi/Ndbif.cpp
@@ -954,23 +954,25 @@ Ndb::pollCompleted(NdbConnection** aCopyArray)
void
Ndb::check_send_timeout()
{
+ Uint32 timeout = TransporterFacade::instance()->m_waitfor_timeout;
NDB_TICKS current_time = NdbTick_CurrentMillisecond();
if (current_time - the_last_check_time > 1000) {
the_last_check_time = current_time;
Uint32 no_of_sent = theNoOfSentTransactions;
for (Uint32 i = 0; i < no_of_sent; i++) {
NdbConnection* a_con = theSentTransactionsArray[i];
- if ((current_time - a_con->theStartTransTime) >
- WAITFOR_RESPONSE_TIMEOUT) {
+ if ((current_time - a_con->theStartTransTime) > timeout)
+ {
#ifdef VM_TRACE
a_con->printState();
Uint32 t1 = a_con->theTransactionId;
Uint32 t2 = a_con->theTransactionId >> 32;
- ndbout_c("[%.8x %.8x]", t1, t2);
- abort();
+ ndbout_c("4012 [%.8x %.8x]", t1, t2);
+ //abort();
#endif
+ a_con->theReleaseOnClose = true;
a_con->setOperationErrorCodeAbort(4012);
- a_con->theCommitStatus = NdbConnection::Aborted;
+ a_con->theCommitStatus = NdbConnection::NeedAbort;
a_con->theCompletionStatus = NdbConnection::CompletedFailure;
a_con->handleExecuteCompletion();
remove_sent_list(i);
diff --git a/ndb/src/ndbapi/TransporterFacade.cpp b/ndb/src/ndbapi/TransporterFacade.cpp
index b6fb2d6cded..30d0eec1e4a 100644
--- a/ndb/src/ndbapi/TransporterFacade.cpp
+++ b/ndb/src/ndbapi/TransporterFacade.cpp
@@ -567,6 +567,19 @@ TransporterFacade::init(Uint32 nodeId, const ndb_mgm_configuration* props)
}
#endif
+ Uint32 timeout = 120000;
+ iter.first();
+ for (iter.first(); iter.valid(); iter.next())
+ {
+ Uint32 tmp1 = 0, tmp2 = 0;
+ iter.get(CFG_DB_TRANSACTION_CHECK_INTERVAL, &tmp1);
+ iter.get(CFG_DB_TRANSACTION_DEADLOCK_TIMEOUT, &tmp2);
+ tmp1 += tmp2;
+ if (tmp1 > timeout)
+ timeout = tmp1;
+ }
+ m_waitfor_timeout = timeout;
+
if (!theTransporterRegistry->start_service(m_socket_server)){
ndbout_c("Unable to start theTransporterRegistry->start_service");
DBUG_RETURN(false);
diff --git a/ndb/src/ndbapi/TransporterFacade.hpp b/ndb/src/ndbapi/TransporterFacade.hpp
index 99edea846c1..1e7377a3b4d 100644
--- a/ndb/src/ndbapi/TransporterFacade.hpp
+++ b/ndb/src/ndbapi/TransporterFacade.hpp
@@ -172,6 +172,7 @@ private:
*/
public:
STATIC_CONST( MAX_NO_THREADS = 4711 );
+ Uint32 m_waitfor_timeout; // in milli seconds...
private:
struct ThreadData {
diff --git a/ndb/test/ndbapi/testNodeRestart.cpp b/ndb/test/ndbapi/testNodeRestart.cpp
index a741e6233d9..eebd631af94 100644
--- a/ndb/test/ndbapi/testNodeRestart.cpp
+++ b/ndb/test/ndbapi/testNodeRestart.cpp
@@ -535,6 +535,52 @@ err:
return NDBT_FAILED;
}
+int
+runBug16772(NDBT_Context* ctx, NDBT_Step* step){
+
+ NdbRestarter restarter;
+ if (restarter.getNumDbNodes() < 2)
+ {
+ ctx->stopTest();
+ return NDBT_OK;
+ }
+
+ int aliveNodeId = restarter.getRandomNotMasterNodeId(rand());
+ int deadNodeId = aliveNodeId;
+ while (deadNodeId == aliveNodeId)
+ deadNodeId = restarter.getDbNodeId(rand() % restarter.getNumDbNodes());
+
+ if (restarter.insertErrorInNode(aliveNodeId, 930))
+ return NDBT_FAILED;
+
+ if (restarter.restartOneDbNode(deadNodeId,
+ /** initial */ false,
+ /** nostart */ true,
+ /** abort */ true))
+ return NDBT_FAILED;
+
+ if (restarter.waitNodesNoStart(&deadNodeId, 1))
+ return NDBT_FAILED;
+
+ if (restarter.startNodes(&deadNodeId, 1))
+ return NDBT_FAILED;
+
+ // It should now be hanging since we throw away NDB_FAILCONF
+ int ret = restarter.waitNodesStartPhase(&deadNodeId, 1, 3, 10);
+ // So this should fail...i.e it should not reach startphase 3
+
+ // Now send a NDB_FAILCONF for deadNo
+ int dump[] = { 7020, 323, 252, 0 };
+ dump[3] = deadNodeId;
+ if (restarter.dumpStateOneNode(aliveNodeId, dump, 4))
+ return NDBT_FAILED;
+
+ if (restarter.waitNodesStarted(&deadNodeId, 1))
+ return NDBT_FAILED;
+
+ return ret ? NDBT_OK : NDBT_FAILED;
+}
+
NDBT_TESTSUITE(testNodeRestart);
TESTCASE("NoLoad",
@@ -820,6 +866,10 @@ TESTCASE("Bug15685",
STEP(runBug15685);
FINALIZER(runClearTable);
}
+TESTCASE("Bug16772",
+ "Test bug with restarting before NF handling is complete"){
+ STEP(runBug16772);
+}
NDBT_TESTSUITE_END(testNodeRestart);
int main(int argc, const char** argv){
diff --git a/ndb/test/ndbapi/testTimeout.cpp b/ndb/test/ndbapi/testTimeout.cpp
index 71c11b25859..25392698642 100644
--- a/ndb/test/ndbapi/testTimeout.cpp
+++ b/ndb/test/ndbapi/testTimeout.cpp
@@ -24,6 +24,7 @@
#define TIMEOUT (Uint32)3000
Uint32 g_org_timeout = 3000;
+Uint32 g_org_deadlock = 3000;
int
setTransactionTimeout(NDBT_Context* ctx, NDBT_Step* step){
@@ -59,6 +60,60 @@ resetTransactionTimeout(NDBT_Context* ctx, NDBT_Step* step){
return NDBT_OK;
}
+int
+setDeadlockTimeout(NDBT_Context* ctx, NDBT_Step* step){
+ NdbRestarter restarter;
+ int timeout = ctx->getProperty("TransactionDeadlockTimeout", TIMEOUT);
+
+ NdbConfig conf(GETNDB(step)->getNodeId()+1);
+ unsigned int nodeId = conf.getMasterNodeId();
+ if (!conf.getProperty(nodeId,
+ NODE_TYPE_DB,
+ CFG_DB_TRANSACTION_DEADLOCK_TIMEOUT,
+ &g_org_deadlock))
+ return NDBT_FAILED;
+
+ g_err << "Setting timeout: " << timeout << endl;
+ int val[] = { DumpStateOrd::TcSetTransactionTimeout, timeout };
+ if(restarter.dumpStateAllNodes(val, 2) != 0){
+ return NDBT_FAILED;
+ }
+
+ return NDBT_OK;
+}
+
+int
+getDeadlockTimeout(NDBT_Context* ctx, NDBT_Step* step){
+ NdbRestarter restarter;
+
+ Uint32 val = 0;
+ NdbConfig conf(GETNDB(step)->getNodeId()+1);
+ unsigned int nodeId = conf.getMasterNodeId();
+ if (!conf.getProperty(nodeId,
+ NODE_TYPE_DB,
+ CFG_DB_TRANSACTION_DEADLOCK_TIMEOUT,
+ &val))
+ return NDBT_FAILED;
+
+ if (val < 120000)
+ val = 120000;
+ ctx->setProperty("TransactionDeadlockTimeout", 4*val);
+
+ return NDBT_OK;
+}
+
+int
+resetDeadlockTimeout(NDBT_Context* ctx, NDBT_Step* step){
+ NdbRestarter restarter;
+
+ int val[] = { DumpStateOrd::TcSetTransactionTimeout, g_org_deadlock };
+ if(restarter.dumpStateAllNodes(val, 2) != 0){
+ return NDBT_FAILED;
+ }
+
+ return NDBT_OK;
+}
+
int runLoadTable(NDBT_Context* ctx, NDBT_Step* step){
@@ -374,6 +429,43 @@ int runBuddyTransNoTimeout(NDBT_Context* ctx, NDBT_Step* step){
return result;
}
+int
+runError4012(NDBT_Context* ctx, NDBT_Step* step){
+ int result = NDBT_OK;
+ int loops = ctx->getNumLoops();
+ int stepNo = step->getStepNo();
+
+ int timeout = ctx->getProperty("TransactionDeadlockTimeout", TIMEOUT);
+
+ HugoOperations hugoOps(*ctx->getTab());
+ Ndb* pNdb = GETNDB(step);
+
+ do{
+ // Commit transaction
+ CHECK(hugoOps.startTransaction(pNdb) == 0);
+ CHECK(hugoOps.pkUpdateRecord(pNdb, 0) == 0);
+ int ret = hugoOps.execute_NoCommit(pNdb);
+ if (ret == 0)
+ {
+ int sleep = timeout;
+ ndbout << "Sleeping for " << sleep << " milliseconds" << endl;
+ NdbSleep_MilliSleep(sleep);
+
+ // Expect that transaction has NOT timed-out
+ CHECK(hugoOps.execute_Commit(pNdb) == 0);
+ }
+ else
+ {
+ CHECK(ret == 4012);
+ }
+ } while(false);
+
+ hugoOps.closeTransaction(pNdb);
+
+ return result;
+}
+
+
NDBT_TESTSUITE(testTimeout);
TESTCASE("DontTimeoutTransaction",
"Test that the transaction does not timeout "\
@@ -465,6 +557,15 @@ TESTCASE("BuddyTransNoTimeout5",
FINALIZER(resetTransactionTimeout);
FINALIZER(runClearTable);
}
+TESTCASE("Error4012", ""){
+ TC_PROPERTY("TransactionDeadlockTimeout", 120000);
+ INITIALIZER(runLoadTable);
+ INITIALIZER(getDeadlockTimeout);
+ INITIALIZER(setDeadlockTimeout);
+ STEPS(runError4012, 2);
+ FINALIZER(runClearTable);
+}
+
NDBT_TESTSUITE_END(testTimeout);
int main(int argc, const char** argv){
diff --git a/ndb/test/run-test/daily-basic-tests.txt b/ndb/test/run-test/daily-basic-tests.txt
index 6378b4a06d3..70518f7881d 100644
--- a/ndb/test/run-test/daily-basic-tests.txt
+++ b/ndb/test/run-test/daily-basic-tests.txt
@@ -236,6 +236,10 @@ max-time: 500
cmd: testTimeout
args: -n TimeoutRandTransaction T1
+max-time: 600
+cmd: testTimeout
+args: -n Error4012 T1
+
# SCAN TESTS
#
max-time: 500
@@ -446,6 +450,10 @@ max-time: 500
cmd: testNodeRestart
args: -n Bug15685 T1
+max-time: 500
+cmd: testNodeRestart
+args: -n Bug16772 T1
+
# OLD FLEX
max-time: 500
cmd: flexBench