Below is the list of changes that have just been committed into a local
5.1 repository of jonas. When jonas does a push these changes will
be propagated to the main repository and, within 24 hours after the
push, to the public repository.
For information on how to access the public repository
see http://dev.mysql.com/doc/mysql/en/installing-source-tree.html
ChangeSet@stripped, 2007-01-11 21:09:27+01:00, jonas@stripped +3 -0
Merge perch.ndb.mysql.com:/home/jonas/src/51-work
into perch.ndb.mysql.com:/home/jonas/src/mysql-5.1-new-ndb
MERGE: 1.2344.3.19
storage/ndb/src/kernel/blocks/dbdict/Dbdict.cpp@stripped, 2007-01-11 21:09:24+01:00, jonas@stripped +0 -0
Auto merged
MERGE: 1.110.1.2
storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp@stripped, 2007-01-11 21:09:24+01:00, jonas@stripped +0 -0
Auto merged
MERGE: 1.94.1.2
storage/ndb/test/run-test/daily-basic-tests.txt@stripped, 2007-01-11 21:09:24+01:00, jonas@stripped +0 -0
Auto merged
MERGE: 1.58.1.4
# This is a BitKeeper patch. What follows are the unified diffs for the
# set of deltas contained in the patch. The rest of the patch, the part
# that BitKeeper cares about, is below these diffs.
# User: jonas
# Host: perch.ndb.mysql.com
# Root: /home/jonas/src/mysql-5.1-new-ndb/RESYNC
--- 1.60/storage/ndb/test/run-test/daily-basic-tests.txt 2007-01-11 21:09:32 +01:00
+++ 1.61/storage/ndb/test/run-test/daily-basic-tests.txt 2007-01-11 21:09:32 +01:00
@@ -513,6 +513,14 @@
cmd: testNodeRestart
args: -n Bug24717 T1
+max-time: 1000
+cmd: testNodeRestart
+args: -n Bug25364 T1
+
+max-time: 1000
+cmd: testNodeRestart
+args: -n Bug25554 T1
+
#
# DICT TESTS
max-time: 1500
@@ -763,6 +771,10 @@
max-time: 1500
cmd: testSystemRestart
args: -n Bug24664
+
+max-time: 1000
+cmd: testNodeRestart
+args: -n Bug25468 T1
# OLD FLEX
max-time: 500
--- 1.111/storage/ndb/src/kernel/blocks/dbdict/Dbdict.cpp 2007-01-11 21:09:32 +01:00
+++ 1.112/storage/ndb/src/kernel/blocks/dbdict/Dbdict.cpp 2007-01-11 21:09:32 +01:00
@@ -10050,9 +10050,20 @@
}
OpSubEventPtr subbPtr;
Uint32 errCode = 0;
+
+ DictLockPtr loopPtr;
+ if (c_dictLockQueue.first(loopPtr) &&
+ loopPtr.p->lt->lockType == DictLockReq::NodeRestartLock)
+ {
+ jam();
+ errCode = 1405;
+ goto busy;
+ }
+
if (!c_opSubEvent.seize(subbPtr)) {
errCode = SubStartRef::Busy;
busy:
+ jam();
SubStartRef * ref = (SubStartRef *)signal->getDataPtrSend();
{ // fix
@@ -10151,6 +10162,7 @@
SubStartRef* ref = (SubStartRef*) signal->getDataPtrSend();
ref->senderRef = reference();
ref->senderData = subbPtr.p->m_senderData;
+ ref->errorCode = err;
sendSignal(subbPtr.p->m_senderRef, GSN_SUB_START_REF,
signal, SubStartRef::SignalLength2, JBB);
c_opSubEvent.release(subbPtr);
@@ -10213,6 +10225,7 @@
#ifdef EVENT_PH3_DEBUG
ndbout_c("DBDICT(Coordinator) got GSN_SUB_START_CONF = (%d)", subbPtr.i);
#endif
+ subbPtr.p->m_sub_start_conf = *conf;
subbPtr.p->m_reqTracker.reportConf(c_counterMgr, refToNode(senderRef));
completeSubStartReq(signal,subbPtr.i,0);
}
@@ -10252,6 +10265,9 @@
#ifdef EVENT_DEBUG
ndbout_c("SUB_START_CONF");
#endif
+
+ SubStartConf* conf = (SubStartConf*)signal->getDataPtrSend();
+ * conf = subbPtr.p->m_sub_start_conf;
sendSignal(subbPtr.p->m_senderRef, GSN_SUB_START_CONF,
signal, SubStartConf::SignalLength, JBB);
c_opSubEvent.release(subbPtr);
@@ -10373,6 +10389,7 @@
SubStopRef* ref = (SubStopRef*) signal->getDataPtrSend();
ref->senderRef = reference();
ref->senderData = subbPtr.p->m_senderData;
+ ref->errorCode = err;
sendSignal(subbPtr.p->m_senderRef, GSN_SUB_STOP_REF,
signal, SubStopRef::SignalLength, JBB);
c_opSubEvent.release(subbPtr);
@@ -10425,6 +10442,7 @@
* Coordinator
*/
ndbrequire(refToBlock(senderRef) == DBDICT);
+ subbPtr.p->m_sub_stop_conf = *conf;
subbPtr.p->m_reqTracker.reportConf(c_counterMgr, refToNode(senderRef));
completeSubStopReq(signal,subbPtr.i,0);
}
@@ -10465,6 +10483,8 @@
#ifdef EVENT_DEBUG
ndbout_c("SUB_STOP_CONF");
#endif
+ SubStopConf* conf = (SubStopConf*)signal->getDataPtrSend();
+ * conf = subbPtr.p->m_sub_stop_conf;
sendSignal(subbPtr.p->m_senderRef, GSN_SUB_STOP_CONF,
signal, SubStopConf::SignalLength, JBB);
c_opSubEvent.release(subbPtr);
@@ -10713,6 +10733,7 @@
SubRemoveRef* ref = (SubRemoveRef*) signal->getDataPtrSend();
ref->senderRef = reference();
ref->senderData = subbPtr.p->m_senderData;
+ ref->errorCode = err;
sendSignal(subbPtr.p->m_senderRef, GSN_SUB_REMOVE_REF,
signal, SubRemoveRef::SignalLength, JBB);
}
--- 1.95/storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp 2007-01-11 21:09:32 +01:00
+++ 1.96/storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp 2007-01-11 21:09:32 +01:00
@@ -3741,7 +3741,6 @@
takeOverPtr.i = takeOverPtrI;
ptrCheckGuard(takeOverPtr, MAX_NDB_NODES, takeOverRecord);
- releaseTakeOver(takeOverPtrI);
if ((takeOverPtr.p->toMasterStatus != TakeOverRecord::IDLE) &&
(takeOverPtr.p->toMasterStatus != TakeOverRecord::TO_WAIT_START_TAKE_OVER)) {
jam();
@@ -3755,6 +3754,7 @@
}//if
setAllowNodeStart(takeOverPtr.p->toStartingNode, true);
initTakeOver(takeOverPtr);
+ releaseTakeOver(takeOverPtrI);
}//Dbdih::endTakeOver()
void Dbdih::releaseTakeOver(Uint32 takeOverPtrI)
@@ -4046,6 +4046,11 @@
Uint32 newMasterId = nodeFail->masterNodeId;
const Uint32 noOfFailedNodes = nodeFail->noOfNodes;
+ if (ERROR_INSERTED(7179))
+ {
+ CLEAR_ERROR_INSERT_VALUE;
+ }
+
/*-------------------------------------------------------------------------*/
// The first step is to convert from a bit mask to an array of failed nodes.
/*-------------------------------------------------------------------------*/
@@ -4909,6 +4914,7 @@
break;
}
ndbrequire(ok);
+ endTakeOver(takeOverPtr.i);
}//if
}//Dbdih::handleTakeOverNewMaster()
@@ -10256,12 +10262,42 @@
Uint32 fragId = lcpReport->fragId;
jamEntry();
+
+ if (ERROR_INSERTED(7178) && nodeId != getOwnNodeId())
+ {
+ jam();
+ Uint32 owng =Sysfile::getNodeGroup(getOwnNodeId(), SYSFILE->nodeGroups);
+ Uint32 nodeg = Sysfile::getNodeGroup(nodeId, SYSFILE->nodeGroups);
+ if (owng == nodeg)
+ {
+ jam();
+ ndbout_c("throwing away LCP_FRAG_REP from (and killing) %d", nodeId);
+ SET_ERROR_INSERT_VALUE(7179);
+ signal->theData[0] = 9999;
+ sendSignal(numberToRef(CMVMI, nodeId),
+ GSN_NDB_TAMPER, signal, 1, JBA);
+ return;
+ }
+ }
+ if (ERROR_INSERTED(7179) && nodeId != getOwnNodeId())
+ {
+ jam();
+ Uint32 owng =Sysfile::getNodeGroup(getOwnNodeId(), SYSFILE->nodeGroups);
+ Uint32 nodeg = Sysfile::getNodeGroup(nodeId, SYSFILE->nodeGroups);
+ if (owng == nodeg)
+ {
+ jam();
+ ndbout_c("throwing away LCP_FRAG_REP from %d", nodeId);
+ return;
+ }
+ }
+
CRASH_INSERTION2(7025, isMaster());
CRASH_INSERTION2(7016, !isMaster());
-
+
bool fromTimeQueue = (signal->senderBlockRef() == reference());
-
+
TabRecordPtr tabPtr;
tabPtr.i = tableId;
ptrCheckGuard(tabPtr, ctabFileSize, tabRecord);
@@ -10463,6 +10499,37 @@
ndbrequire(false);
}//Dbdih::findReplica()
+
+int
+Dbdih::handle_invalid_lcp_no(const LcpFragRep* rep,
+ ReplicaRecordPtr replicaPtr)
+{
+ ndbrequire(!isMaster());
+ Uint32 lcpNo = rep->lcpNo;
+ Uint32 lcpId = rep->lcpId;
+ Uint32 replicaLcpNo = replicaPtr.p->nextLcp;
+ Uint32 prevReplicaLcpNo = prevLcpNo(replicaLcpNo);
+
+ warningEvent("Detected previous node failure of %d during lcp",
+ rep->nodeId);
+ replicaPtr.p->nextLcp = lcpNo;
+ replicaPtr.p->lcpId[lcpNo] = 0;
+ replicaPtr.p->lcpStatus[lcpNo] = ZINVALID;
+
+ for (Uint32 i = lcpNo; i != lcpNo; i = nextLcpNo(i))
+ {
+ jam();
+ if (replicaPtr.p->lcpStatus[i] == ZVALID &&
+ replicaPtr.p->lcpId[i] >= lcpId)
+ {
+ ndbout_c("i: %d lcpId: %d", i, replicaPtr.p->lcpId[i]);
+ ndbrequire(false);
+ }
+ }
+
+ return 0;
+}
+
/**
* Return true if table is all fragment replicas have been checkpointed
* to disk (in all LQHs)
@@ -10491,9 +10558,12 @@
ndbrequire(replicaPtr.p->lcpOngoingFlag == true);
if(lcpNo != replicaPtr.p->nextLcp){
- ndbout_c("lcpNo = %d replicaPtr.p->nextLcp = %d",
- lcpNo, replicaPtr.p->nextLcp);
- ndbrequire(false);
+ if (handle_invalid_lcp_no(lcpReport, replicaPtr))
+ {
+ ndbout_c("lcpNo = %d replicaPtr.p->nextLcp = %d",
+ lcpNo, replicaPtr.p->nextLcp);
+ ndbrequire(false);
+ }
}
ndbrequire(lcpNo == replicaPtr.p->nextLcp);
ndbrequire(lcpNo < MAX_LCP_STORED);
| Thread |
|---|
| • bk commit into 5.1 tree (jonas:1.2371) | jonas | 11 Jan |
