List:Commits« Previous MessageNext Message »
From:jonas Date:February 19 2007 8:29pm
Subject:bk commit into 5.1 tree (jonas:1.2422)
View as plain text  
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, 2007-02-19 21:29:18+01:00, jonas@stripped +6 -0
  Merge perch.ndb.mysql.com:/home/jonas/src/51-telco-gca
  into  perch.ndb.mysql.com:/home/jonas/src/telco-6.1
  MERGE: 1.2403.4.7

  storage/ndb/src/kernel/blocks/dbdih/Dbdih.hpp@stripped, 2007-02-19 21:24:34+01:00, jonas@stripped +0 -0
    Auto merged
    MERGE: 1.24.1.2

  storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp@stripped, 2007-02-19 21:29:16+01:00, jonas@stripped +1 -2
    merge
    MERGE: 1.96.1.5

  storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp@stripped, 2007-02-19 21:24:34+01:00, jonas@stripped +0 -0
    Auto merged
    MERGE: 1.129.1.2

  storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp@stripped, 2007-02-19 21:24:34+01:00, jonas@stripped +0 -0
    Auto merged
    MERGE: 1.41.1.2

  storage/ndb/test/ndbapi/testNodeRestart.cpp@stripped, 2007-02-19 21:24:34+01:00, jonas@stripped +0 -0
    Auto merged
    MERGE: 1.36.1.3

  storage/ndb/test/run-test/daily-basic-tests.txt@stripped, 2007-02-19 21:24:35+01:00, jonas@stripped +0 -0
    Auto merged
    MERGE: 1.61.1.3

# 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/telco-6.1/RESYNC

--- 1.66/storage/ndb/test/run-test/daily-basic-tests.txt	2007-02-19 21:29:24 +01:00
+++ 1.67/storage/ndb/test/run-test/daily-basic-tests.txt	2007-02-19 21:29:24 +01:00
@@ -525,6 +525,10 @@
 cmd: testNodeRestart
 args: -n Bug25554 T1
 
+max-time: 1000
+cmd: testNodeRestart
+args: -n Bug26457 T1
+
 #
 # DICT TESTS
 max-time: 1500

--- 1.26/storage/ndb/src/kernel/blocks/dbdih/Dbdih.hpp	2007-02-19 21:29:24 +01:00
+++ 1.27/storage/ndb/src/kernel/blocks/dbdih/Dbdih.hpp	2007-02-19 21:29:24 +01:00
@@ -1381,6 +1381,7 @@
   Uint32 csystemnodes;
   Uint32 currentgcp;
   Uint32 c_newest_restorable_gci;
+  Uint32 c_set_initial_start_flag;
 
   enum GcpMasterTakeOverState {
     GMTOS_IDLE = 0,

--- 1.102/storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp	2007-02-19 21:29:24 +01:00
+++ 1.103/storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp	2007-02-19 21:29:24 +01:00
@@ -677,6 +677,12 @@
     Uint32 tmp= SYSFILE->m_restart_seq;
     memcpy(sysfileData, cdata, sizeof(sysfileData));
     SYSFILE->m_restart_seq = tmp;
+
+    if (c_set_initial_start_flag)
+    {
+      jam();
+      Sysfile::setInitialStartOngoing(SYSFILE->systemRestartBits);
+    }
   }
 
   c_copyGCISlave.m_copyReason = reason;
@@ -1290,6 +1296,11 @@
     // The permission is given by the master node in the alive set.  
     /*-----------------------------------------------------------------------*/
     createMutexes(signal, 0);
+    if (cstarttype == NodeState::ST_INITIAL_NODE_RESTART)
+    {
+      jam();
+      c_set_initial_start_flag = TRUE; // In sysfile...
+    }
     break;
     
   case ZNDB_SPH3:
@@ -4829,6 +4840,8 @@
 Dbdih::startLcpMasterTakeOver(Signal* signal, Uint32 nodeId){
   jam();
 
+  Uint32 oldNode = c_lcpMasterTakeOverState.failedNodeId;
+
   c_lcpMasterTakeOverState.minTableId = ~0;
   c_lcpMasterTakeOverState.minFragId = ~0;
   c_lcpMasterTakeOverState.failedNodeId = nodeId;
@@ -4847,7 +4860,20 @@
     /**
      * Node failure during master take over...
      */
-    g_eventLogger.info("Nodefail during master take over");
+    g_eventLogger.info("Nodefail during master take over (old: %)", oldNode);
+  }
+  
+  NodeRecordPtr nodePtr;
+  nodePtr.i = oldNode;
+  if (oldNode > 0 && oldNode < MAX_NDB_NODES)
+  {
+    jam();
+    ptrCheckGuard(nodePtr, MAX_NDB_NODES, nodeRecord);
+    if (nodePtr.p->m_nodefailSteps.get(NF_LCP_TAKE_OVER))
+    {
+      jam();
+      checkLocalNodefailComplete(signal, oldNode, NF_LCP_TAKE_OVER);
+    }
   }
   
   setLocalNodefailHandling(signal, nodeId, NF_LCP_TAKE_OVER);
@@ -5864,6 +5890,14 @@
   jamEntry();
   const BlockReference newMasterBlockref = req->masterRef;
 
+  if (newMasterBlockref != cmasterdihref)
+  {
+    jam();
+    ndbout_c("resending GSN_MASTER_LCPREQ");
+    sendSignalWithDelay(reference(), GSN_MASTER_LCPREQ, signal,
+			signal->getLength(), 50);
+    return;
+  }
   Uint32 failedNodeId = req->failedNodeId;
 
   /**
@@ -6160,6 +6194,8 @@
   ptrCheckGuard(nodePtr, MAX_NDB_NODES, nodeRecord);
   nodePtr.p->lcpStateAtTakeOver = lcpState;
 
+  CRASH_INSERTION(7180);
+  
 #ifdef VM_TRACE
   g_eventLogger.info("MASTER_LCPCONF");
   printMASTER_LCP_CONF(stdout, &signal->theData[0], 0, 0);
@@ -10718,6 +10754,17 @@
   
   sendSignal(c_lcpState.m_masterLcpDihRef, GSN_LCP_COMPLETE_REP, signal, 
 	     LcpCompleteRep::SignalLength, JBB);
+
+  /**
+   * Say that an initial node restart does not need to be redone
+   *   once node has been part of first LCP
+   */
+  if (c_set_initial_start_flag &&
+      c_lcpState.m_participatingLQH.get(getOwnNodeId()))
+  {
+    jam();
+    c_set_initial_start_flag = FALSE;
+  }
 }
 
 /*-------------------------------------------------------------------------- */

--- 1.134/storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp	2007-02-19 21:29:24 +01:00
+++ 1.135/storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp	2007-02-19 21:29:24 +01:00
@@ -11904,7 +11904,7 @@
     return;
   }
 
-  if (getNodeState().getNodeRestartInProgress())
+  if (getNodeState().getNodeRestartInProgress() && cstartRecReq == ZFALSE)
   {
     GCPSaveRef * const saveRef = (GCPSaveRef*)&signal->theData[0];
     saveRef->dihPtr = dihPtr;
@@ -11951,7 +11951,6 @@
   }//if
   
   ndbrequire(ccurrentGcprec == RNIL);
-    
   ccurrentGcprec = 0;
   gcpPtr.i = ccurrentGcprec;
   ptrCheckGuard(gcpPtr, cgcprecFileSize, gcpRecord);

--- 1.44/storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp	2007-02-19 21:29:24 +01:00
+++ 1.45/storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp	2007-02-19 21:29:24 +01:00
@@ -75,8 +75,8 @@
   { DBTUP_REF,   1 ,  4000,  4007 },
   { DBDICT_REF,  1 ,  6000,  6003 },
   { NDBCNTR_REF, 0 ,  1000,  1999 },
+  { CMVMI_REF,   1 ,  9000,  9999 }, // before QMGR
   { QMGR_REF,    1 ,     1,   999 },
-  { CMVMI_REF,   1 ,  9000,  9999 },
   { TRIX_REF,    1 ,     0,     0 },
   { BACKUP_REF,  1 , 10000, 10999 },
   { DBUTIL_REF,  1 , 11000, 11999 },

--- 1.40/storage/ndb/test/ndbapi/testNodeRestart.cpp	2007-02-19 21:29:24 +01:00
+++ 1.41/storage/ndb/test/ndbapi/testNodeRestart.cpp	2007-02-19 21:29:24 +01:00
@@ -1178,6 +1178,85 @@
   return NDBT_OK;
 }
 
+int
+runBug26457(NDBT_Context* ctx, NDBT_Step* step)
+{
+  NdbRestarter res;
+  if (res.getNumDbNodes() < 4)
+    return NDBT_OK;
+
+  int loops = ctx->getNumLoops();
+  while (loops --)
+  {
+retry:
+    int master = res.getMasterNodeId();
+    int next = res.getNextMasterNodeId(master);
+
+    ndbout_c("master: %d next: %d", master, next);
+
+    if (res.getNodeGroup(master) == res.getNodeGroup(next))
+    {
+      res.restartOneDbNode(next, false, false, true);
+      if (res.waitClusterStarted())
+	return NDBT_FAILED;
+      goto retry;
+    }
+
+    int val2[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 2 };
+    
+    if (res.dumpStateOneNode(next, val2, 2))
+      return NDBT_FAILED;
+    
+    if (res.insertErrorInNode(next, 7180))
+      return NDBT_FAILED;
+    
+    res.restartOneDbNode(master, false, false, true);
+    if (res.waitClusterStarted())
+      return NDBT_FAILED;
+  }
+  
+  return NDBT_OK;
+}
+
+int 
+runBug26481(NDBT_Context* ctx, NDBT_Step* step)
+{
+  
+  int result = NDBT_OK;
+  int loops = ctx->getNumLoops();
+  int records = ctx->getNumRecords();
+  NdbRestarter res;
+  
+  int node = res.getRandomNotMasterNodeId(rand());
+  ndbout_c("node: %d", node);
+  if (res.restartOneDbNode(node, true, true, true))
+    return NDBT_FAILED;
+
+  if (res.waitNodesNoStart(&node, 1))
+    return NDBT_FAILED;
+
+  int val2[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 1 };
+  if (res.dumpStateOneNode(node, val2, 2))
+    return NDBT_FAILED;
+
+  if (res.insertErrorInNode(node, 7018))
+    return NDBT_FAILED;
+
+  if (res.startNodes(&node, 1))
+    return NDBT_FAILED;
+
+  res.waitNodesStartPhase(&node, 1, 3);
+  
+  if (res.waitNodesNoStart(&node, 1))
+    return NDBT_FAILED;
+
+  res.startNodes(&node, 1);
+  
+  if (res.waitClusterStarted())
+    return NDBT_FAILED;
+  
+  return NDBT_OK;
+}
 
 NDBT_TESTSUITE(testNodeRestart);
 TESTCASE("NoLoad", 
@@ -1513,6 +1592,12 @@
 }
 TESTCASE("Bug25554", ""){
   INITIALIZER(runBug25554);
+}
+TESTCASE("Bug26457", ""){
+  INITIALIZER(runBug26457);
+}
+TESTCASE("Bug26481", ""){
+  INITIALIZER(runBug26481);
 }
 NDBT_TESTSUITE_END(testNodeRestart);
 
Thread
bk commit into 5.1 tree (jonas:1.2422)jonas19 Feb