Commit be0d6c94 authored by unknown's avatar unknown
Browse files

ndb - wl1760/bug#18216 

  add two new start options that will decrease likelyhood of bug#18612
  push cntr-sp2 logic down into qmgr-sp1 to decrease likelyhood of bug#18612


ndb/include/kernel/signaldata/CmRegSignalData.hpp:
  Expand CmRegReq with lots of stuff
ndb/include/mgmapi/ndb_logevent.h:
  Add Start report during sp1
ndb/src/common/debugger/EventLogger.cpp:
  Add Start report during sp1
ndb/src/kernel/blocks/dbdih/DbdihInit.cpp:
  Init cntrref
ndb/src/kernel/blocks/dbdih/DbdihMain.cpp:
  Fix small bugs related to partial initial start
ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp:
  Remove timeouts as they are handled in QMGR nowadays
ndb/src/kernel/blocks/qmgr/Qmgr.hpp:
  Push cntr sp2 logic down to QMGR to prevent the forming of multiple QMGR cluster
    that will lead to start problems (crashes...)
ndb/src/kernel/blocks/qmgr/QmgrInit.cpp:
  Push cntr sp2 logic down to QMGR to prevent the forming of multiple QMGR cluster
    that will lead to start problems (crashes...)
ndb/src/kernel/blocks/qmgr/QmgrMain.cpp:
  Push cntr sp2 logic down to QMGR to prevent the forming of multiple QMGR cluster
    that will lead to start problems (crashes...)
ndb/src/kernel/vm/Configuration.cpp:
  Add new flags
    --initial-start
    --nowait-nodes
parent e40bdb5b

ndb/include/kernel/signaldata/CmRegSignalData.hpp

+23 −6
Original line number Diff line number Diff line
@@ -30,12 +30,17 @@ class CmRegReq { 
  friend class Qmgr;

  

public:

  STATIC_CONST( SignalLength = 3 );

  STATIC_CONST( SignalLength = 5 + NdbNodeBitmask::Size );

private:

  

  Uint32 blockRef;

  Uint32 nodeId;

  Uint32 version;    // See ndb_version.h



  Uint32 start_type; // As specified by cmd-line or mgm, NodeState::StartType

  Uint32 latest_gci; // 0 means no fs

  Uint32 skip_nodes[NdbNodeBitmask::Size]; // Nodes that does not _need_ 

                                           // to be part of restart

};



/**
@@ -59,7 +64,6 @@ private: 
   * The dynamic id that the node reciving this signal has

   */

  Uint32 dynamicId;

  

  Uint32 allNdbNodes[NdbNodeBitmask::Size];  

};
@@ -73,7 +77,7 @@ class CmRegRef { 
  friend class Qmgr;

  

public:

  STATIC_CONST( SignalLength = 4 );

  STATIC_CONST( SignalLength = 7 + NdbNodeBitmask::Size );

  

  enum ErrorCode {

    ZBUSY = 0,          /* Only the president can send this */
@@ -85,14 +89,27 @@ public: 
                         * as president. */

    ZNOT_PRESIDENT = 5, /* We are not president */

    ZNOT_DEAD = 6,       /* We are not dead when we are starting  */

    ZINCOMPATIBLE_VERSION = 7

    ZINCOMPATIBLE_VERSION = 7,

    ZINCOMPATIBLE_START_TYPE = 8

  };

private:

  

  Uint32 blockRef;

  Uint32 nodeId;

  Uint32 errorCode;

  /**

   * Applicable if ZELECTION

   */

  Uint32 presidentCandidate;

  Uint32 candidate_latest_gci; // 0 means non



  /**

   * Data for sending node sending node

   */

  Uint32 latest_gci; 

  Uint32 start_type; 

  Uint32 skip_nodes[NdbNodeBitmask::Size]; // Nodes that does not _need_ 

                                           // to be part of restart

};



class CmAdd {

ndb/include/mgmapi/ndb_logevent.h

+12 −1
Original line number Diff line number Diff line
@@ -166,10 +166,14 @@ extern "C" { 
    /** NDB_MGM_EVENT_CATEGORY_BACKUP */

    NDB_LE_BackupCompleted = 56,

    /** NDB_MGM_EVENT_CATEGORY_BACKUP */

    NDB_LE_BackupAborted = 57

    NDB_LE_BackupAborted = 57,



    /* 58 used in 5.1  */

    /* 59 used */



    /** NDB_MGM_EVENT_CATEGORY_STARTUP */

    NDB_LE_StartReport = 60



    /* 60 unused */

    /* 61 unused */

    /* 62 unused */
@@ -625,6 +629,13 @@ extern "C" { 
        unsigned type;

        unsigned node_id;

      } SingleUser;

      /** Log even data @ref NDB_LE_StartReport */

      struct {

	unsigned report_type;

	unsigned remaining_time;

	unsigned bitmask_size;

	unsigned bitmask_data[1];

      } StartReport;

#ifndef DOXYGEN_FIX

    };

#else

ndb/src/common/debugger/EventLogger.cpp

+85 −0
Original line number Diff line number Diff line
@@ -707,6 +707,90 @@ void getTextSingleUser(QQQQ) { 
  }

}



void getTextStartReport(QQQQ) {

  Uint32 time = theData[2];

  Uint32 sz = theData[3];

  char mask1[100];

  char mask2[100];

  char mask3[100];

  char mask4[100];

  BitmaskImpl::getText(sz, theData + 4 + (0 * sz), mask1);

  BitmaskImpl::getText(sz, theData + 4 + (1 * sz), mask2);

  BitmaskImpl::getText(sz, theData + 4 + (2 * sz), mask3);

  BitmaskImpl::getText(sz, theData + 4 + (3 * sz), mask4);

  switch(theData[1]){

  case 1: // Wait initial

    BaseString::snprintf

      (m_text, m_text_len,

       "Initial start, waiting for %s to connect, "

       " nodes [ all: %s connected: %s no-wait: %s ]",

       mask4, mask1, mask2, mask3);

    break;

  case 2: // Wait partial

    BaseString::snprintf

      (m_text, m_text_len,

       "Waiting until nodes: %s connects, "

       "nodes [ all: %s connected: %s no-wait: %s ]",

       mask4, mask1, mask2, mask3);

    break;

  case 3: // Wait partial timeout

    BaseString::snprintf

      (m_text, m_text_len,

       "Waiting %u sec for nodes %s to connect, "

       "nodes [ all: %s connected: %s no-wait: %s ]",

       

       time, mask4, mask1, mask2, mask3);

    break;

  case 4: // Wait partioned

    BaseString::snprintf

      (m_text, m_text_len,

       "Waiting for non partitioned start, "

       "nodes [ all: %s connected: %s missing: %s no-wait: %s ]",

       

       mask1, mask2, mask4, mask3);

    break;

  case 5:

    BaseString::snprintf

      (m_text, m_text_len,

       "Waiting %u sec for non partitioned start, "

       "nodes [ all: %s connected: %s missing: %s no-wait: %s ]",

       

       time, mask1, mask2, mask4, mask3);

    break;

  case 0x8000: // Do initial

    BaseString::snprintf

      (m_text, m_text_len,

       "Initial start with nodes %s [ missing: %s no-wait: %s ]",

       mask2, mask4, mask3);

    break;

  case 0x8001: // Do start

    BaseString::snprintf

      (m_text, m_text_len,

       "Start with all nodes %s",

       mask2);

    break;

  case 0x8002: // Do partial

    BaseString::snprintf

      (m_text, m_text_len,

       "Start with nodes %s [ missing: %s no-wait: %s ]",

       mask2, mask4, mask3);

    break;

  case 0x8003: // Do partioned

    BaseString::snprintf

      (m_text, m_text_len,

       "Start potentially partitioned with nodes %s "

       " [ missing: %s no-wait: %s ]",

       mask2, mask4, mask3);

    break;

  default:

    BaseString::snprintf

      (m_text, m_text_len,

       "Unknown startreport: 0x%x [ %s %s %s %s ]", 

       theData[1],

       mask1, mask2, mask3, mask4);

  }

}



#if 0

BaseString::snprintf(m_text, 

		     m_text_len,
@@ -755,6 +839,7 @@ const EventLoggerBase::EventRepLogLevelMatrix EventLoggerBase::matrix[] = { 
  ROW(StartREDOLog,            LogLevel::llStartUp,    10, Logger::LL_INFO ),

  ROW(StartLog,                LogLevel::llStartUp,    10, Logger::LL_INFO ),

  ROW(UNDORecordsExecuted,     LogLevel::llStartUp,    15, Logger::LL_INFO ),

  ROW(StartReport,             LogLevel::llStartUp,     4, Logger::LL_INFO ),

  

  // NODERESTART

  ROW(NR_CopyDict,             LogLevel::llNodeRestart, 8, Logger::LL_INFO ),

ndb/src/kernel/blocks/dbdih/DbdihInit.cpp

+1 −0
Original line number Diff line number Diff line
@@ -71,6 +71,7 @@ void Dbdih::initData() 
  cwaitLcpSr       = false;

  c_blockCommit    = false;

  c_blockCommitNo  = 1;

  cntrlblockref    = RNIL;

}//Dbdih::initData()



void Dbdih::initRecords()

ndb/src/kernel/blocks/dbdih/DbdihMain.cpp

+52 −26
Original line number Diff line number Diff line
@@ -11659,7 +11659,7 @@ void Dbdih::makeNodeGroups(Uint32 nodeArray[]) 
  Uint32 tmngNode;

  Uint32 tmngNodeGroup;

  Uint32 tmngLimit;

  Uint32 i;

  Uint32 i, j;



  /**-----------------------------------------------------------------------

   * ASSIGN ALL ACTIVE NODES INTO NODE GROUPS. HOT SPARE NODES ARE ASSIGNED
@@ -11705,6 +11705,38 @@ void Dbdih::makeNodeGroups(Uint32 nodeArray[]) 
      Sysfile::setNodeGroup(mngNodeptr.i, SYSFILE->nodeGroups, mngNodeptr.p->nodeGroup);

    }//if

  }//for



  for (i = 0; i<cnoOfNodeGroups; i++)

  {

    jam();

    bool alive = false;

    NodeGroupRecordPtr NGPtr;

    NGPtr.i = i;

    ptrCheckGuard(NGPtr, MAX_NDB_NODES, nodeGroupRecord);

    for (j = 0; j<NGPtr.p->nodeCount; j++)

    {

      jam();

      mngNodeptr.i = NGPtr.p->nodesInGroup[j];

      ptrCheckGuard(mngNodeptr, MAX_NDB_NODES, nodeRecord);

      if (checkNodeAlive(NGPtr.p->nodesInGroup[j]))

      {

	alive = true;

	break;

      }

    }



    if (!alive)

    {

      char buf[255];

      BaseString::snprintf

	(buf, sizeof(buf), 

	 "Illegal initial start, no alive node in nodegroup %u", i);

      progError(__LINE__, 

		NDBD_EXIT_SR_RESTARTCONFLICT,

		buf);

      

    }

  }

}//Dbdih::makeNodeGroups()



/**
@@ -12512,7 +12544,6 @@ void Dbdih::sendStartFragreq(Signal* signal, 
void Dbdih::setInitialActiveStatus()

{

  NodeRecordPtr siaNodeptr;

  Uint32 tsiaNodeActiveStatus;

  Uint32 tsiaNoActiveNodes;



  tsiaNoActiveNodes = csystemnodes - cnoHotSpare;
@@ -12520,39 +12551,34 @@ void Dbdih::setInitialActiveStatus() 
    SYSFILE->nodeStatus[i] = 0;

  for (siaNodeptr.i = 1; siaNodeptr.i < MAX_NDB_NODES; siaNodeptr.i++) {

    ptrAss(siaNodeptr, nodeRecord);

    if (siaNodeptr.p->nodeStatus == NodeRecord::ALIVE) {

    switch(siaNodeptr.p->nodeStatus){

    case NodeRecord::ALIVE:

    case NodeRecord::DEAD:

      if (tsiaNoActiveNodes == 0) {

        jam();

        siaNodeptr.p->activeStatus = Sysfile::NS_HotSpare;

      } else {

        jam();

        tsiaNoActiveNodes = tsiaNoActiveNodes - 1;

        siaNodeptr.p->activeStatus = Sysfile::NS_Active;

      }//if

    } else {

      jam();

      siaNodeptr.p->activeStatus = Sysfile::NS_NotDefined;

    }//if

    switch (siaNodeptr.p->activeStatus) {

    case Sysfile::NS_Active:

      jam();

      tsiaNodeActiveStatus = Sysfile::NS_Active;

      break;

    case Sysfile::NS_HotSpare:

      jam();

      tsiaNodeActiveStatus = Sysfile::NS_HotSpare;

      break;

    case Sysfile::NS_NotDefined:

        if (siaNodeptr.p->nodeStatus == NodeRecord::ALIVE)

	{

	  jam();

      tsiaNodeActiveStatus = Sysfile::NS_NotDefined;

	  siaNodeptr.p->activeStatus = Sysfile::NS_Active;

	} 

	else

	{

	  siaNodeptr.p->activeStatus = Sysfile::NS_NotActive_NotTakenOver;

	}

      }

      break;

    default:

      ndbrequire(false);

      return;

      jam();

      siaNodeptr.p->activeStatus = Sysfile::NS_NotDefined;

      break;

    }//switch

    Sysfile::setNodeStatus(siaNodeptr.i, SYSFILE->nodeStatus,

                           tsiaNodeActiveStatus);

    }//if

    Sysfile::setNodeStatus(siaNodeptr.i, 

			   SYSFILE->nodeStatus,

                           siaNodeptr.p->activeStatus);

  }//for

}//Dbdih::setInitialActiveStatus()