ndb - bug#22696

  Fix handling of multiple redo phases
  - set correct state
  - handle 4 phases, by moving START_FRAGCONF to end of phase instead of beginning...


storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp:
  remove unused state ACTIVE_REMOVE_AFTER
  remove unused variable nextLcp
  remove unused methods (BUILD_INDX)
  add counter of outstanding GSN_EXEC_FRAGREQ
  remove c_redo_complete_fragments cause it fits badly with functional multi-redo-phases
storage/ndb/src/kernel/blocks/dblqh/DblqhInit.cpp:
  remove unused stuff
storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
  1) set correct state before starting next non-empty REDO phase
  2) move GSN_START_FRAGCONF to execFRAGCONF instead of when sending GSN_EXEC_FRAGREQ
     or fragments with 4 redo-phases would never be complete...
storage/ndb/test/ndbapi/testSystemRestart.cpp:
  testcase
storage/ndb/test/run-test/daily-basic-tests.txt:
  testcase

storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp

@@ -585,7 +585,6 @@ public:
 
     enum ExecSrStatus {
       IDLE = 0,
-      ACTIVE_REMOVE_AFTER = 1,
       ACTIVE = 2
     };
     /**
@@ -869,11 +868,6 @@ public:
      *       heard of.
      */
     Uint8 fragDistributionKey;
-    /**
-     *       The identity of the next local checkpoint this fragment
-     *       should perform.
-     */
-    Uint8 nextLcp;
    /**
      *       How many local checkpoints does the fragment contain
      */
@@ -2097,10 +2091,6 @@ private:
   void execEXEC_SRCONF(Signal* signal);
   void execREAD_PSEUDO_REQ(Signal* signal);
 
-  void build_acc(Signal*, Uint32 fragPtrI);
-  void execBUILDINDXREF(Signal*signal);
-  void execBUILDINDXCONF(Signal*signal);
-  
   void execDUMP_STATE_ORD(Signal* signal);
   void execACC_ABORTCONF(Signal* signal);
   void execNODE_FAILREP(Signal* signal);
@@ -2780,7 +2770,13 @@ private:
 /*THIS VARIABLE KEEPS TRACK OF HOW MANY FRAGMENTS THAT PARTICIPATE IN        */
 /*EXECUTING THE LOG. IF ZERO WE DON'T NEED TO EXECUTE THE LOG AT ALL.        */
 /* ------------------------------------------------------------------------- */
-  UintR cnoFragmentsExecSr;
+  Uint32 cnoFragmentsExecSr;
+
+  /**
+   * This is no of sent GSN_EXEC_FRAGREQ during this log phase
+   */
+  Uint32 cnoOutstandingExecFragReq;
+
 /* ------------------------------------------------------------------------- */
 /*THIS VARIABLE KEEPS TRACK OF WHICH OF THE FIRST TWO RESTART PHASES THAT    */
 /*HAVE COMPLETED.                                                            */
@@ -2801,7 +2797,6 @@ private:
   DLFifoList<Fragrecord> c_lcp_waiting_fragments;  // StartFragReq'ed
   DLFifoList<Fragrecord> c_lcp_restoring_fragments; // Restoring as we speek
   DLFifoList<Fragrecord> c_lcp_complete_fragments;  // Restored
-  DLFifoList<Fragrecord> c_redo_complete_fragments; // Redo'ed
   
 /* ------------------------------------------------------------------------- */
 /*USED DURING SYSTEM RESTART, INDICATES THE OLDEST GCI THAT CAN BE RESTARTED */

storage/ndb/src/kernel/blocks/dblqh/DblqhInit.cpp

@@ -168,7 +168,6 @@ Dblqh::Dblqh(Block_context& ctx):
   c_lcp_waiting_fragments(c_fragment_pool),
   c_lcp_restoring_fragments(c_fragment_pool),
   c_lcp_complete_fragments(c_fragment_pool),
-  c_redo_complete_fragments(c_fragment_pool),
   m_commitAckMarkerHash(m_commitAckMarkerPool),
   c_scanTakeOverHash(c_scanRecordPool)
 {
@@ -295,9 +294,6 @@ Dblqh::Dblqh(Block_context& ctx):
 
   addRecSignal(GSN_READ_PSEUDO_REQ, &Dblqh::execREAD_PSEUDO_REQ);
 
-  addRecSignal(GSN_BUILDINDXREF, &Dblqh::execBUILDINDXREF);
-  addRecSignal(GSN_BUILDINDXCONF, &Dblqh::execBUILDINDXCONF);
-
   addRecSignal(GSN_DEFINE_BACKUP_REF, &Dblqh::execDEFINE_BACKUP_REF);
   addRecSignal(GSN_DEFINE_BACKUP_CONF, &Dblqh::execDEFINE_BACKUP_CONF);
 

storage/ndb/test/ndbapi/testSystemRestart.cpp

@@ -1501,6 +1501,38 @@ int runSR_DD_2(NDBT_Context* ctx, NDBT_Step* step)
   return result;
 }
 
+int runBug22696(NDBT_Context* ctx, NDBT_Step* step)
+{
+  Ndb* pNdb = GETNDB(step);
+  int result = NDBT_OK;
+  Uint32 loops = ctx->getNumLoops();
+  Uint32 rows = ctx->getNumRecords();
+  NdbRestarter restarter;
+  HugoTransactions hugoTrans(*ctx->getTab());
+
+  Uint32 i = 0;
+  while(i<=loops && result != NDBT_FAILED)
+  {
+    for (Uint32 j = 0; j<10 && result != NDBT_FAILED; j++)
+      CHECK(hugoTrans.scanUpdateRecords(pNdb, rows) == 0);
+    
+    CHECK(restarter.restartAll(false, true, i > 0 ? true : false) == 0);
+    CHECK(restarter.waitClusterNoStart() == 0);
+    CHECK(restarter.insertErrorInAllNodes(7072) == 0);
+    CHECK(restarter.startAll() == 0);
+    CHECK(restarter.waitClusterStarted() == 0);
+
+    i++;
+    if (i < loops)
+    {
+      NdbSleep_SecSleep(5); // Wait for a few gcp
+    }
+  }
+  
+  ctx->stopTest();  
+  return result;
+}
+
 int 
 runBug27434(NDBT_Context* ctx, NDBT_Step* step)
 {
@@ -1813,8 +1845,13 @@ TESTCASE("Bug28770",
   STEP(runBug28770);
   FINALIZER(runClearTable);
 }
-
-
+TESTCASE("Bug22696", "")
+{
+  INITIALIZER(runWaitStarted);
+  INITIALIZER(runLoadTable);
+  INITIALIZER(runBug22696);
+  FINALIZER(runClearTable);
+}
 NDBT_TESTSUITE_END(testSystemRestart);
 
 int main(int argc, const char** argv){

storage/ndb/test/run-test/daily-basic-tests.txt

@@ -1042,4 +1042,7 @@ max-time: 300
 cmd: test_event
 args: -n Bug31701 T1
 
+max-time: 300
+cmd: testSystemRestart
+args: -n Bug22696 T1