bk commit into 5.1 tree (jonas:1.2115) BUG#27003

[jonas]

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, 2007-03-12 15:28:07+01:00, jonas@stripped +5 -0
  ndb - bug#27003
    Handle random(not in order) LQHKEYREQ failures during node-restart

  storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp@stripped, 2007-03-12 15:28:05+01:00,
jonas@stripped +16 -3
    Handle random(not in order) LQHKEYREQ failures during node-restart

  storage/ndb/src/kernel/blocks/dbtup/DbtupExecQuery.cpp@stripped, 2007-03-12 15:28:06+01:00,
jonas@stripped +24 -0
    Error inserts for random TUPKEY errors

  storage/ndb/src/kernel/blocks/dbtup/DbtupGen.cpp@stripped, 2007-03-12 15:28:06+01:00,
jonas@stripped +1 -1
    Move CLEAR_ERROR_INSERT_VALUE to constructor so that it's reasonable to use it for
restart testing

  storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp@stripped, 2007-03-12 15:28:06+01:00,
jonas@stripped +1 -0
    Add error insert for CopyFragRef

  storage/ndb/test/ndbapi/testNodeRestart.cpp@stripped, 2007-03-12 15:28:06+01:00,
jonas@stripped +56 -0
    Test prg for bug#27003

# 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/drop5

--- 1.92/storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp	2007-03-12 15:28:11 +01:00
+++ 1.93/storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp	2007-03-12 15:28:11 +01:00
@@ -9573,6 +9573,15 @@
     closeCopyLab(signal);
     return;
   }//if
+
+  if (scanptr.p->scanState == ScanRecord::WAIT_LQHKEY_COPY &&
+      scanptr.p->scanErrorCounter)
+  {
+    jam();
+    closeCopyLab(signal);
+    return;
+  }
+  
   if (scanptr.p->scanState == ScanRecord::WAIT_LQHKEY_COPY) {
     jam();
 /*---------------------------------------------------------------------------*/
@@ -9649,13 +9658,16 @@
 void Dblqh::copyLqhKeyRefLab(Signal* signal) 
 {
   ndbrequire(tcConnectptr.p->transid[1] == signal->theData[4]);
-  tcConnectptr.p->copyCountWords -= signal->theData[3];
+  Uint32 copyWords = signal->theData[3];
   scanptr.i = tcConnectptr.p->tcScanRec;
   c_scanRecordPool.getPtr(scanptr);
   scanptr.p->scanErrorCounter++;
   tcConnectptr.p->errorCode = terrorCode;
-  closeCopyLab(signal);
-  return;
+  
+  LqhKeyConf* conf = (LqhKeyConf*)signal->getDataPtrSend();
+  conf->transId1 = copyWords;
+  conf->transId2 = tcConnectptr.p->transid[1];
+  copyCompletedLab(signal);
 }//Dblqh::copyLqhKeyRefLab()
 
 void Dblqh::closeCopyLab(Signal* signal) 
@@ -9666,6 +9678,7 @@
 // Wait until all of those have arrived until we start the
 // close process.
 /*---------------------------------------------------------------------------*/
+    scanptr.p->scanState = ScanRecord::WAIT_LQHKEY_COPY;
     jam();
     return;
   }//if

--- 1.22/storage/ndb/src/kernel/blocks/dbtup/DbtupExecQuery.cpp	2007-03-12 15:28:11 +01:00
+++ 1.23/storage/ndb/src/kernel/blocks/dbtup/DbtupExecQuery.cpp	2007-03-12 15:28:11 +01:00
@@ -213,6 +213,30 @@
 //---------------------------------------------------
   PagePtr pagePtr;
   Uint32 pageOffset;
+
+  if (ERROR_INSERTED(4025))
+  {
+    signal->theData[0] = 827;
+    return;
+  }
+  if (ERROR_INSERTED(4026))
+  {
+    CLEAR_ERROR_INSERT_VALUE;
+    signal->theData[0] = 827;
+    return;
+  }
+  if (ERROR_INSERTED(4027) && (rand() % 100) > 25)
+  {
+    signal->theData[0] = 827;
+    return;
+  }
+  if (ERROR_INSERTED(4028) && (rand() % 100) > 25)
+  {
+    CLEAR_ERROR_INSERT_VALUE;
+    signal->theData[0] = 827;
+    return;
+  }
+  
   if (!allocTh(regFragPtr.p,
                regTabPtr.p,
                NORMAL_PAGE,

--- 1.25/storage/ndb/src/kernel/blocks/dbtup/DbtupGen.cpp	2007-03-12 15:28:11 +01:00
+++ 1.26/storage/ndb/src/kernel/blocks/dbtup/DbtupGen.cpp	2007-03-12 15:28:11 +01:00
@@ -66,6 +66,7 @@
   undoPage = 0;
   totNoOfPagesAllocated = 0;
   cnoOfAllocatedPages = 0;
+  CLEAR_ERROR_INSERT_VALUE;
   
   // Records with constant sizes
 }//Dbtup::initData()
@@ -576,7 +577,6 @@
   switch (startPhase) {
   case ZSTARTPHASE1:
     ljam();
-    CLEAR_ERROR_INSERT_VALUE;
     cownref = calcTupBlockRef(0);
     break;
   default:

--- 1.28/storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp	2007-03-12 15:28:11 +01:00
+++ 1.29/storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp	2007-03-12 15:28:11 +01:00
@@ -190,6 +190,7 @@
     break;
 
   case SystemError::CopyFragRefError:
+    CRASH_INSERTION(1000);
     BaseString::snprintf(buf, sizeof(buf), 
 	     "Node %d killed this node because "
 	     "it could not copy a fragment during node restart. "

--- 1.32/storage/ndb/test/ndbapi/testNodeRestart.cpp	2007-03-12 15:28:11 +01:00
+++ 1.33/storage/ndb/test/ndbapi/testNodeRestart.cpp	2007-03-12 15:28:11 +01:00
@@ -1304,6 +1304,59 @@
   return NDBT_OK;
 }
 
+int
+runBug27003(NDBT_Context* ctx, NDBT_Step* step)
+{
+  int result = NDBT_OK;
+  int loops = ctx->getNumLoops();
+  int records = ctx->getNumRecords();
+  NdbRestarter res;
+  
+  static const int errnos[] = { 4025, 4026, 4027, 4028, 0 };
+
+  int node = res.getRandomNotMasterNodeId(rand());
+  ndbout_c("node: %d", node);
+  if (res.restartOneDbNode(node, false, true, true))
+    return NDBT_FAILED;
+
+  Uint32 pos = 0;
+  for (Uint32 i = 0; i<loops; i++)
+  {
+    while (errnos[pos] != 0)
+    {
+      ndbout_c("Tesing err: %d", errnos[pos]);
+      
+      if (res.waitNodesNoStart(&node, 1))
+	return NDBT_FAILED;
+
+      if (res.insertErrorInNode(node, 1000))
+	return NDBT_FAILED;
+      
+      if (res.insertErrorInNode(node, errnos[pos]))
+	return NDBT_FAILED;
+      
+      int val2[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 1 };
+      if (res.dumpStateOneNode(node, val2, 2))
+	return NDBT_FAILED;
+      
+      res.startNodes(&node, 1);
+      res.waitNodesStartPhase(&node, 1, 2);
+      pos++;
+    }
+    pos = 0;
+  }
+
+  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", 
 	 "Test that one node at a time can be stopped and then restarted "\
@@ -1640,6 +1693,9 @@
 }
 TESTCASE("Bug26481", ""){
   INITIALIZER(runBug26481);
+}
+TESTCASE("Bug27003", ""){
+  INITIALIZER(runBug27003);
 }
 NDBT_TESTSUITE_END(testNodeRestart);