ndb - bug#24543

  Resend if master dies during alloc node id


storage/ndb/src/kernel/blocks/ERROR_codes.txt:
  Add new error code
storage/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp:
  Add new error code
storage/ndb/src/mgmsrv/MgmtSrvr.cpp:
  Make sure to resend if master died
storage/ndb/test/ndbapi/testNodeRestart.cpp:
  add testcase
storage/ndb/test/run-test/daily-basic-tests.txt:
  add testcase

storage/ndb/src/kernel/blocks/ERROR_codes.txt

@@ -21,6 +21,9 @@ Crash president when he starts to run in ArbitState 1-9.
 
 910: Crash new president after node crash
 
+934 : Crash president in ALLOC_NODE_ID_REQ
+
+
 ERROR CODES FOR TESTING NODE FAILURE, GLOBAL CHECKPOINT HANDLING:
 -----------------------------------------------------------------
 

storage/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp

@@ -4832,6 +4832,11 @@ Qmgr::execALLOC_NODEID_REQ(Signal * signal)
       return;
     }
 
+    if (ERROR_INSERTED(934) && nodeId != getOwnNodeId())
+    {
+      CRASH_INSERTION(934);
+    }
+    
     opAllocNodeIdReq.m_req = *req;
     opAllocNodeIdReq.m_error = 0;
     opAllocNodeIdReq.m_connectCount = getNodeInfo(refToNode(senderRef)).m_connectCount;

storage/ndb/src/mgmsrv/MgmtSrvr.cpp

@@ -2051,7 +2051,10 @@ MgmtSrvr::alloc_node_id_req(NodeId free_node_id, enum ndb_mgm_node_type type)
       ndbout_c("Node %d fail completed", rep->failedNodeId);
 #endif
       if (rep->failedNodeId == nodeId)
+      {
+	do_send = 1;
         nodeId = 0;
+      }
       continue;
     }
     case GSN_NODE_FAILREP:{

storage/ndb/test/ndbapi/testNodeRestart.cpp

@@ -966,6 +966,37 @@ runBug21271(NDBT_Context* ctx, NDBT_Step* step){
   return NDBT_OK;
 }
 
+int 
+runBug24543(NDBT_Context* ctx, NDBT_Step* step){
+  NdbRestarter restarter;
+  
+  int val2[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 1 };
+  if (restarter.dumpStateAllNodes(val2, 2))
+    return NDBT_FAILED;
+
+  int nodes[2];
+  nodes[0] = restarter.getMasterNodeId();
+  restarter.insertErrorInNode(nodes[0], 934);
+
+  nodes[1] = restarter.getRandomNodeOtherNodeGroup(nodes[0], rand());
+  if (nodes[1] == -1)
+  {
+    nodes[1] = restarter.getRandomNodeSameNodeGroup(nodes[0], rand());
+  }
+  
+  restarter.restartOneDbNode(nodes[1], false, true, true);
+  if (restarter.waitNodesNoStart(nodes, 2))
+    return NDBT_FAILED;
+  
+  restarter.startNodes(nodes, 2);
+  if (restarter.waitNodesStarted(nodes, 2))
+  {
+    return NDBT_FAILED;
+  }
+  
+  return NDBT_OK;
+}
+
 
 NDBT_TESTSUITE(testNodeRestart);
 TESTCASE("NoLoad", 
@@ -1279,6 +1310,10 @@ TESTCASE("Bug20185",
   STEP(runBug20185);
   FINALIZER(runClearTable);
 }
+TESTCASE("Bug24543", "")
+{
+  INITIALIZER(runBug24543);
+}
 TESTCASE("Bug21271",
 	 ""){
   INITIALIZER(runLoadTable);

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

@@ -744,6 +744,10 @@ max-time: 1000
 cmd: testSRBank
 args: -n Mix -l 300 -r 15 T1
 
+max-time: 300
+cmd: testNodeRestart
+args: -n Bug24543 T1
+
 # OLD FLEX
 max-time: 500
 cmd: flexBench