bug#9757 - ndb

  Impl. stupid mgm lock against shutting down 2 nodes simultaniously
  Add some more dump


ndb/src/kernel/blocks/dbdih/DbdihMain.cpp:
  Impl. 2 new dump 
  1) to query nf handling state
  2) that impl. sending of arbitrary signal...
     to be used with extreme care
ndb/src/kernel/main.cpp:
  enable critical and warnings in out file
ndb/src/mgmsrv/MgmtSrvr.cpp:
  call handleStopReply on nfCompleteRep instead
ndb/src/mgmsrv/MgmtSrvr.hpp:
  call handleStopReply on nfCompleteRep instead

4 files changed, 57 insertions, 5 deletions

diff --git a/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp b/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp
index 0bc8351a9db..efd1740651c 100644
--- a/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp
+++ b/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp
@@ -69,6 +69,9 @@
 #include <signaldata/FsOpenReq.hpp>
 #include <DebuggerNames.hpp>
 
+#include <EventLogger.hpp>
+extern EventLogger g_eventLogger;
+
 #define SYSFILE ((Sysfile *)&sysfileData[0])
 
 #define RETURN_IF_NODE_NOT_ALIVE(node) \
@@ -13103,6 +13106,48 @@ Dbdih::execDUMP_STATE_ORD(Signal* signal)
       }
     }
   }
+
+  if(dumpState->args[0] == 7019 && signal->getLength() == 2)
+  {
+    char buf2[8+1];
+    NodeRecordPtr nodePtr;
+    nodePtr.i = signal->theData[1];
+    ptrCheckGuard(nodePtr, MAX_NDB_NODES, nodeRecord);
+    infoEvent("NF Node %d tc: %d lqh: %d dih: %d dict: %d recNODE_FAILREP: %d",
+	      nodePtr.i,
+	      nodePtr.p->dbtcFailCompleted,
+	      nodePtr.p->dblqhFailCompleted,
+	      nodePtr.p->dbdihFailCompleted,
+	      nodePtr.p->dbdictFailCompleted,
+	      nodePtr.p->recNODE_FAILREP);
+    infoEvent(" m_NF_COMPLETE_REP: %s m_nodefailSteps: %s",
+	      nodePtr.p->m_NF_COMPLETE_REP.getText(),
+	      nodePtr.p->m_nodefailSteps.getText(buf2));
+  }
+  
+  if(dumpState->args[0] == 7020 && signal->getLength() > 3)
+  {
+    Uint32 gsn= signal->theData[1];
+    Uint32 block= signal->theData[2];
+    Uint32 length= signal->length() - 3;
+    memmove(signal->theData, signal->theData+3, 4*length);
+    sendSignal(numberToRef(block, getOwnNodeId()), gsn, signal, length, JBB);
+    
+    warningEvent("-- SENDING CUSTOM SIGNAL --");
+    char buf[100], buf2[100];
+    buf2[0]= 0;
+    for(Uint32 i = 0; i<length; i++)
+    {
+      snprintf(buf, 100, "%s %.8x", buf2, signal->theData[i]);
+      snprintf(buf2, 100, "%s", buf);
+    }
+    warningEvent("gsn: %d block: %s, length: %d theData: %s", 
+		 gsn, getBlockName(block, "UNKNOWN"), length, buf);
+
+    g_eventLogger.warning("-- SENDING CUSTOM SIGNAL --");
+    g_eventLogger.warning("gsn: %d block: %s, length: %d theData: %s", 
+			  gsn, getBlockName(block, "UNKNOWN"), length, buf);
+  }
   
   if(dumpState->args[0] == DumpStateOrd::DihDumpLCPState){
     infoEvent("-- Node %d LCP STATE --", getOwnNodeId());
diff --git a/ndb/src/kernel/main.cpp b/ndb/src/kernel/main.cpp
index e6e0a7ca877..1e5cd5270e4 100644
--- a/ndb/src/kernel/main.cpp
+++ b/ndb/src/kernel/main.cpp
@@ -58,7 +58,9 @@ int main(int argc, char** argv)
   // Print to stdout/console
   g_eventLogger.createConsoleHandler();
   g_eventLogger.setCategory("NDB");
+  g_eventLogger.enable(Logger::LL_ON, Logger::LL_CRITICAL);
   g_eventLogger.enable(Logger::LL_ON, Logger::LL_ERROR);
+  g_eventLogger.enable(Logger::LL_ON, Logger::LL_WARNING);
 
   globalEmulatorData.create();
 
diff --git a/ndb/src/mgmsrv/MgmtSrvr.cpp b/ndb/src/mgmsrv/MgmtSrvr.cpp
index 3bc00b18f50..98cdcdca2f7 100644
--- a/ndb/src/mgmsrv/MgmtSrvr.cpp
+++ b/ndb/src/mgmsrv/MgmtSrvr.cpp
@@ -2061,7 +2061,7 @@ MgmtSrvr::handleStopReply(NodeId nodeId, Uint32 errCode)
 }
 
 void
-MgmtSrvr::handleStatus(NodeId nodeId, bool alive)
+MgmtSrvr::handleStatus(NodeId nodeId, bool alive, bool nfComplete)
 {
   DBUG_ENTER("MgmtSrvr::handleStatus");
   Uint32 theData[25];
@@ -2070,9 +2070,15 @@ MgmtSrvr::handleStatus(NodeId nodeId, bool alive)
     m_started_nodes.push_back(nodeId);
     theData[0] = EventReport::Connected;
   } else {
-    handleStopReply(nodeId, 0);
     theData[0] = EventReport::Disconnected;
+    if(nfComplete)
+    {
+      theData[0] = EventReport::NodeFailCompleted;
+      handleStopReply(nodeId, 0);
+      return;
+    }
   }
+  
   eventReport(_ownNodeId, theData);
   DBUG_VOID_RETURN;
 }
@@ -2097,8 +2103,7 @@ MgmtSrvr::nodeStatusNotification(void* mgmSrv, Uint32 nodeId,
 {
   DBUG_ENTER("MgmtSrvr::nodeStatusNotification");
   DBUG_PRINT("enter",("nodeid= %d, alive= %d, nfComplete= %d", nodeId, alive, nfComplete));
-  if(!(!alive && nfComplete))
-    ((MgmtSrvr*)mgmSrv)->handleStatus(nodeId, alive);
+  ((MgmtSrvr*)mgmSrv)->handleStatus(nodeId, alive, nfComplete);
   DBUG_VOID_RETURN;
 }
 
diff --git a/ndb/src/mgmsrv/MgmtSrvr.hpp b/ndb/src/mgmsrv/MgmtSrvr.hpp
index 6378e763363..a05b29b7f31 100644
--- a/ndb/src/mgmsrv/MgmtSrvr.hpp
+++ b/ndb/src/mgmsrv/MgmtSrvr.hpp
@@ -573,7 +573,7 @@ private:
   // Returns: -
   //**************************************************************************
 
-  void handleStatus(NodeId nodeId, bool alive);
+  void handleStatus(NodeId nodeId, bool alive, bool nfComplete);
   //**************************************************************************
   // Description: Handle the death of a process
   // Parameters: