bk commit into 4.1 tree (jonas:1.2548) BUG#15303

[jonas]

Below is the list of changes that have just been committed into a local
4.1 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, 2006-09-26 15:20:10+02:00, jonas@stripped +1 -0
  ndb - bug#15303
    Fix take-over during SR, remove coupling to lcp/gcp (if systemRestartOngoing())

  ndb/src/kernel/blocks/dbdih/DbdihMain.cpp@stripped, 2006-09-26 15:20:09+02:00, jonas@stripped +30 -7
    Fix take-over during SR

# 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/41-work

--- 1.39/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp	2006-09-26 15:20:13 +02:00
+++ 1.40/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp	2006-09-26 15:20:13 +02:00
@@ -1265,9 +1265,9 @@
       if (isMaster()) {
 	jam();
 	systemRestartTakeOverLab(signal);
-	if (anyActiveTakeOver() && false) {
+	if (anyActiveTakeOver())
+	{
 	  jam();
-	  ndbout_c("1 - anyActiveTakeOver == true");
 	  return;
 	}
       }
@@ -2260,6 +2260,8 @@
 	// NOT ACTIVE NODES THAT HAVE NOT YET BEEN TAKEN OVER NEEDS TAKE OVER
 	// IMMEDIATELY. IF WE ARE ALIVE WE TAKE OVER OUR OWN NODE.
 	/*-------------------------------------------------------------------*/
+	infoEvent("Take over of node %d started", 
+		  nodePtr.i);
 	startTakeOver(signal, RNIL, nodePtr.i, nodePtr.i);
       }//if
       break;
@@ -2372,6 +2374,12 @@
      *--------------------------------------------------------------------*/
     Uint32 takeOverNode = Sysfile::getTakeOverNode(startNodeId, 
 						   SYSFILE->takeOver);
+    if(takeOverNode == 0){
+      jam();
+      warningEvent("Bug in take-over code restarting");
+      takeOverNode = startNodeId;
+    }
+
     startTakeOver(signal, RNIL, startNodeId, takeOverNode);
     break;
   }
@@ -2525,7 +2533,14 @@
   Sysfile::setTakeOverNode(takeOverPtr.p->toFailedNode, SYSFILE->takeOver,
 			   startNode);
   takeOverPtr.p->toMasterStatus = TakeOverRecord::TO_START_COPY;
-  
+
+  if (getNodeState().getSystemRestartInProgress())
+  {
+    jam();
+    checkToCopy();
+    checkToCopyCompleted(signal);
+    return;
+  }
   cstartGcpNow = true;
 }//Dbdih::startTakeOver()
 
@@ -3273,6 +3288,18 @@
   signal->theData[1] = takeOverPtr.p->toStartingNode;
   sendSignal(CMVMI_REF, GSN_EVENT_REP, signal, 2, JBB);
 
+  if (getNodeState().getSystemRestartInProgress())
+  {
+    jam();
+    infoEvent("Take over of node %d complete", takeOverPtr.p->toStartingNode);
+    setNodeActiveStatus(takeOverPtr.p->toStartingNode, Sysfile::NS_Active);
+    takeOverPtr.p->toMasterStatus = TakeOverRecord::WAIT_LCP;
+    takeOverCompleted(takeOverPtr.p->toStartingNode);
+    checkToCopy();
+    checkToCopyCompleted(signal);
+    return;
+  }
+  
   c_lcpState.immediateLcpStart = true;
   takeOverPtr.p->toMasterStatus = TakeOverRecord::WAIT_LCP;
   
@@ -3379,16 +3406,12 @@
   }//if
   endTakeOver(takeOverPtr.i);
 
-  ndbout_c("2 - endTakeOver");
   if (cstartPhase == ZNDB_SPH4) {
     jam();
-    ndbrequire(false);
     if (anyActiveTakeOver()) {
       jam();
-      ndbout_c("4 - anyActiveTakeOver == true");
       return;
     }//if
-    ndbout_c("5 - anyActiveTakeOver == false -> ndbsttorry10Lab");
     ndbsttorry10Lab(signal, __LINE__);
     return;
   }//if