ndb - bug#15685

  Error in abort handling in TC when timeout during abort
  


ndb/src/kernel/blocks/ERROR_codes.txt:
  New error codes
ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
  New error codes
ndb/src/kernel/blocks/dbtc/DbtcMain.cpp:
  Dont release transaction record to early
ndb/test/ndbapi/testNodeRestart.cpp:
  Test case
ndb/test/run-test/daily-basic-tests.txt:
  Test case

5 files changed, 67 insertions, 3 deletions

diff --git a/ndb/src/kernel/blocks/ERROR_codes.txt b/ndb/src/kernel/blocks/ERROR_codes.txt
index e11c5ef4c5d..5d7c8d758fc 100644
--- a/ndb/src/kernel/blocks/ERROR_codes.txt
+++ b/ndb/src/kernel/blocks/ERROR_codes.txt
@@ -165,6 +165,7 @@ handling in DBTC to ensure that node failures are also well handled in
 time-out handling. They can also be used to test multiple node failure
 handling.
 
+
 ERROR CODES FOR TESTING TIME-OUT HANDLING IN DBLQH
 -------------------------------------------------
 5011:
@@ -201,6 +202,9 @@ Delay execution of ABORTREQ signal 2 seconds to generate time-out.
 8048: Make TC not choose own node for simple/dirty read
 5041: Crash is receiving simple read from other TC on different node
 
+5100,5101: Drop ABORT req in primary replica
+           Crash on "next" ABORT
+
 ERROR CODES FOR TESTING TIME-OUT HANDLING IN DBTC
 -------------------------------------------------
 8040:
diff --git a/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp b/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp
index 6cae512274d..fc6a470e0ef 100644
--- a/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp
+++ b/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp
@@ -5870,12 +5870,21 @@ void Dblqh::execABORT(Signal* signal)
     warningReport(signal, 8);
     return;
   }//if
+  
+  TcConnectionrec * const regTcPtr = tcConnectptr.p;
+
+  if (ERROR_INSERTED(5100))
+  {
+    SET_ERROR_INSERT_VALUE(5101);
+    return;
+  }
+  CRASH_INSERTION2(5101, regTcPtr->nextReplica != ZNIL);
+  
 /* ------------------------------------------------------------------------- */
 /*A GUIDING DESIGN PRINCIPLE IN HANDLING THESE ERROR SITUATIONS HAVE BEEN    */
 /*KEEP IT SIMPLE. THUS WE RATHER INSERT A WAIT AND SET THE ABORT_STATE TO    */
 /*ACTIVE RATHER THAN WRITE NEW CODE TO HANDLE EVERY SPECIAL SITUATION.       */
 /* ------------------------------------------------------------------------- */
-  TcConnectionrec * const regTcPtr = tcConnectptr.p;
   if (regTcPtr->nextReplica != ZNIL) {
 /* ------------------------------------------------------------------------- */
 // We will immediately send the ABORT message also to the next LQH node in line.
diff --git a/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp b/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp
index 30a50e7695e..d9d1f01b213 100644
--- a/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp
+++ b/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp
@@ -6129,7 +6129,6 @@ void Dbtc::timeOutFoundLab(Signal* signal, Uint32 TapiConPtr)
 	<< " - place: " << c_apiConTimer_line[apiConnectptr.i]);
   switch (apiConnectptr.p->apiConnectstate) {
   case CS_STARTED:
-    ndbrequire(c_apiConTimer_line[apiConnectptr.i] != 3615);
     if(apiConnectptr.p->lqhkeyreqrec == apiConnectptr.p->lqhkeyconfrec){
       jam();
       /*
@@ -6389,8 +6388,8 @@ void Dbtc::sendAbortedAfterTimeout(Signal* signal, int Tcheck)
 	warningEvent(buf);
 	ndbout_c(buf);
 	ndbrequire(false);
+	releaseAbortResources(signal);
       }
-      releaseAbortResources(signal);
       return;
     }//if
     TloopCount++;
diff --git a/ndb/test/ndbapi/testNodeRestart.cpp b/ndb/test/ndbapi/testNodeRestart.cpp
index 5daf1fcfea0..a741e6233d9 100644
--- a/ndb/test/ndbapi/testNodeRestart.cpp
+++ b/ndb/test/ndbapi/testNodeRestart.cpp
@@ -496,6 +496,45 @@ int runBug15632(NDBT_Context* ctx, NDBT_Step* step){
   return NDBT_OK;
 }
 
+int runBug15685(NDBT_Context* ctx, NDBT_Step* step){
+
+  Ndb* pNdb = GETNDB(step);
+  HugoOperations hugoOps(*ctx->getTab());
+  NdbRestarter restarter;
+
+  HugoTransactions hugoTrans(*ctx->getTab());
+  if (hugoTrans.loadTable(GETNDB(step), 10) != 0){
+    return NDBT_FAILED;
+  }
+
+  if(hugoOps.startTransaction(pNdb) != 0)
+    goto err;
+  
+  if(hugoOps.pkUpdateRecord(pNdb, 0, 1, rand()) != 0)
+    goto err;
+
+  if(hugoOps.execute_NoCommit(pNdb) != 0)
+    goto err;
+
+  if (restarter.insertErrorInAllNodes(5100))
+    return NDBT_FAILED;
+  
+  hugoOps.execute_Rollback(pNdb);
+
+  if (restarter.waitClusterStarted() != 0)
+    goto err;
+
+  if (restarter.insertErrorInAllNodes(0))
+    return NDBT_FAILED;
+  
+  ctx->stopTest();
+  return NDBT_OK;
+  
+err:
+  ctx->stopTest();
+  return NDBT_FAILED;
+}
+
 
 NDBT_TESTSUITE(testNodeRestart);
 TESTCASE("NoLoad", 
@@ -776,6 +815,11 @@ TESTCASE("Bug15632",
   STEP(runBug15632);
   FINALIZER(runClearTable);
 }
+TESTCASE("Bug15685",
+	 "Test bug with NF during abort"){
+  STEP(runBug15685);
+  FINALIZER(runClearTable);
+}
 NDBT_TESTSUITE_END(testNodeRestart);
 
 int main(int argc, const char** argv){
diff --git a/ndb/test/run-test/daily-basic-tests.txt b/ndb/test/run-test/daily-basic-tests.txt
index c538b01c6d7..6378b4a06d3 100644
--- a/ndb/test/run-test/daily-basic-tests.txt
+++ b/ndb/test/run-test/daily-basic-tests.txt
@@ -438,6 +438,14 @@ max-time: 500
 cmd: testNodeRestart
 args: -n Bug15587 T1
 
+max-time: 500
+cmd: testNodeRestart
+args: -n Bug15632 T1
+
+max-time: 500
+cmd: testNodeRestart
+args: -n Bug15685 T1
+
 # OLD FLEX
 max-time: 500
 cmd: flexBench