bzr commit into mysql-5.1-telco-6.2 branch (jonas:2622) Bug#37338

[jonas]

#At file:///home/jonas/src/telco-6.2/

 2622 jonas@stripped	2008-06-11
      ndb - bug#37338
        Fix weird case with node failure and api failure
        cause subsequent node restart to fail
modified:
  storage/ndb/src/kernel/blocks/suma/Suma.cpp
  storage/ndb/test/ndbapi/test_event.cpp
  storage/ndb/test/run-test/daily-basic-tests.txt

per-file comments:
  storage/ndb/src/kernel/blocks/suma/Suma.cpp
    1) put gci in buffer if resend is ongoing
    2) fix correct interval for resend
       (currently incorrect if no active subscriber)
  storage/ndb/test/ndbapi/test_event.cpp
    testcase
  storage/ndb/test/run-test/daily-basic-tests.txt
    testcase
=== modified file 'storage/ndb/src/kernel/blocks/suma/Suma.cpp'
--- a/storage/ndb/src/kernel/blocks/suma/Suma.cpp	2008-06-11 08:23:36 +0000
+++ b/storage/ndb/src/kernel/blocks/suma/Suma.cpp	2008-06-11 19:49:15 +0000
@@ -3781,12 +3781,13 @@ Suma::execSUB_GCP_COMPLETE_REP(Signal* s
   /**
    * Add GCP COMPLETE REP to buffer
    */
+  bool subscribers = !c_subscriber_nodes.isclear();
   for(Uint32 i = 0; i<c_no_of_buckets; i++)
   {
     if(m_active_buckets.get(i))
       continue;
 
-    if (!c_subscriber_nodes.isclear())
+    if (subscribers || (c_buckets[i].m_state & Bucket::BUCKET_RESEND))
     {
       //Uint32* dst;
       get_buffer_ptr(signal, i, gci, 0);
@@ -5020,7 +5021,7 @@ Suma::start_resend(Signal* signal, Uint3
   }
 
   Uint64 min= bucket->m_max_acked_gci + 1;
-  Uint64 max = pos.m_max_gci;
+  Uint64 max = m_max_seen_gci;
 
   ndbrequire(max <= m_max_seen_gci);
 
@@ -5029,7 +5030,9 @@ Suma::start_resend(Signal* signal, Uint3
     ndbrequire(pos.m_page_id == bucket->m_buffer_tail);
     m_active_buckets.set(buck);
     m_gcp_complete_rep_count ++;
-    ndbout_c("empty bucket -> active");
+    ndbout_c("empty bucket (%u/%u %u/%u) -> active", 
+             Uint32(min >> 32), Uint32(min),
+             Uint32(max >> 32), Uint32(max));
     return;
   }
 

=== modified file 'storage/ndb/test/ndbapi/test_event.cpp'
--- a/storage/ndb/test/ndbapi/test_event.cpp	2008-06-11 08:23:36 +0000
+++ b/storage/ndb/test/ndbapi/test_event.cpp	2008-06-11 19:49:15 +0000
@@ -2579,6 +2579,72 @@ runBug37279(NDBT_Context* ctx, NDBT_Step
   return NDBT_OK;
 }
 
+int
+runBug37338(NDBT_Context* ctx, NDBT_Step* step)
+{
+  NdbRestarter res;
+  if (res.getNumDbNodes() < 2)
+  {
+    ctx->stopTest();
+    return NDBT_OK;
+  }
+
+  int nodeId = res.getDbNodeId(rand() % res.getNumDbNodes());
+
+  Ndb* pNdb = GETNDB(step);
+  NdbDictionary::Dictionary* dict = pNdb->getDictionary();
+  const NdbDictionary::Table* tab = dict->getTable(ctx->getTab()->getName());
+
+  const char * name = "BugXXX";
+  NdbDictionary::Table copy = * tab;
+  copy.setName(name);
+  dict->dropTable(name);
+
+  for (int i = 0; i<ctx->getNumLoops(); i++)
+  {
+    Ndb* ndb0;
+    Ndb_cluster_connection *con0;
+    NdbEventOperation* pOp0;
+    NdbDictionary::Dictionary * dict0;
+
+    cc(&con0, &ndb0);
+    dict0 = ndb0->getDictionary();
+    if (dict0->createTable(copy) != 0)
+    {
+      ndbout << dict0->getNdbError() << endl;
+      return NDBT_FAILED;
+    }
+
+    const NdbDictionary::Table * copyptr = dict0->getTable(name);
+    if (copyptr == 0)
+    {
+      return NDBT_FAILED;
+    }
+    createEvent(ndb0, *copyptr, ctx);
+    pOp0 = createEventOperation(ndb0, *copyptr);
+    dict0 = ndb0->getDictionary();dict->dropTable(name);
+    
+    res.restartOneDbNode(nodeId,
+                         /** initial */ false,
+                         /** nostart */ true,
+                         /** abort   */ true);
+    
+    res.waitNodesNoStart(&nodeId, 1);
+    res.startNodes(&nodeId, 1);
+    if (res.waitClusterStarted())
+    {
+      return NDBT_FAILED;
+    }
+    
+    ndb0->dropEventOperation(pOp0);
+    
+    delete ndb0;
+    delete con0;
+  }
+  
+  return NDBT_OK;
+}
+
 NDBT_TESTSUITE(test_event);
 TESTCASE("BasicEventOperation", 
 	 "Verify that we can listen to Events"
@@ -2762,6 +2828,10 @@ TESTCASE("Bug37279", "")
 {
   INITIALIZER(runBug37279);
 }
+TESTCASE("Bug37338", "")
+{
+  INITIALIZER(runBug37338);
+}
 NDBT_TESTSUITE_END(test_event);
 
 int main(int argc, const char** argv){

=== modified file 'storage/ndb/test/run-test/daily-basic-tests.txt'
--- a/storage/ndb/test/run-test/daily-basic-tests.txt	2008-06-11 08:23:36 +0000
+++ b/storage/ndb/test/run-test/daily-basic-tests.txt	2008-06-11 19:49:15 +0000
@@ -1102,6 +1102,10 @@ max-time: 300
 cmd: test_event
 args: -n Bug37279 T1
 
+max-time: 300
+cmd: test_event
+args: -n Bug37338 T1
+
 # 2008-04-22
 max-time: 1500
 cmd: testNodeRestart