Bug#20617891: NDB : SUSPICIOUS HANDLING OF SIGNAL-WAIT TIMEOUT

IN NDBAPI In ndbapi, a signal timeout should be handled by setting the state to timed-out. A timed-out state can then be mapped to the error code 4008. However, the timed-out state is always overwritten to a success state. Modified the signal-timeout code so that the timed-out state is not overwritten.

Author: Priyanka Sangam

mysql-test/suite/ndb/r/ndb_err4012.result

@@ -33,3 +33,14 @@ ERROR HY000: Got error 4012 'Request ndbd time-out, maybe due to high load or co
 SET SESSION debug=@save_debug;
 drop table t4;
 # bulk update ignore is not possible because 'ignore' disables bulk updates
+# mysqld gets timeout error 4008 on CREATE TABLE
+set @save_debug = @@session.debug;
+SET SESSION debug="+d,ndb_timeout_gettabinforeq";
+create table t4(id int primary key, val int)engine=ndb;
+ERROR HY000: Can't create table 'test.t4' (errno: 1625)
+show warnings;
+Level	Code	Message
+Warning	1296	Got error 4008 'Receive from NDB failed' from NDB
+Warning	1625	Bad schema for mysql.ndb_replication table. Message: Unable to retrieve mysql.ndb_replication, logging and conflict r
+Error	1005	Can't create table 'test.t4' (errno: 1625)
+SET SESSION debug=@save_debug;

mysql-test/suite/ndb/t/ndb_err4012.test

@@ -45,3 +45,13 @@ SET SESSION debug=@save_debug;
 drop table t4;
 
 --echo # bulk update ignore is not possible because 'ignore' disables bulk updates
+
+--echo # mysqld gets timeout error 4008 on CREATE TABLE
+--exec $NDB_MGM --no-defaults --ndb-connectstring="$NDB_CONNECTSTRING" -e "all error 6216" >> $NDB_TOOLS_OUTPUT
+set @save_debug = @@session.debug;
+SET SESSION debug="+d,ndb_timeout_gettabinforeq";
+--error 1005
+create table t4(id int primary key, val int)engine=ndb;
+show warnings;
+--exec $NDB_MGM --no-defaults --ndb-connectstring="$NDB_CONNECTSTRING" -e "all error 0" >> $NDB_TOOLS_OUTPUT
+SET SESSION debug=@save_debug;

storage/ndb/src/kernel/blocks/dbdict/Dbdict.cpp

@@ -10185,6 +10185,15 @@ void Dbdict::execGET_TABINFOREQ(Signal* signal)
 
  SectionHandle handle(this, signal);
 
+ if(ERROR_INSERTED(6216))
+ {
+ ndbout_c("Delaying GSN_GET_TABINFOREQ\n");
+ sendSignalWithDelay(reference(), GSN_GET_TABINFOREQ, signal, 10000,
+ signal->length(),
+ &handle);
+ return;
+ }
+
  const bool useLongSig = (req->requestType & GetTabInfoReq::LongSignalConf);
  const bool byName = (req->requestType & GetTabInfoReq::RequestByName);
  const Uint32 transId = req->schemaTransId;

storage/ndb/src/ndbapi/NdbDictionaryImpl.cpp

@@ -2544,10 +2544,18 @@ NdbDictInterface::getTable(class NdbApiSignal * signal,
  Uint32 noOfSections, bool fullyQualifiedNames)
 {
  int errCodes[] = {GetTabInfoRef::Busy, 0 };
+ int timeout = DICT_WAITFOR_TIMEOUT;
+ DBUG_EXECUTE_IF("ndb_timeout_gettabinforeq", {
+ fprintf(stderr, "NdbDictInterface::getTable() times out in dictSignal WAIT_GET_TAB_INFO_REQ\n");
+ timeout = 1000; 
+ });
+
  int r = dictSignal(signal, ptr, noOfSections,
  -1, // any node
  WAIT_GET_TAB_INFO_REQ,
- DICT_WAITFOR_TIMEOUT, 100, errCodes);
+ timeout, // parse stage
+ 100, 
+ errCodes); 
 
  if (r)
  return 0;

storage/ndb/src/ndbapi/trp_client.cpp

@@ -1,5 +1,5 @@
 /*
- Copyright (c) 2010, 2014, Oracle and/or its affiliates. All rights reserved.
+ Copyright (c) 2010, 2015, Oracle and/or its affiliates. All rights reserved.
 
  This program is free software; you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published by
@@ -198,6 +198,7 @@ int PollGuard::wait_for_input_in_loop(int wait_time, bool forceSend)
  else if (state == WAIT_NODE_FAILURE)
  {
  ret_val= -2;
+ m_waiter->set_state(NO_WAIT);
  break;
  }
  if (wait_time == -1)
@@ -222,7 +223,6 @@ int PollGuard::wait_for_input_in_loop(int wait_time, bool forceSend)
  ndbout << "ERR: receiveResponse - theImpl->theWaiter.m_state = ";
  ndbout << m_waiter->get_state() << endl;
 #endif
- m_waiter->set_state(NO_WAIT);
  return ret_val;
 }