Merge mskold@bk-internal.mysql.com:/home/bk/mysql-5.1-ndb

into  mysql.com:/windows/Linux_space/MySQL/mysql-5.1-new-ndb


sql/ha_ndbcluster.cc:
  Auto merged
storage/ndb/include/ndbapi/Ndb.hpp:
  Auto merged

extra/perror.c

@@ -25,6 +25,7 @@
 #ifdef WITH_NDBCLUSTER_STORAGE_ENGINE
 #include "../storage/ndb/src/ndbapi/ndberror.c"
 #include "../storage/ndb/src/kernel/error/ndbd_exit_codes.c"
+#include "../storage/ndb/include/mgmapi/mgmapi_error.h"
 #endif
 
 static my_bool verbose, print_all_codes;
@@ -32,6 +33,20 @@ static my_bool verbose, print_all_codes;
 #ifdef WITH_NDBCLUSTER_STORAGE_ENGINE
 static my_bool ndb_code;
 static char ndb_string[1024];
+int mgmapi_error_string(int err_no, char *str, int size)
+{
+  int i;
+  for (i= 0; i < ndb_mgm_noOfErrorMsgs; i++)
+  {
+    if (ndb_mgm_error_msgs[i].code == err_no)
+    {
+      my_snprintf(str, size-1, "%s", ndb_mgm_error_msgs[i].msg);
+      str[size-1]= '\0';
+      return 0;
+    }
+  }
+  return -1;
+}
 #endif
 
 static struct my_option my_long_options[] =
@@ -238,8 +253,9 @@ int main(int argc,char *argv[])
 #ifdef WITH_NDBCLUSTER_STORAGE_ENGINE
       if (ndb_code)
       {
-	if ((ndb_error_string(code, ndb_string, sizeof(ndb_string)) < 0) &&
+        if ((ndb_error_string(code, ndb_string, sizeof(ndb_string)) < 0) &&
-	    (ndbd_exit_string(code, ndb_string, sizeof(ndb_string)) < 0))
+            (ndbd_exit_string(code, ndb_string, sizeof(ndb_string)) < 0) &&
+            (mgmapi_error_string(code, ndb_string, sizeof(ndb_string)) < 0))
 	{
           msg= 0;
 	}

mysql-test/ndb/ndb_config_1_node.ini

@@ -10,7 +10,8 @@ DataDir= CHOOSE_FILESYSTEM
 MaxNoOfOrderedIndexes= CHOOSE_MaxNoOfOrderedIndexes
 MaxNoOfAttributes= CHOOSE_MaxNoOfAttributes
 TimeBetweenGlobalCheckpoints= 500
-NoOfFragmentLogFiles= 3
+NoOfFragmentLogFiles= 8
+FragmentLogFileSize= 6M
 DiskPageBufferMemory= CHOOSE_DiskPageBufferMemory 
 
 #

mysql-test/ndb/ndb_config_2_node.ini

@@ -10,8 +10,10 @@ DataDir= CHOOSE_FILESYSTEM
 MaxNoOfOrderedIndexes= CHOOSE_MaxNoOfOrderedIndexes
 MaxNoOfAttributes= CHOOSE_MaxNoOfAttributes
 TimeBetweenGlobalCheckpoints= 500
-NoOfFragmentLogFiles= 3
+NoOfFragmentLogFiles= 4
+FragmentLogFileSize=12M
 DiskPageBufferMemory= CHOOSE_DiskPageBufferMemory
+ODirect= 1
 # the following parametes just function as a small regression
 # test that the parameter exists
 InitialNoOfOpenFiles= 27

mysql-test/r/ndb_backup_print.result

@@ -0,0 +1,64 @@
+use test;
+drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9,t10;
+Connected to Management Server at: :
+Waiting for completed, this may take several minutes
+Backup  started from node 
+Backup  started from node  completed
+ StartGCP:  StopGCP: 
+ #Records:  #LogRecords: 
+ Data:  bytes Log:  bytes
+create table t1
+(pk int key
+,a1 BIT(1), a2 BIT(5), a3 BIT(33), a4 BIT(63), a5 BIT(64)
+,b1 TINYINT, b2 TINYINT UNSIGNED
+,c1 SMALLINT, c2 SMALLINT UNSIGNED
+,d1 INT, d2 INT UNSIGNED
+,e1 BIGINT, e2 BIGINT UNSIGNED
+,f1 CHAR(1) BINARY, f2 CHAR(32) BINARY, f3 CHAR(255) BINARY
+,g1 VARCHAR(32) BINARY, g2 VARCHAR(255) BINARY, g3 VARCHAR(1000) BINARY
+,h1 BINARY(1), h2 BINARY(8), h3 BINARY(255)
+,i1 VARBINARY(32), i2 VARBINARY(255), i3 VARBINARY(1000)
+) engine ndb;
+insert into t1 values
+(1
+,0x1, 0x17, 0x789a, 0x789abcde, 0xfedc0001
+,127, 255
+,32767, 65535
+,2147483647, 4294967295
+,9223372036854775807, 18446744073709551615
+,'1','12345678901234567890123456789012','123456789'
+  ,'1','12345678901234567890123456789012','123456789'
+  ,0x12,0x123456789abcdef0, 0x012345
+,0x12,0x123456789abcdef0, 0x00123450
+);
+insert into t1 values
+(2
+,0, 0, 0, 0, 0
+,-128, 0
+,-32768, 0
+,-2147483648, 0
+,-9223372036854775808, 0
+,'','',''
+  ,'','',''
+  ,0x0,0x0,0x0
+,0x0,0x0,0x0
+);
+insert into t1 values
+(3
+,NULL,NULL,NULL,NULL,NULL
+,NULL,NULL
+,NULL,NULL
+,NULL,NULL
+,NULL,NULL
+,NULL,NULL,NULL
+,NULL,NULL,NULL
+,NULL,NULL,NULL
+,NULL,NULL,NULL
+);
+Connected to Management Server at: :
+Waiting for completed, this may take several minutes
+Backup  started from node 
+Backup  started from node  completed
+ StartGCP:  StopGCP: 
+ #Records:  #LogRecords: 
+ Data:  bytes Log:  bytes

mysql-test/r/ndb_basic.result

@@ -6,6 +6,34 @@ attr1 INT NOT NULL,
 attr2 INT,
 attr3 VARCHAR(10)
 ) ENGINE=ndbcluster;
+drop table t1;
+SHOW GLOBAL STATUS LIKE 'ndb%';
+Variable_name	Value
+Ndb_cluster_node_id	#
+Ndb_config_from_host	#
+Ndb_config_from_port	#
+Ndb_number_of_data_nodes	#
+SHOW GLOBAL VARIABLES LIKE 'ndb%';
+Variable_name	Value
+ndb_autoincrement_prefetch_sz	#
+ndb_cache_check_time	#
+ndb_connectstring	#
+ndb_extra_logging	#
+ndb_force_send	#
+ndb_index_stat_cache_entries	#
+ndb_index_stat_enable	#
+ndb_index_stat_update_freq	#
+ndb_report_thresh_binlog_epoch_slip	#
+ndb_report_thresh_binlog_mem_usage	#
+ndb_use_copying_alter_table	#
+ndb_use_exact_count	#
+ndb_use_transactions	#
+CREATE TABLE t1 (
+pk1 INT NOT NULL PRIMARY KEY,
+attr1 INT NOT NULL,
+attr2 INT,
+attr3 VARCHAR(10)
+) ENGINE=ndbcluster;
 SHOW INDEX FROM t1;
 Table	Non_unique	Key_name	Seq_in_index	Column_name	Collation	Cardinality	Sub_part	Packed	Null	Index_type	Comment
 t1	0	PRIMARY	1	pk1	A	0	NULL	NULL		BTREE	

mysql-test/r/ndb_restore.result

@@ -18,7 +18,7 @@ CREATE TABLE `t2_c` (
 PRIMARY KEY  (`capgotod`),
 KEY `i quadaddsvr` (`gotod`)
 ) ENGINE=ndbcluster DEFAULT CHARSET=latin1;
-INSERT INTO `t2_c` VALUES (500,4,'','q3.net','addavp:MK_CASELECTOR=1','postorod rattoaa'),(2,1,'4','','addavp:MK_BRANDTAD=345','REDS Brandtad'),(3,2,'4','q3.net','execorder','fixedRatediPO REDS'),(1,1,'3','','addavp:MK_BRANDTAD=123','TEST Brandtad'),(6,5,'','told.q3.net','addavp:MK_BRANDTAD=123','Brandtad Toldzone'),(4,3,'3','q3.net','addavp:MK_POOLHINT=2','ratedi PO TEST');
+INSERT INTO `t2_c` VALUES (500,4,'','q3.net','addavp:MK_CASELECTOR=1','postorod rattoaa'),(2,1,'4','','addavp:MK_BRANDTAD=345','REDS Brandtad'),(3,2,'4','q3.net','execorder','fixedRatediPO REDS'),(1,1,'3','','addavp:MK_BRANDTAD=123','TEST Brandtad'),(6,5,'','told.q3.net','addavp:MK_BRANDTAD=123','Brandtad Toldzone'),(4,3,'3','q3.net','addavp:MK_POOLHINT=2','ratedi PO TEST'),(5,0,'',NULL,NULL,'');
 CREATE TABLE `t3_c` (
 `CapGoaledatta` smallint(5) unsigned NOT NULL default '0',
 `capgotod` smallint(5) unsigned NOT NULL default '0',
@@ -154,15 +154,15 @@ count(*)
 5
 select count(*) from t2;
 count(*)
-6
+7
 select count(*) from t2_c;
 count(*)
-6
+7
 select count(*)
 from (select * from t2 union 
 select * from t2_c) a;
 count(*)
-6
+7
 select count(*) from t3;
 count(*)
 4
@@ -286,15 +286,15 @@ count(*)
 5
 select count(*) from t2;
 count(*)
-6
+7
 select count(*) from t2_c;
 count(*)
-6
+7
 select count(*)
 from (select * from t2 union 
 select * from t2_c) a;
 count(*)
-6
+7
 select count(*) from t3;
 count(*)
 4
@@ -386,15 +386,15 @@ count(*)
 5
 select count(*) from t2;
 count(*)
-6
+7
 select count(*) from t2_c;
 count(*)
-6
+7
 select count(*)
 from (select * from t2 union 
 select * from t2_c) a;
 count(*)
-6
+7
 select count(*) from t3;
 count(*)
 4

mysql-test/t/disabled.def

@@ -39,5 +39,5 @@ synchronization          : Bug#24529  	Test 'synchronization' fails on Mac pushb
 
 #rpl_ndb_dd_advance	 : Bug#25913 rpl_ndb_dd_advance fails randomly
 
-rpl_ndb_stm_innodb       : Bug#26783
 ndb_partition_error2	 : HF is not sure if the test can work as internded on all the platforms
+mysql_upgrade		 : Bug#28560 test links to /usr/local/mysql/lib libraries, causes non-determinism and failures on ABI breakage

mysql-test/t/ndb_backup_print.test

@@ -0,0 +1,66 @@
+-- source include/have_ndb.inc
+-- source include/ndb_default_cluster.inc
+-- source include/not_embedded.inc
+
+--disable_warnings
+use test;
+drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9,t10;
+--enable_warnings
+
+#NO.1 test output of backup
+--exec $NDB_TOOLS_DIR/../src/mgmclient/ndb_mgm -e "start backup" |sed -e 's/[0-9]//g' |sed -e 's/localhost//g' |sed -e 's/\.\.\.*//g'
+
+create table t1
+ (pk int key
+  ,a1 BIT(1), a2 BIT(5), a3 BIT(33), a4 BIT(63), a5 BIT(64)
+  ,b1 TINYINT, b2 TINYINT UNSIGNED
+  ,c1 SMALLINT, c2 SMALLINT UNSIGNED
+  ,d1 INT, d2 INT UNSIGNED
+  ,e1 BIGINT, e2 BIGINT UNSIGNED
+  ,f1 CHAR(1) BINARY, f2 CHAR(32) BINARY, f3 CHAR(255) BINARY
+  ,g1 VARCHAR(32) BINARY, g2 VARCHAR(255) BINARY, g3 VARCHAR(1000) BINARY
+  ,h1 BINARY(1), h2 BINARY(8), h3 BINARY(255)
+  ,i1 VARBINARY(32), i2 VARBINARY(255), i3 VARBINARY(1000)
+ ) engine ndb;
+
+insert into t1 values
+ (1
+  ,0x1, 0x17, 0x789a, 0x789abcde, 0xfedc0001
+  ,127, 255
+  ,32767, 65535
+  ,2147483647, 4294967295
+  ,9223372036854775807, 18446744073709551615
+  ,'1','12345678901234567890123456789012','123456789'
+  ,'1','12345678901234567890123456789012','123456789'
+  ,0x12,0x123456789abcdef0, 0x012345
+  ,0x12,0x123456789abcdef0, 0x00123450
+ );
+
+insert into t1 values
+ (2
+  ,0, 0, 0, 0, 0
+  ,-128, 0
+  ,-32768, 0
+  ,-2147483648, 0
+  ,-9223372036854775808, 0
+  ,'','',''
+  ,'','',''
+  ,0x0,0x0,0x0
+  ,0x0,0x0,0x0
+ );
+
+insert into t1 values
+ (3
+  ,NULL,NULL,NULL,NULL,NULL
+  ,NULL,NULL
+  ,NULL,NULL
+  ,NULL,NULL
+  ,NULL,NULL
+  ,NULL,NULL,NULL
+  ,NULL,NULL,NULL
+  ,NULL,NULL,NULL
+  ,NULL,NULL,NULL
+ );
+
+#NO.2 test output of backup after some simple SQL operations
+--exec $NDB_TOOLS_DIR/../src/mgmclient/ndb_mgm -e "start backup" |sed -e 's/[0-9]//g' |sed -e 's/localhost//g' |sed -e 's/\.\.\.*//g'

mysql-test/t/ndb_basic.test

@@ -6,22 +6,30 @@ DROP TABLE IF EXISTS t1,t2,t3,t4,t5,t6,t7;
 drop database if exists mysqltest;
 --enable_warnings
 
-## workaround for bug#16445
+# workaround for bug#16445
-## remove to reproduce bug and run tests from ndb start
+# remove to reproduce bug and run tests from ndb start
-## and with ndb_autodiscover disabled. Fails on Linux 50 % of the times
+# and with ndb_autodiscover disabled. Fails on Linux 50 % of the times
-#CREATE TABLE t1 (
+CREATE TABLE t1 (
-#  pk1 INT NOT NULL PRIMARY KEY,
+  pk1 INT NOT NULL PRIMARY KEY,
-#  attr1 INT NOT NULL,
+  attr1 INT NOT NULL,
-#  attr2 INT,
+  attr2 INT,
-#  attr3 VARCHAR(10)
+  attr3 VARCHAR(10)
-#) ENGINE=ndbcluster;
+) ENGINE=ndbcluster;
-#drop table t1;
+drop table t1;
 
 #
 # Basic test to show that the NDB 
 # table handler is working
 #
 
+#
+# Show status and variables
+#
+--replace_column 2 #
+SHOW GLOBAL STATUS LIKE 'ndb%';
+--replace_column 2 #
+SHOW GLOBAL VARIABLES LIKE 'ndb%';
+
 #
 # Create a normal table with primary key
 #

mysql-test/t/ndb_restore.test

@@ -33,7 +33,7 @@ CREATE TABLE `t2_c` (
   PRIMARY KEY  (`capgotod`),
   KEY `i quadaddsvr` (`gotod`)
 ) ENGINE=ndbcluster DEFAULT CHARSET=latin1;
-INSERT INTO `t2_c` VALUES (500,4,'','q3.net','addavp:MK_CASELECTOR=1','postorod rattoaa'),(2,1,'4','','addavp:MK_BRANDTAD=345','REDS Brandtad'),(3,2,'4','q3.net','execorder','fixedRatediPO REDS'),(1,1,'3','','addavp:MK_BRANDTAD=123','TEST Brandtad'),(6,5,'','told.q3.net','addavp:MK_BRANDTAD=123','Brandtad Toldzone'),(4,3,'3','q3.net','addavp:MK_POOLHINT=2','ratedi PO TEST');
+INSERT INTO `t2_c` VALUES (500,4,'','q3.net','addavp:MK_CASELECTOR=1','postorod rattoaa'),(2,1,'4','','addavp:MK_BRANDTAD=345','REDS Brandtad'),(3,2,'4','q3.net','execorder','fixedRatediPO REDS'),(1,1,'3','','addavp:MK_BRANDTAD=123','TEST Brandtad'),(6,5,'','told.q3.net','addavp:MK_BRANDTAD=123','Brandtad Toldzone'),(4,3,'3','q3.net','addavp:MK_POOLHINT=2','ratedi PO TEST'),(5,0,'',NULL,NULL,'');
 
 # Added ROW_FORMAT=FIXED to use below to see that setting is preserved
 # by restore

sql/ha_ndbcluster.cc

@@ -4248,8 +4248,6 @@ THR_LOCK_DATA **ha_ndbcluster::store_lock(THD *thd,
 extern MASTER_INFO *active_mi;
 static int ndbcluster_update_apply_status(THD *thd, int do_update)
 {
-  return 0;
-
   Thd_ndb *thd_ndb= get_thd_ndb(thd);
   Ndb *ndb= thd_ndb->ndb;
   NDBDICT *dict= ndb->getDictionary();

storage/ndb/MAINTAINERS

@@ -0,0 +1,163 @@
+MySQL Cluster MAINTAINERS
+-------------------------
+
+This is a list of knowledgable people in parts of the NDB code.
+
+In changing that area of code, you probably want to talk to the
+people who know a lot about it to look over the patch.
+
+When sending patches and queries, always CC the mailing list.
+
+If no list specified, assume internals@lists.mysql.com
+
+P: Person
+M: Mail
+L: Mailing list
+W: Web page with status/info
+C: Comment
+SRC: Source directory (relative to this directory)
+T: SCM tree type and location
+S: Status, one of:
+
+	Supported: Somebody is paid to maintain this.
+	Maintained: Not their primary job, but maintained.
+	Orphan: No current obvious maintainer.
+	Obsolete: Replaced by something else.
+
+-------------------------------------------------------------
+
+Binlog Injector
+SRC: ha_ndbcluster_binlog.cc
+C: see also row based replication
+P: Stewart Smith
+M: stewart@mysql.com
+C: Original author
+P: Tomas Ulin
+M: tomas@mysql.com
+C: Lots of updates
+P: Martin Skold
+M: martin@mysql.com
+C: Metadata ops
+S: Supported
+
+BLOBs
+SRC: ha_ndbcluster.cc
+SRC: src/ndbapi/NdbBlob*
+P: Pekka
+M: pekka@mysql.com
+S: Supported
+
+cpcd/cpcc
+SRC: src/cw/cpcd
+SRC: src/cw/cpcc
+C: Maintained only as part of autotest
+P: Jonas Orland
+M: jonas@mysql.com
+S: Maintained
+
+cpcc-win32
+SRC: src/cw/cpcc-win32
+S: Obsolete
+
+Handler
+SRC: ha_ndbcluster.cc
+P: Martin Skold
+M: martin@mysql.com
+S: Supported
+
+Management Server
+SRC: src/mgmsrv/
+P: Stewart Smith
+M: stewart@mysql.com
+S: Supported
+
+Management Client
+SRC: src/mgmclient/
+P: Stewart Smith
+M: stewart@mysql.com
+S: Supported
+
+Management API
+SRC: src/mgmapi/
+P: Stewart Smith
+M: stewart@mysql.com
+S: Supported
+
+NDB API Examples
+SRC: ndbapi-examples/
+P: Tomas Ulin
+M: tomas@mysql.com
+C: Originally by Lars
+P: Lars Thalmann
+M: lars@mysql.com
+S: Maintained
+
+NDB API NdbRecord Examples
+SRC: ndbapi-examples/
+P: Kristian Nielsen
+M: knielsen@mysql.com
+S: Maintained
+
+tsman
+C: Disk Data (Table Space MANager)
+SRC: src/kernel/blocks/tsman.cpp
+SRC: src/kernel/blocks/tsman.hpp
+P: Jonas Oreland
+M: jonas@mysql.com
+S: Supported
+
+lgman
+C: Disk Data (LoG MANager)
+SRC: src/kernel/blocks/lgman.cpp
+SRC: src/kernel/blocks/lgman.hpp
+P: Jonas Oreland
+M: jonas@mysql.com
+S: Supported
+
+pgman
+C: Disk Data (PaGe MANager)
+SRC: src/kernel/blocks/lgman.cpp
+SRC: src/kernel/blocks/lgman.hpp
+P: Jonas Oreland
+M: jonas@mysql.com
+S: Supported
+
+SUMA
+C: SUbscription MAnager
+C: Used for replication
+SRC: src/kernel/blocks/suma/
+P: Tomas Ulin
+P: tomas@mysql.com
+P: Jonas Oreland
+P: jonas@mysql.com
+S: Supported
+
+TRIX
+C: TRiggers and IndeXs (but only online Index build)
+SRC: src/kernel/blocks/trix
+P: Martin Skold
+P: mskold@mysql.com
+S: Supported
+
+QMGR
+C: Cluster (with a Q) ManaGeR
+C: Heartbeats etc
+SRC: src/kernel/blocks/qmgr
+S: Supported
+
+NDBFS
+C: NDB FileSystem
+C: File System abstraction
+SRC: src/kernel/blocks/ndbfs
+S: Supported
+
+TRIX
+C: TRiggers and IndeXs (but only online Index build)
+SRC: src/kernel/blocks/trix
+S: Supported
+
+TRIX
+C: TRiggers and IndeXs (but only online Index build)
+SRC: src/kernel/blocks/trix
+S: Supported
+

storage/ndb/include/Makefile.am

@@ -45,6 +45,7 @@ ndbapi/ndberror.h
 
 mgmapiinclude_HEADERS = \
 mgmapi/mgmapi.h \
+mgmapi/mgmapi_error.h \
 mgmapi/mgmapi_debug.h \
 mgmapi/mgmapi_config_parameters.h \
 mgmapi/mgmapi_config_parameters_debug.h \

storage/ndb/include/debugger/EventLogger.hpp

@@ -173,5 +173,5 @@ private:
   STATIC_CONST(MAX_TEXT_LENGTH = 256);
 };
 
-
+extern void getRestartAction(Uint32 action, BaseString &str);
 #endif

storage/ndb/include/mgmapi/mgmapi.h

@@ -18,6 +18,7 @@
 
 #include "mgmapi_config_parameters.h"
 #include "ndb_logevent.h"
+#include "mgmapi_error.h"
 
 #define MGM_LOGLEVELS CFG_MAX_LOGLEVEL - CFG_MIN_LOGLEVEL + 1
 #define NDB_MGM_MAX_LOGLEVEL 15
@@ -211,105 +212,6 @@ extern "C" {
 #endif
   };
 
-  /**
-   *    Error codes
-   */
-  enum ndb_mgm_error {
-    /** Not an error */
-    NDB_MGM_NO_ERROR = 0,
-
-    /* Request for service errors */
-    /** Supplied connectstring is illegal */
-    NDB_MGM_ILLEGAL_CONNECT_STRING = 1001,
-    /** Supplied NdbMgmHandle is illegal */
-    NDB_MGM_ILLEGAL_SERVER_HANDLE = 1005,
-    /** Illegal reply from server */
-    NDB_MGM_ILLEGAL_SERVER_REPLY = 1006,
-    /** Illegal number of nodes */
-    NDB_MGM_ILLEGAL_NUMBER_OF_NODES = 1007,
-    /** Illegal node status */
-    NDB_MGM_ILLEGAL_NODE_STATUS = 1008,
-    /** Memory allocation error */
-    NDB_MGM_OUT_OF_MEMORY = 1009,
-    /** Management server not connected */
-    NDB_MGM_SERVER_NOT_CONNECTED = 1010,
-    /** Could not connect to socker */
-    NDB_MGM_COULD_NOT_CONNECT_TO_SOCKET = 1011,
-    /** Could not bind local address */
-    NDB_MGM_BIND_ADDRESS = 1012,
-    
-    /* Alloc node id failures */
-    /** Generic error, retry may succeed */
-    NDB_MGM_ALLOCID_ERROR = 1101,
-    /** Non retriable error */
-    NDB_MGM_ALLOCID_CONFIG_MISMATCH = 1102,
-
-    /* Service errors - Start/Stop Node or System */
-    /** Start failed */
-    NDB_MGM_START_FAILED = 2001,
-    /** Stop failed */
-    NDB_MGM_STOP_FAILED = 2002,
-    /** Restart failed */
-    NDB_MGM_RESTART_FAILED = 2003,
-
-    /* Service errors - Backup */
-    /** Unable to start backup */
-    NDB_MGM_COULD_NOT_START_BACKUP = 3001,
-    /** Unable to abort backup */
-    NDB_MGM_COULD_NOT_ABORT_BACKUP = 3002,
-
-    /* Service errors - Single User Mode */
-    /** Unable to enter single user mode */
-    NDB_MGM_COULD_NOT_ENTER_SINGLE_USER_MODE = 4001,
-    /** Unable to exit single user mode */
-    NDB_MGM_COULD_NOT_EXIT_SINGLE_USER_MODE = 4002,
-
-    /* Usage errors */
-    /** Usage error */
-    NDB_MGM_USAGE_ERROR = 5001
-  };
-
-#ifndef DOXYGEN_SHOULD_SKIP_INTERNAL
-  struct Ndb_Mgm_Error_Msg {
-    enum ndb_mgm_error  code;
-    const char *        msg;
-  };
-  const struct Ndb_Mgm_Error_Msg ndb_mgm_error_msgs[] = {
-    { NDB_MGM_NO_ERROR, "No error" },
-
-    /* Request for service errors */
-    { NDB_MGM_ILLEGAL_CONNECT_STRING, "Illegal connect string" },
-    { NDB_MGM_ILLEGAL_SERVER_HANDLE, "Illegal server handle" },
-    { NDB_MGM_ILLEGAL_SERVER_REPLY, "Illegal reply from server" },
-    { NDB_MGM_ILLEGAL_NUMBER_OF_NODES, "Illegal number of nodes" },
-    { NDB_MGM_ILLEGAL_NODE_STATUS, "Illegal node status" },
-    { NDB_MGM_OUT_OF_MEMORY, "Out of memory" },
-    { NDB_MGM_SERVER_NOT_CONNECTED, "Management server not connected" },
-    { NDB_MGM_COULD_NOT_CONNECT_TO_SOCKET, "Could not connect to socket" },
-
-    /* Service errors - Start/Stop Node or System */
-    { NDB_MGM_START_FAILED, "Start failed" },
-    { NDB_MGM_STOP_FAILED, "Stop failed" },
-    { NDB_MGM_RESTART_FAILED, "Restart failed" },
-
-    /* Service errors - Backup */
-    { NDB_MGM_COULD_NOT_START_BACKUP, "Could not start backup" },
-    { NDB_MGM_COULD_NOT_ABORT_BACKUP, "Could not abort backup" },
-
-    /* Service errors - Single User Mode */
-    { NDB_MGM_COULD_NOT_ENTER_SINGLE_USER_MODE,
-      "Could not enter single user mode" },
-    { NDB_MGM_COULD_NOT_EXIT_SINGLE_USER_MODE,
-      "Could not exit single user mode" },
-
-    /* Usage errors */
-    { NDB_MGM_USAGE_ERROR,
-      "Usage error" }
-  };
-  const int ndb_mgm_noOfErrorMsgs =
-  sizeof(ndb_mgm_error_msgs)/sizeof(struct Ndb_Mgm_Error_Msg);
-#endif
-
   /**
    *   Status of a node in the cluster.
    *

storage/ndb/include/mgmapi/mgmapi_config_parameters.h

@@ -64,6 +64,7 @@
 
 #define CFG_DB_FILESYSTEM_PATH        125
 #define CFG_DB_NO_REDOLOG_FILES       126
+#define CFG_DB_REDOLOG_FILE_SIZE      140
 
 #define CFG_DB_LCP_DISC_PAGES_TUP     127
 #define CFG_DB_LCP_DISC_PAGES_TUP_SR  128
@@ -81,6 +82,8 @@
 #define CFG_DB_BACKUP_WRITE_SIZE          136
 #define CFG_DB_BACKUP_MAX_WRITE_SIZE      139
 
+#define CFG_DB_WATCHDOG_INTERVAL_INITIAL  141
+
 #define CFG_LOG_DESTINATION           147
 
 #define CFG_DB_DISCLESS               148
@@ -113,6 +116,10 @@
 
 #define CFG_DB_MEMREPORT_FREQUENCY    166
 
+#define CFG_DB_O_DIRECT               168
+
+#define CFG_DB_MAX_ALLOCATE           169
+
 #define CFG_DB_SGA                    198 /* super pool mem */
 #define CFG_DB_DATA_MEM_2             199 /* used in special build in 5.1 */
 

storage/ndb/include/mgmapi/mgmapi_error.h

@@ -0,0 +1,121 @@
+/* Copyright (C) 2003 MySQL AB
+
+   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
+   the Free Software Foundation; version 2 of the License.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
+
+#ifndef MGMAPI_ERROR_H
+#define MGMAPI_ERROR_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+  /**
+   *    Error codes
+   */
+  enum ndb_mgm_error {
+    /** Not an error */
+    NDB_MGM_NO_ERROR = 0,
+
+    /* Request for service errors */
+    /** Supplied connectstring is illegal */
+    NDB_MGM_ILLEGAL_CONNECT_STRING = 1001,
+    /** Supplied NdbMgmHandle is illegal */
+    NDB_MGM_ILLEGAL_SERVER_HANDLE = 1005,
+    /** Illegal reply from server */
+    NDB_MGM_ILLEGAL_SERVER_REPLY = 1006,
+    /** Illegal number of nodes */
+    NDB_MGM_ILLEGAL_NUMBER_OF_NODES = 1007,
+    /** Illegal node status */
+    NDB_MGM_ILLEGAL_NODE_STATUS = 1008,
+    /** Memory allocation error */
+    NDB_MGM_OUT_OF_MEMORY = 1009,
+    /** Management server not connected */
+    NDB_MGM_SERVER_NOT_CONNECTED = 1010,
+    /** Could not connect to socker */
+    NDB_MGM_COULD_NOT_CONNECT_TO_SOCKET = 1011,
+    /** Could not bind local address */
+    NDB_MGM_BIND_ADDRESS = 1012,
+    
+    /* Alloc node id failures */
+    /** Generic error, retry may succeed */
+    NDB_MGM_ALLOCID_ERROR = 1101,
+    /** Non retriable error */
+    NDB_MGM_ALLOCID_CONFIG_MISMATCH = 1102,
+
+    /* Service errors - Start/Stop Node or System */
+    /** Start failed */
+    NDB_MGM_START_FAILED = 2001,
+    /** Stop failed */
+    NDB_MGM_STOP_FAILED = 2002,
+    /** Restart failed */
+    NDB_MGM_RESTART_FAILED = 2003,
+
+    /* Service errors - Backup */
+    /** Unable to start backup */
+    NDB_MGM_COULD_NOT_START_BACKUP = 3001,
+    /** Unable to abort backup */
+    NDB_MGM_COULD_NOT_ABORT_BACKUP = 3002,
+
+    /* Service errors - Single User Mode */
+    /** Unable to enter single user mode */
+    NDB_MGM_COULD_NOT_ENTER_SINGLE_USER_MODE = 4001,
+    /** Unable to exit single user mode */
+    NDB_MGM_COULD_NOT_EXIT_SINGLE_USER_MODE = 4002,
+
+    /* Usage errors */
+    /** Usage error */
+    NDB_MGM_USAGE_ERROR = 5001
+  };
+  struct Ndb_Mgm_Error_Msg {
+    enum ndb_mgm_error  code;
+    const char *        msg;
+  };
+  const struct Ndb_Mgm_Error_Msg ndb_mgm_error_msgs[] = {
+    { NDB_MGM_NO_ERROR, "No error" },
+
+    /* Request for service errors */
+    { NDB_MGM_ILLEGAL_CONNECT_STRING, "Illegal connect string" },
+    { NDB_MGM_ILLEGAL_SERVER_HANDLE, "Illegal server handle" },
+    { NDB_MGM_ILLEGAL_SERVER_REPLY, "Illegal reply from server" },
+    { NDB_MGM_ILLEGAL_NUMBER_OF_NODES, "Illegal number of nodes" },
+    { NDB_MGM_ILLEGAL_NODE_STATUS, "Illegal node status" },
+    { NDB_MGM_OUT_OF_MEMORY, "Out of memory" },
+    { NDB_MGM_SERVER_NOT_CONNECTED, "Management server not connected" },
+    { NDB_MGM_COULD_NOT_CONNECT_TO_SOCKET, "Could not connect to socket" },
+
+    /* Service errors - Start/Stop Node or System */
+    { NDB_MGM_START_FAILED, "Start failed" },
+    { NDB_MGM_STOP_FAILED, "Stop failed" },
+    { NDB_MGM_RESTART_FAILED, "Restart failed" },
+
+    /* Service errors - Backup */
+    { NDB_MGM_COULD_NOT_START_BACKUP, "Could not start backup" },
+    { NDB_MGM_COULD_NOT_ABORT_BACKUP, "Could not abort backup" },
+
+    /* Service errors - Single User Mode */
+    { NDB_MGM_COULD_NOT_ENTER_SINGLE_USER_MODE,
+      "Could not enter single user mode" },
+    { NDB_MGM_COULD_NOT_EXIT_SINGLE_USER_MODE,
+      "Could not exit single user mode" },
+
+    /* Usage errors */
+    { NDB_MGM_USAGE_ERROR,
+      "Usage error" }
+  };
+  const int ndb_mgm_noOfErrorMsgs =
+  sizeof(ndb_mgm_error_msgs)/sizeof(struct Ndb_Mgm_Error_Msg);
+#ifdef __cplusplus
+}
+#endif
+
+#endif

storage/ndb/include/ndb_global.h.in

@@ -144,4 +144,6 @@ extern "C" {
 #define MAX(x,y) (((x)>(y))?(x):(y))
 #endif
 
+#define NDB_O_DIRECT_WRITE_ALIGNMENT 512
+
 #endif

storage/ndb/include/ndbapi/Ndb.hpp

@@ -1055,6 +1055,7 @@ class Ndb
   friend class NdbDictInterface;
   friend class NdbBlob;
   friend class NdbImpl;
+  friend class Ndb_internal;
 #endif
 
 public:

storage/ndb/include/ndbapi/NdbOperation.hpp

@@ -1042,6 +1042,13 @@ protected:
    */
   Int8 m_abortOption;
 
+  /*
+   * For blob impl, option to not propagate error to trans level.
+   * Could be AO_IgnoreError variant if we want it public.
+   * Ignored unless AO_IgnoreError is also set.
+   */
+  Int8 m_noErrorPropagation;
+
   friend struct Ndb_free_list_t<NdbOperation>;
 };
 

storage/ndb/include/portlib/NdbTick.h

@@ -37,9 +37,6 @@ NDB_TICKS NdbTick_CurrentMillisecond(void);
  */
 int NdbTick_CurrentMicrosecond(NDB_TICKS * secs, Uint32 * micros);
 
-  /*#define TIME_MEASUREMENT*/
-#ifdef TIME_MEASUREMENT
-
 struct MicroSecondTimer {
   NDB_TICKS seconds;
   NDB_TICKS micro_seconds;
@@ -54,7 +51,6 @@ struct MicroSecondTimer {
 NDB_TICKS NdbTick_getMicrosPassed(struct MicroSecondTimer start,
                             struct MicroSecondTimer stop);
 int NdbTick_getMicroTimer(struct MicroSecondTimer* time_now);
-#endif
 
 #ifdef	__cplusplus
 }

storage/ndb/ndbapi-examples/mgmapi_logevent/Makefile

@@ -1,6 +1,6 @@
 TARGET = mgmapi_logevent
-SRCS = $(TARGET).cpp
+SRCS = main.cpp
-OBJS = $(TARGET).o
+OBJS = main.o
 CXX = g++
 CFLAGS = -c -Wall -fno-rtti -fno-exceptions
 CXXFLAGS = 
@@ -17,7 +17,7 @@ SYS_LIB =
 $(TARGET): $(OBJS)
 	$(CXX) $(CXXFLAGS) $(LFLAGS) $(LIB_DIR) $(OBJS) -lndbclient -lmysqlclient_r -lmysys -lmystrings -lz $(SYS_LIB) -o $(TARGET)
 
-$(TARGET).o: $(SRCS)
+$(OBJS): $(SRCS)
 	$(CXX) $(CFLAGS) -I$(TOP_SRCDIR)/include -I$(INCLUDE_DIR) -I$(INCLUDE_DIR)/mgmapi  -I$(INCLUDE_DIR)/ndbapi $(SRCS)
 
 clean:

storage/ndb/ndbapi-examples/mgmapi_logevent2/Makefile

@@ -1,6 +1,6 @@
 TARGET = mgmapi_logevent2
-SRCS = $(TARGET).cpp
+SRCS = main.cpp
-OBJS = $(TARGET).o
+OBJS = main.o
 CXX = g++
 CFLAGS = -c -Wall -fno-rtti -fno-exceptions
 CXXFLAGS = 
@@ -17,7 +17,7 @@ SYS_LIB =
 $(TARGET): $(OBJS)
 	$(CXX) $(CXXFLAGS) $(LFLAGS) $(LIB_DIR) $(OBJS) -lndbclient -lmysqlclient_r -lmysys -lmystrings -lz $(SYS_LIB) -o $(TARGET)
 
-$(TARGET).o: $(SRCS)
+$(OBJS): $(SRCS)
 	$(CXX) $(CFLAGS) -I$(TOP_SRCDIR)/include -I$(INCLUDE_DIR) -I$(INCLUDE_DIR)/mgmapi  -I$(INCLUDE_DIR)/ndbapi $(SRCS)
 
 clean:

storage/ndb/ndbapi-examples/ndbapi_simple_dual/Makefile

@@ -1,6 +1,6 @@
 TARGET = ndbapi_simple_dual
-SRCS = $(TARGET).cpp
+SRCS = main.cpp
-OBJS = $(TARGET).o
+OBJS = main.o
 CXX = g++
 CFLAGS = -c -Wall -fno-rtti -fno-exceptions
 CXXFLAGS = 
@@ -17,7 +17,7 @@ SYS_LIB =
 $(TARGET): $(OBJS)
 	$(CXX) $(CXXFLAGS) $(LFLAGS) $(LIB_DIR) $(OBJS) -lndbclient -lmysqlclient_r -lmysys -lmystrings  -lz $(SYS_LIB) -o $(TARGET)
 
-$(TARGET).o: $(SRCS)
+$(OBJS): $(SRCS)
 	$(CXX) $(CFLAGS) -I$(TOP_SRCDIR)/include -I$(INCLUDE_DIR) -I$(INCLUDE_DIR)/ndbapi $(SRCS)
 
 clean:

storage/ndb/ndbapi-examples/ndbapi_simple_index/Makefile

@@ -1,6 +1,6 @@
 TARGET = ndbapi_simple_index
-SRCS = $(TARGET).cpp
+SRCS = main.cpp
-OBJS = $(TARGET).o
+OBJS = main.o
 CXX = g++
 CFLAGS = -c -Wall -fno-rtti -fno-exceptions
 CXXFLAGS = 
@@ -17,7 +17,7 @@ SYS_LIB =
 $(TARGET): $(OBJS)
 	$(CXX) $(CXXFLAGS) $(LFLAGS) $(LIB_DIR) $(OBJS) -lndbclient -lmysqlclient_r -lmysys -lmystrings -lz $(SYS_LIB) -o $(TARGET)
 
-$(TARGET).o: $(SRCS)
+$(OBJS): $(SRCS)
 	$(CXX) $(CFLAGS) -I$(INCLUDE_DIR)/include -I$(INCLUDE_DIR)/storage/ndb/include -I$(INCLUDE_DIR)/storage/ndb/include/ndbapi $(SRCS)
 
 clean:

storage/ndb/src/common/portlib/NdbTick.c

@@ -15,7 +15,7 @@
 
 
 #include <ndb_global.h>
-#include "NdbTick.h"
+#include <NdbTick.h>
 
 #define NANOSEC_PER_SEC  1000000000
 #define MICROSEC_PER_SEC 1000000
@@ -71,7 +71,6 @@ NdbTick_CurrentMicrosecond(NDB_TICKS * secs, Uint32 * micros){
 }
 
 #endif
-#ifdef TIME_MEASUREMENT
 int
 NdbTick_getMicroTimer(struct MicroSecondTimer* input_timer)
 {
@@ -102,4 +101,3 @@ NdbTick_getMicrosPassed(struct MicroSecondTimer start,
   }
   return ret_value;
 }
-#endif

storage/ndb/src/common/transporter/TransporterRegistry.cpp

@@ -818,6 +818,7 @@ TransporterRegistry::performReceive()
         {
           Uint32 * ptr;
           Uint32 sz = t->getReceiveData(&ptr);
+          transporter_recv_from(callbackObj, nodeId);
           Uint32 szUsed = unpack(ptr, sz, nodeId, ioStates[nodeId]);
           t->updateReceiveDataPtr(szUsed);
 	}

storage/ndb/src/kernel/blocks/ERROR_codes.txt

@@ -1,5 +1,5 @@
 Next QMGR 1
-Next NDBCNTR 1001
+Next NDBCNTR 1002
 Next NDBFS 2000
 Next DBACC 3002
 Next DBTUP 4029
@@ -523,3 +523,4 @@ Dbtup:
 NDBCNTR:
 
 1000: Crash insertion on SystemError::CopyFragRef
+1001: Delay sending NODE_FAILREP (to own node), until error is cleared

storage/ndb/src/kernel/blocks/backup/Backup.cpp

@@ -448,6 +448,41 @@ Backup::execDUMP_STATE_ORD(Signal* signal)
 		  filePtr.p->m_flags);
       }
     }
+
+    ndbout_c("m_curr_disk_write_speed: %u  m_words_written_this_period: %u  m_overflow_disk_write: %u",
+              m_curr_disk_write_speed, m_words_written_this_period, m_overflow_disk_write);
+    ndbout_c("m_reset_delay_used: %u  m_reset_disk_speed_time: %llu",
+             m_reset_delay_used, (Uint64)m_reset_disk_speed_time);
+    for(c_backups.first(ptr); ptr.i != RNIL; c_backups.next(ptr))
+    {
+      ndbout_c("BackupRecord %u:  BackupId: %u  MasterRef: %x  ClientRef: %x",
+               ptr.i, ptr.p->backupId, ptr.p->masterRef, ptr.p->clientRef);
+      ndbout_c(" State: %u", ptr.p->slaveState.getState());
+      ndbout_c(" noOfByte: %llu  noOfRecords: %llu",
+               ptr.p->noOfBytes, ptr.p->noOfRecords);
+      ndbout_c(" noOfLogBytes: %llu  noOfLogRecords: %llu",
+               ptr.p->noOfLogBytes, ptr.p->noOfLogRecords);
+      ndbout_c(" errorCode: %u", ptr.p->errorCode);
+      BackupFilePtr filePtr;
+      for(ptr.p->files.first(filePtr); filePtr.i != RNIL; 
+	  ptr.p->files.next(filePtr))
+      {
+	ndbout_c(" file %u:  type: %u  flags: H'%x  tableId: %u  fragmentId: %u",
+                 filePtr.i, filePtr.p->fileType, filePtr.p->m_flags,
+                 filePtr.p->tableId, filePtr.p->fragmentNo);
+      }
+      if (ptr.p->slaveState.getState() == SCANNING && ptr.p->dataFilePtr != RNIL)
+      {
+        c_backupFilePool.getPtr(filePtr, ptr.p->dataFilePtr);
+        OperationRecord & op = filePtr.p->operation;
+        Uint32 *tmp = NULL;
+        Uint32 sz = 0;
+        bool eof = FALSE;
+        bool ready = op.dataBuffer.getReadPtr(&tmp, &sz, &eof);
+        ndbout_c("ready: %s  eof: %s", ready ? "TRUE" : "FALSE", eof ? "TRUE" : "FALSE");
+      }
+    }
+    return;
   }
   if(signal->theData[0] == 24){
     /**
@@ -2771,6 +2806,8 @@ Backup::openFiles(Signal* signal, BackupRecordPtr ptr)
   c_backupFilePool.getPtr(filePtr, ptr.p->dataFilePtr);
   filePtr.p->m_flags |= BackupFile::BF_OPENING;
 
+  if (c_defaults.m_o_direct)
+    req->fileFlags |= FsOpenReq::OM_DIRECT;
   req->userPointer = filePtr.i;
   FsOpenReq::setVersion(req->fileNumber, 2);
   FsOpenReq::setSuffix(req->fileNumber, FsOpenReq::S_DATA);
@@ -3745,12 +3782,31 @@ Backup::OperationRecord::newFragment(Uint32 tableId, Uint32 fragNo)
 }
 
 bool
-Backup::OperationRecord::fragComplete(Uint32 tableId, Uint32 fragNo)
+Backup::OperationRecord::fragComplete(Uint32 tableId, Uint32 fragNo, bool fill_record)
 {
   Uint32 * tmp;
   const Uint32 footSz = sizeof(BackupFormat::DataFile::FragmentFooter) >> 2;
+  Uint32 sz = footSz + 1;
 
-  if(dataBuffer.getWritePtr(&tmp, footSz + 1)) {
+  if (fill_record)
+  {
+    Uint32 * new_tmp;
+    if (!dataBuffer.getWritePtr(&tmp, sz))
+      return false;
+    new_tmp = tmp + sz;
+
+    if ((UintPtr)new_tmp & (sizeof(Page32)-1))
+    {
+      /* padding is needed to get full write */
+      new_tmp += 2 /* to fit empty header minimum 2 words*/;
+      new_tmp = (Uint32 *)(((UintPtr)new_tmp + sizeof(Page32)-1) &
+                            ~(UintPtr)(sizeof(Page32)-1));
+      /* new write sz */
+      sz = new_tmp - tmp;
+    }
+  }
+
+  if(dataBuffer.getWritePtr(&tmp, sz)) {
     jam();
     * tmp = 0; // Finish record stream
     tmp++;
@@ -3762,7 +3818,17 @@ Backup::OperationRecord::fragComplete(Uint32 tableId, Uint32 fragNo)
     foot->FragmentNo    = htonl(fragNo);
     foot->NoOfRecords   = htonl(noOfRecords);
     foot->Checksum      = htonl(0);
-    dataBuffer.updateWritePtr(footSz + 1);
+
+    if (sz != footSz + 1)
+    {
+      tmp += footSz;
+      memset(tmp, 0, (sz - footSz - 1) * 4);
+      *tmp = htonl(BackupFormat::EMPTY_ENTRY);
+      tmp++;
+      *tmp = htonl(sz - footSz - 1);
+    }
+
+    dataBuffer.updateWritePtr(sz);
     return true;
   }//if
   return false;
@@ -3864,8 +3930,13 @@ Backup::fragmentCompleted(Signal* signal, BackupFilePtr filePtr)
     return;
   }//if
     
+  BackupRecordPtr ptr LINT_SET_PTR;
+  c_backupPool.getPtr(ptr, filePtr.p->backupPtr);
+
   OperationRecord & op = filePtr.p->operation;
-  if(!op.fragComplete(filePtr.p->tableId, filePtr.p->fragmentNo)) {
+  if(!op.fragComplete(filePtr.p->tableId, filePtr.p->fragmentNo,
+                      c_defaults.m_o_direct))
+  {
     jam();
     signal->theData[0] = BackupContinueB::BUFFER_FULL_FRAG_COMPLETE;
     signal->theData[1] = filePtr.i;
@@ -3875,9 +3946,6 @@ Backup::fragmentCompleted(Signal* signal, BackupFilePtr filePtr)
   
   filePtr.p->m_flags &= ~(Uint32)BackupFile::BF_SCAN_THREAD;
   
-  BackupRecordPtr ptr LINT_SET_PTR;
-  c_backupPool.getPtr(ptr, filePtr.p->backupPtr);
-
   if (ptr.p->is_lcp())
   {
     ptr.p->slaveState.setState(STOPPING);
@@ -4914,6 +4982,8 @@ Backup::lcp_open_file(Signal* signal, BackupRecordPtr ptr)
     FsOpenReq::OM_CREATE | 
     FsOpenReq::OM_APPEND |
     FsOpenReq::OM_AUTOSYNC;
+  if (c_defaults.m_o_direct)
+    req->fileFlags |= FsOpenReq::OM_DIRECT;
   FsOpenReq::v2_setCount(req->fileNumber, 0xFFFFFFFF);
   req->auto_sync_size = c_defaults.m_disk_synch_size;
   

storage/ndb/src/kernel/blocks/backup/Backup.hpp

@@ -240,7 +240,7 @@ public:
      * Once per fragment
      */
     bool newFragment(Uint32 tableId, Uint32 fragNo);
-    bool fragComplete(Uint32 tableId, Uint32 fragNo);
+    bool fragComplete(Uint32 tableId, Uint32 fragNo, bool fill_record);
     
     /**
      * Once per scan frag (next) req/conf
@@ -534,6 +534,7 @@ public:
     Uint32 m_disk_write_speed;
     Uint32 m_disk_synch_size;
     Uint32 m_diskless;
+    Uint32 m_o_direct;
   };
   
   /**

storage/ndb/src/kernel/blocks/backup/BackupFormat.hpp

@@ -32,7 +32,8 @@ struct BackupFormat {
     TABLE_LIST        = 4,
     TABLE_DESCRIPTION = 5,
     GCP_ENTRY         = 6,
-    FRAGMENT_INFO     = 7
+    FRAGMENT_INFO     = 7,
+    EMPTY_ENTRY       = 8
   };
 
   struct FileHeader {
@@ -93,6 +94,13 @@ struct BackupFormat {
       Uint32 NoOfRecords;
       Uint32 Checksum;
     };
+
+    /* optional padding for O_DIRECT */
+    struct EmptyEntry {
+      Uint32 SectionType;
+      Uint32 SectionLength;
+      /* not used data */
+    };
   };
 
   /**

storage/ndb/src/kernel/blocks/backup/BackupInit.cpp

@@ -148,10 +148,13 @@ Backup::execREAD_CONFIG_REQ(Signal* signal)
   c_defaults.m_disk_write_speed = 10 * (1024 * 1024);
   c_defaults.m_disk_write_speed_sr = 100 * (1024 * 1024);
   c_defaults.m_disk_synch_size = 4 * (1024 * 1024);
+  c_defaults.m_o_direct = true;
 
   Uint32 noBackups = 0, noTables = 0, noAttribs = 0, noFrags = 0;
   ndbrequire(!ndb_mgm_get_int_parameter(p, CFG_DB_DISCLESS, 
 					&c_defaults.m_diskless));
+  ndb_mgm_get_int_parameter(p, CFG_DB_O_DIRECT,
+                            &c_defaults.m_o_direct);
   ndb_mgm_get_int_parameter(p, CFG_DB_CHECKPOINT_SPEED_SR,
 			    &c_defaults.m_disk_write_speed_sr);
   ndb_mgm_get_int_parameter(p, CFG_DB_CHECKPOINT_SPEED,
@@ -204,7 +207,7 @@ Backup::execREAD_CONFIG_REQ(Signal* signal)
     / sizeof(Page32);
   // We need to allocate an additional of 2 pages. 1 page because of a bug in
   // ArrayPool and another one for DICTTAINFO.
-  c_pagePool.setSize(noPages + NO_OF_PAGES_META_FILE + 2); 
+  c_pagePool.setSize(noPages + NO_OF_PAGES_META_FILE + 2, true); 
   
   { // Init all tables
     SLList<Table> tables(c_tablePool);

storage/ndb/src/kernel/blocks/backup/FsBuffer.hpp

@@ -270,8 +270,8 @@ FsBuffer::getReadPtr(Uint32 ** ptr, Uint32 * sz, bool * _eof){
     
     * ptr = &Tp[Tr];
 
-    DEBUG(ndbout_c("getReadPtr() Tr: %d Tw: %d Ts: %d Tm: %d sz1: %d -> %d",
+    DEBUG(ndbout_c("getReadPtr() Tr: %d Tmw: %d Ts: %d Tm: %d sz1: %d -> %d",
-		   Tr, Tw, Ts, Tm, sz1, * sz));
+		   Tr, Tmw, Ts, Tm, sz1, * sz));
 
     return true;
   }
@@ -279,8 +279,8 @@ FsBuffer::getReadPtr(Uint32 ** ptr, Uint32 * sz, bool * _eof){
   if(!m_eof){
     * _eof = false;
     
-    DEBUG(ndbout_c("getReadPtr() Tr: %d Tw: %d Ts: %d Tm: %d sz1: %d -> false",
+    DEBUG(ndbout_c("getReadPtr() Tr: %d Tmw: %d Ts: %d Tm: %d sz1: %d -> false",
-		   Tr, Tw, Ts, Tm, sz1));
+		   Tr, Tmw, Ts, Tm, sz1));
     
     return false;
   }
@@ -289,8 +289,8 @@ FsBuffer::getReadPtr(Uint32 ** ptr, Uint32 * sz, bool * _eof){
   * _eof = true;
   * ptr = &Tp[Tr];
 
-  DEBUG(ndbout_c("getReadPtr() Tr: %d Tw: %d Ts: %d Tm: %d sz1: %d -> %d eof",
+  DEBUG(ndbout_c("getReadPtr() Tr: %d Tmw: %d Ts: %d Tm: %d sz1: %d -> %d eof",
-		 Tr, Tw, Ts, Tm, sz1, * sz));
+		 Tr, Tmw, Ts, Tm, sz1, * sz));
   
   return false;
 }
@@ -316,13 +316,13 @@ FsBuffer::getWritePtr(Uint32 ** ptr, Uint32 sz){
   if(sz1 > sz){ // Note at least 1 word of slack
     * ptr = &Tp[Tw];
 
-    DEBUG(ndbout_c("getWritePtr(%d) Tr: %d Tw: %d Ts: %d sz1: %d -> true",
+    DEBUG(ndbout_c("getWritePtr(%d) Tw: %d sz1: %d -> true",
-		   sz, Tr, Tw, Ts, sz1));
+		   sz, Tw, sz1));
     return true;
   }
 
-  DEBUG(ndbout_c("getWritePtr(%d) Tr: %d Tw: %d Ts: %d sz1: %d -> false",
+  DEBUG(ndbout_c("getWritePtr(%d) Tw: %d sz1: %d -> false",
-		 sz, Tr, Tw, Ts, sz1));
+		 sz, Tw, sz1));
 
   return false;
 }
@@ -339,11 +339,15 @@ FsBuffer::updateWritePtr(Uint32 sz){
   m_free -= sz;
   if(Tnew < Ts){
     m_writeIndex = Tnew;
+    DEBUG(ndbout_c("updateWritePtr(%d) m_writeIndex: %d",
+                   sz, m_writeIndex));
     return;
   }
 
   memcpy(Tp, &Tp[Ts], (Tnew - Ts) << 2);
   m_writeIndex = Tnew - Ts;
+  DEBUG(ndbout_c("updateWritePtr(%d) m_writeIndex: %d",
+                 sz, m_writeIndex));
 }
 
 inline

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

@@ -698,6 +698,9 @@ void Dbdict::execFSCLOSECONF(Signal* signal)
   case FsConnectRecord::OPEN_READ_SCHEMA2:
     openSchemaFile(signal, 1, fsPtr.i, false, false);
     break;
+  case FsConnectRecord::OPEN_READ_TAB_FILE2:
+    openTableFile(signal, 1, fsPtr.i, c_readTableRecord.tableId, false);
+    break;
   default:
     jamLine((fsPtr.p->fsState & 0xFFF));
     ndbrequire(false);
@@ -1073,8 +1076,11 @@ void Dbdict::readTableConf(Signal* signal,
 void Dbdict::readTableRef(Signal* signal,
                           FsConnectRecordPtr fsPtr)
 {
+  /**
+   * First close corrupt file
+   */
   fsPtr.p->fsState = FsConnectRecord::OPEN_READ_TAB_FILE2;
-  openTableFile(signal, 1, fsPtr.i, c_readTableRecord.tableId, false);
+  closeFile(signal, fsPtr.p->filePtr, fsPtr.i);
   return;
 }//Dbdict::readTableRef()
 

storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp

@@ -4741,12 +4741,18 @@ void Dbdih::failedNodeLcpHandling(Signal* signal, NodeRecordPtr failedNodePtr)
   jam();
   const Uint32 nodeId = failedNodePtr.i;
 
-  if (c_lcpState.m_participatingLQH.get(failedNodePtr.i)){
+  if (isMaster() && c_lcpState.m_participatingLQH.get(failedNodePtr.i))
+  {
     /*----------------------------------------------------*/
     /*  THE NODE WAS INVOLVED IN A LOCAL CHECKPOINT. WE   */
     /* MUST UPDATE THE ACTIVE STATUS TO INDICATE THAT     */
     /* THE NODE HAVE MISSED A LOCAL CHECKPOINT.           */
     /*----------------------------------------------------*/
+
+    /**
+     * Bug#28717, Only master should do this, as this status is copied
+     *   to other nodes
+     */
     switch (failedNodePtr.p->activeStatus) {
     case Sysfile::NS_Active:
       jam();

storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp

@@ -71,7 +71,6 @@ class Dbtup;
 /*       CONSTANTS OF THE LOG PAGES                                          */
 /* ------------------------------------------------------------------------- */
 #define ZPAGE_HEADER_SIZE 32
-#define ZNO_MBYTES_IN_FILE 16
 #define ZPAGE_SIZE 8192
 #define ZPAGES_IN_MBYTE 32
 #define ZTWOLOG_NO_PAGES_IN_MBYTE 5
@@ -115,9 +114,6 @@ class Dbtup;
 /* ------------------------------------------------------------------------- */
 /*       VARIOUS CONSTANTS USED AS FLAGS TO THE FILE MANAGER.                */
 /* ------------------------------------------------------------------------- */
-#define ZOPEN_READ 0
-#define ZOPEN_WRITE 1
-#define ZOPEN_READ_WRITE 2
 #define ZVAR_NO_LOG_PAGE_WORD 1
 #define ZLIST_OF_PAIRS 0
 #define ZLIST_OF_PAIRS_SYNCH 16
@@ -142,7 +138,7 @@ class Dbtup;
 /*       IN THE MBYTE.                                                       */
 /* ------------------------------------------------------------------------- */
 #define ZFD_HEADER_SIZE 3
-#define ZFD_PART_SIZE 48
+#define ZFD_MBYTE_SIZE 3
 #define ZLOG_HEAD_SIZE 8
 #define ZNEXT_LOG_SIZE 2
 #define ZABORT_LOG_SIZE 3
@@ -169,7 +165,6 @@ class Dbtup;
 #define ZPOS_LOG_TYPE 0
 #define ZPOS_NO_FD 1
 #define ZPOS_FILE_NO 2
-#define ZMAX_LOG_FILES_IN_PAGE_ZERO 40
 /* ------------------------------------------------------------------------- */
 /*       THE POSITIONS WITHIN A PREPARE LOG RECORD AND A NEW PREPARE         */
 /*       LOG RECORD.                                                         */
@@ -1437,17 +1432,17 @@ public:
      *       header of each log file.  That information is used during
      *       system restart to find the tail of the log.  
      */
-    UintR logLastPrepRef[16];
+    UintR *logLastPrepRef;
     /**
      *       The max global checkpoint completed before the mbyte in the
      *       log file was started.  One variable per mbyte.  
      */
-    UintR logMaxGciCompleted[16];
+    UintR *logMaxGciCompleted;
     /**
      *       The max global checkpoint started before the mbyte in the log
      *       file was started.  One variable per mbyte.
      */
-    UintR logMaxGciStarted[16];
+    UintR *logMaxGciStarted;
     /**
      *       This variable contains the file name as needed by the file 
      *       system when opening the file.
@@ -2163,6 +2158,7 @@ private:
   void execSTART_RECREF(Signal* signal);
 
   void execGCP_SAVEREQ(Signal* signal);
+  void execFSOPENREF(Signal* signal);
   void execFSOPENCONF(Signal* signal);
   void execFSCLOSECONF(Signal* signal);
   void execFSWRITECONF(Signal* signal);
@@ -2671,6 +2667,8 @@ private:
   LogPartRecord *logPartRecord;
   LogPartRecordPtr logPartPtr;
   UintR clogPartFileSize;
+  Uint32 clogFileSize; // In MBYTE
+  Uint32 cmaxLogFilesInPageZero; //
 
 // Configurable
   LogFileRecord *logFileRecord;
@@ -2686,6 +2684,7 @@ private:
   UintR clfoFileSize;
 
   LogPageRecord *logPageRecord;
+  void *logPageRecordUnaligned;
   LogPageRecordPtr logPagePtr;
   UintR cfirstfreeLogPage;
   UintR clogPageFileSize;
@@ -2889,6 +2888,7 @@ private:
   UintR ctransidHash[1024];
   
   Uint32 c_diskless;
+  Uint32 c_o_direct;
   Uint32 c_error_insert_table_id;
   
 public:

storage/ndb/src/kernel/blocks/dblqh/DblqhInit.cpp

@@ -49,6 +49,7 @@ void Dblqh::initData()
   logFileRecord = 0;
   logFileOperationRecord = 0;
   logPageRecord = 0;
+  logPageRecordUnaligned= 0;
   pageRefRecord = 0;
   tablerec = 0;
   tcConnectionrec = 0;
@@ -60,6 +61,8 @@ void Dblqh::initData()
   cLqhTimeOutCheckCount = 0;
   cbookedAccOps = 0;
   m_backup_ptr = RNIL;
+  clogFileSize = 16;
+  cmaxLogFilesInPageZero = 40;
 }//Dblqh::initData()
 
 void Dblqh::initRecords() 
@@ -105,10 +108,13 @@ void Dblqh::initRecords()
 		sizeof(LogFileOperationRecord), 
 		clfoFileSize);
 
-  logPageRecord = (LogPageRecord*)allocRecord("LogPageRecord",
+  logPageRecord =
-					      sizeof(LogPageRecord),
+    (LogPageRecord*)allocRecordAligned("LogPageRecord",
-					      clogPageFileSize,
+                                       sizeof(LogPageRecord),
-					      false);
+                                       clogPageFileSize,
+                                       &logPageRecordUnaligned,
+                                       NDB_O_DIRECT_WRITE_ALIGNMENT,
+                                       false);
 
   pageRefRecord = (PageRefRecord*)allocRecord("PageRefRecord",
 					      sizeof(PageRefRecord),
@@ -260,6 +266,7 @@ Dblqh::Dblqh(Block_context& ctx):
   addRecSignal(GSN_START_FRAGREQ, &Dblqh::execSTART_FRAGREQ);
   addRecSignal(GSN_START_RECREF, &Dblqh::execSTART_RECREF);
   addRecSignal(GSN_GCP_SAVEREQ, &Dblqh::execGCP_SAVEREQ);
+  addRecSignal(GSN_FSOPENREF, &Dblqh::execFSOPENREF, true);
   addRecSignal(GSN_FSOPENCONF, &Dblqh::execFSOPENCONF);
   addRecSignal(GSN_FSCLOSECONF, &Dblqh::execFSCLOSECONF);
   addRecSignal(GSN_FSWRITECONF, &Dblqh::execFSWRITECONF);
@@ -377,7 +384,7 @@ Dblqh::~Dblqh()
 		sizeof(LogFileOperationRecord), 
 		clfoFileSize);
   
-  deallocRecord((void**)&logPageRecord,
+  deallocRecord((void**)&logPageRecordUnaligned,
 		"LogPageRecord",
 		sizeof(LogPageRecord),
 		clogPageFileSize);

storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp

@@ -1036,11 +1036,41 @@ void Dblqh::execREAD_CONFIG_REQ(Signal* signal)
   cmaxAccOps = cscanrecFileSize * MAX_PARALLEL_OP_PER_SCAN;
 
   ndbrequire(!ndb_mgm_get_int_parameter(p, CFG_DB_DISCLESS, &c_diskless));
+  c_o_direct = true;
+  ndb_mgm_get_int_parameter(p, CFG_DB_O_DIRECT, &c_o_direct);
   
   Uint32 tmp= 0;
   ndbrequire(!ndb_mgm_get_int_parameter(p, CFG_LQH_FRAG, &tmp));
   c_fragment_pool.setSize(tmp);
 
+  if (!ndb_mgm_get_int_parameter(p, CFG_DB_REDOLOG_FILE_SIZE,
+                                 &clogFileSize))
+  {
+    // convert to mbyte
+    clogFileSize = (clogFileSize + 1024*1024 - 1) / (1024 * 1024);
+    ndbrequire(clogFileSize >= 4 && clogFileSize <= 1024);
+  }
+
+  cmaxLogFilesInPageZero = (ZPAGE_SIZE - ZPAGE_HEADER_SIZE - 128) /
+    (ZFD_MBYTE_SIZE * clogFileSize);
+
+  /**
+   * "Old" cmaxLogFilesInPageZero was 40
+   * Each FD need 3 words per mb, require that they can fit into 1 page
+   *   (atleast 1 FD)
+   * Is also checked in ConfigInfo.cpp (max FragmentLogFileSize = 1Gb)
+   *   1Gb = 1024Mb => 3(ZFD_MBYTE_SIZE) * 1024 < 8192 (ZPAGE_SIZE)
+   */
+  if (cmaxLogFilesInPageZero > 40)
+  {
+    jam();
+    cmaxLogFilesInPageZero = 40;
+  }
+  else
+  {
+    ndbrequire(cmaxLogFilesInPageZero);
+  }
+  
   initRecords();
   initialiseRecordsLab(signal, 0, ref, senderData);
 
@@ -11793,9 +11823,9 @@ void Dblqh::sendStartLcp(Signal* signal)
 Uint32 Dblqh::remainingLogSize(const LogFileRecordPtr &sltCurrLogFilePtr,
 			       const LogPartRecordPtr &sltLogPartPtr)
 {
-  Uint32 hf = sltCurrLogFilePtr.p->fileNo*ZNO_MBYTES_IN_FILE+sltCurrLogFilePtr.p->currentMbyte;
+  Uint32 hf = sltCurrLogFilePtr.p->fileNo*clogFileSize+sltCurrLogFilePtr.p->currentMbyte;
-  Uint32 tf = sltLogPartPtr.p->logTailFileNo*ZNO_MBYTES_IN_FILE+sltLogPartPtr.p->logTailMbyte;
+  Uint32 tf = sltLogPartPtr.p->logTailFileNo*clogFileSize+sltLogPartPtr.p->logTailMbyte;
-  Uint32 sz = sltLogPartPtr.p->noLogFiles*ZNO_MBYTES_IN_FILE;
+  Uint32 sz = sltLogPartPtr.p->noLogFiles*clogFileSize;
   if (tf > hf) hf += sz;
   return sz-(hf-tf);
 }
@@ -11853,7 +11883,7 @@ void Dblqh::setLogTail(Signal* signal, Uint32 keepGci)
 /* ------------------------------------------------------------------------- */
   SLT_LOOP:
     for (tsltIndex = tsltStartMbyte;
-	 tsltIndex <= ZNO_MBYTES_IN_FILE - 1; 
+	 tsltIndex <= clogFileSize - 1;
 	 tsltIndex++) {
       if (sltLogFilePtr.p->logMaxGciStarted[tsltIndex] >= keepGci) {
 /* ------------------------------------------------------------------------- */
@@ -11869,7 +11899,7 @@ void Dblqh::setLogTail(Signal* signal, Uint32 keepGci)
 /* ------------------------------------------------------------------------- */
 /*STEPPING BACK INCLUDES ALSO STEPPING BACK TO THE PREVIOUS LOG FILE.        */
 /* ------------------------------------------------------------------------- */
-          tsltMbyte = ZNO_MBYTES_IN_FILE - 1;
+          tsltMbyte = clogFileSize - 1;
           sltLogFilePtr.i = sltLogFilePtr.p->prevLogFile;
           ptrCheckGuard(sltLogFilePtr, clogFileFileSize, logFileRecord);
         }//if
@@ -11907,7 +11937,7 @@ void Dblqh::setLogTail(Signal* signal, Uint32 keepGci)
       UintR ToldTailFileNo = sltLogPartPtr.p->logTailFileNo;
       UintR ToldTailMByte = sltLogPartPtr.p->logTailMbyte;
 
-      arrGuard(tsltMbyte, 16);
+      arrGuard(tsltMbyte, clogFileSize);
       sltLogPartPtr.p->logTailFileNo = 
          sltLogFilePtr.p->logLastPrepRef[tsltMbyte] >> 16;
 /* ------------------------------------------------------------------------- */
@@ -12407,6 +12437,26 @@ void Dblqh::execFSOPENCONF(Signal* signal)
   }//switch
 }//Dblqh::execFSOPENCONF()
 
+void
+Dblqh::execFSOPENREF(Signal* signal)
+{
+  jamEntry();
+  FsRef* ref = (FsRef*)signal->getDataPtr();
+  Uint32 err = ref->errorCode;
+  if (err == FsRef::fsErrInvalidFileSize)
+  {
+    char buf[256];
+    BaseString::snprintf(buf, sizeof(buf),
+                         "Invalid file size for redo logfile, "
+                         " size only changable with --initial");
+    progError(__LINE__,
+              NDBD_EXIT_INVALID_CONFIG,
+              buf);
+    return;
+  }
+
+  SimulatedBlock::execFSOPENREF(signal);
+}
 
 /* ************>> */
 /*  FSREADCONF  > */
@@ -13052,7 +13102,7 @@ void Dblqh::openFileInitLab(Signal* signal)
 {
   logFilePtr.p->logFileStatus = LogFileRecord::OPEN_INIT;
   seizeLogpage(signal);
-  writeSinglePage(signal, (ZNO_MBYTES_IN_FILE * ZPAGES_IN_MBYTE) - 1,
+  writeSinglePage(signal, (clogFileSize * ZPAGES_IN_MBYTE) - 1,
                   ZPAGE_SIZE - 1, __LINE__);
   lfoPtr.p->lfoState = LogFileOperationRecord::INIT_WRITE_AT_END;
   return;
@@ -13115,7 +13165,7 @@ void Dblqh::writeInitMbyteLab(Signal* signal)
 {
   releaseLfo(signal);
   logFilePtr.p->currentMbyte = logFilePtr.p->currentMbyte + 1;
-  if (logFilePtr.p->currentMbyte == ZNO_MBYTES_IN_FILE) {
+  if (logFilePtr.p->currentMbyte == clogFileSize) {
     jam();
     releaseLogpage(signal);
     logFilePtr.p->logFileStatus = LogFileRecord::CLOSING_INIT;
@@ -13235,7 +13285,7 @@ void Dblqh::initLogfile(Signal* signal, Uint32 fileNo)
   logFilePtr.p->lastPageWritten = 0;
   logFilePtr.p->logPageZero = RNIL;
   logFilePtr.p->currentMbyte = 0;
-  for (tilIndex = 0; tilIndex <= 15; tilIndex++) {
+  for (tilIndex = 0; tilIndex < clogFileSize; tilIndex++) {
     logFilePtr.p->logMaxGciCompleted[tilIndex] = (UintR)-1;
     logFilePtr.p->logMaxGciStarted[tilIndex] = (UintR)-1;
     logFilePtr.p->logLastPrepRef[tilIndex] = 0;
@@ -13286,8 +13336,14 @@ void Dblqh::openFileRw(Signal* signal, LogFileRecordPtr olfLogFilePtr)
   signal->theData[3] = olfLogFilePtr.p->fileName[1];
   signal->theData[4] = olfLogFilePtr.p->fileName[2];
   signal->theData[5] = olfLogFilePtr.p->fileName[3];
-  signal->theData[6] = ZOPEN_READ_WRITE | FsOpenReq::OM_AUTOSYNC;
+  signal->theData[6] = FsOpenReq::OM_READWRITE | FsOpenReq::OM_AUTOSYNC | FsOpenReq::OM_CHECK_SIZE;
+  if (c_o_direct)
+    signal->theData[6] |= FsOpenReq::OM_DIRECT;
   req->auto_sync_size = MAX_REDO_PAGES_WITHOUT_SYNCH * sizeof(LogPageRecord);
+  Uint64 sz = clogFileSize;
+  sz *= 1024; sz *= 1024;
+  req->file_size_hi = sz >> 32;
+  req->file_size_lo = sz & 0xFFFFFFFF;
   sendSignal(NDBFS_REF, GSN_FSOPENREQ, signal, FsOpenReq::SignalLength, JBA);
 }//Dblqh::openFileRw()
 
@@ -13306,7 +13362,9 @@ void Dblqh::openLogfileInit(Signal* signal)
   signal->theData[3] = logFilePtr.p->fileName[1];
   signal->theData[4] = logFilePtr.p->fileName[2];
   signal->theData[5] = logFilePtr.p->fileName[3];
-  signal->theData[6] = 0x302 | FsOpenReq::OM_AUTOSYNC;
+  signal->theData[6] = FsOpenReq::OM_READWRITE | FsOpenReq::OM_TRUNCATE | FsOpenReq::OM_CREATE | FsOpenReq::OM_AUTOSYNC;
+  if (c_o_direct)
+    signal->theData[6] |= FsOpenReq::OM_DIRECT;
   req->auto_sync_size = MAX_REDO_PAGES_WITHOUT_SYNCH * sizeof(LogPageRecord);
   sendSignal(NDBFS_REF, GSN_FSOPENREQ, signal, FsOpenReq::SignalLength, JBA);
 }//Dblqh::openLogfileInit()
@@ -13342,8 +13400,14 @@ void Dblqh::openNextLogfile(Signal* signal)
     signal->theData[3] = onlLogFilePtr.p->fileName[1];
     signal->theData[4] = onlLogFilePtr.p->fileName[2];
     signal->theData[5] = onlLogFilePtr.p->fileName[3];
-    signal->theData[6] = 2 | FsOpenReq::OM_AUTOSYNC;
+    signal->theData[6] = FsOpenReq::OM_READWRITE | FsOpenReq::OM_AUTOSYNC | FsOpenReq::OM_CHECK_SIZE;
+    if (c_o_direct)
+      signal->theData[6] |= FsOpenReq::OM_DIRECT;
     req->auto_sync_size = MAX_REDO_PAGES_WITHOUT_SYNCH * sizeof(LogPageRecord);
+    Uint64 sz = clogFileSize;
+    sz *= 1024; sz *= 1024;
+    req->file_size_hi = sz >> 32;
+    req->file_size_lo = sz & 0xFFFFFFFF;
     sendSignal(NDBFS_REF, GSN_FSOPENREQ, signal, FsOpenReq::SignalLength, JBA);
   }//if
 }//Dblqh::openNextLogfile()
@@ -13474,7 +13538,7 @@ void Dblqh::writeFileDescriptor(Signal* signal)
 /* -------------------------------------------------- */
 /*       START BY WRITING TO LOG FILE RECORD          */
 /* -------------------------------------------------- */
-  arrGuard(logFilePtr.p->currentMbyte, 16);
+  arrGuard(logFilePtr.p->currentMbyte, clogFileSize);
   logFilePtr.p->logMaxGciCompleted[logFilePtr.p->currentMbyte] = 
     logPartPtr.p->logPartNewestCompletedGCI;
   logFilePtr.p->logMaxGciStarted[logFilePtr.p->currentMbyte] = cnewestGci;
@@ -13500,10 +13564,7 @@ void Dblqh::writeFileDescriptor(Signal* signal)
 /* ------------------------------------------------------------------------- */
 void Dblqh::writeFileHeaderOpen(Signal* signal, Uint32 wmoType) 
 {
-  LogFileRecordPtr wmoLogFilePtr;
   UintR twmoNoLogDescriptors;
-  UintR twmoLoop;
-  UintR twmoIndex;
 
 /* -------------------------------------------------- */
 /*       WRITE HEADER INFORMATION IN THE NEW FILE.    */
@@ -13511,52 +13572,44 @@ void Dblqh::writeFileHeaderOpen(Signal* signal, Uint32 wmoType)
   logPagePtr.p->logPageWord[ZPAGE_HEADER_SIZE + ZPOS_LOG_TYPE] = ZFD_TYPE;
   logPagePtr.p->logPageWord[ZPAGE_HEADER_SIZE + ZPOS_FILE_NO] = 
     logFilePtr.p->fileNo;
-  if (logPartPtr.p->noLogFiles > ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+  if (logPartPtr.p->noLogFiles > cmaxLogFilesInPageZero) {
     jam();
-    twmoNoLogDescriptors = ZMAX_LOG_FILES_IN_PAGE_ZERO;
+    twmoNoLogDescriptors = cmaxLogFilesInPageZero;
   } else {
     jam();
     twmoNoLogDescriptors = logPartPtr.p->noLogFiles;
   }//if
   logPagePtr.p->logPageWord[ZPAGE_HEADER_SIZE + ZPOS_NO_FD] = 
     twmoNoLogDescriptors;
-  wmoLogFilePtr.i = logFilePtr.i;
+
-  twmoLoop = 0;
+  {
-WMO_LOOP:
+    Uint32 pos = ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE;
-  jam();
+    LogFileRecordPtr filePtr = logFilePtr;
-  if (twmoLoop < twmoNoLogDescriptors) {
+    for (Uint32 fd = 0; fd < twmoNoLogDescriptors; fd++)
-    jam();
+    {
-    ptrCheckGuard(wmoLogFilePtr, clogFileFileSize, logFileRecord);
-    for (twmoIndex = 0; twmoIndex <= ZNO_MBYTES_IN_FILE - 1; twmoIndex++) {
       jam();
-      arrGuard(((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
+      ptrCheckGuard(filePtr, clogFileFileSize, logFileRecord);
-                (twmoLoop * ZFD_PART_SIZE)) + twmoIndex, ZPAGE_SIZE);
+      for (Uint32 mb = 0; mb < clogFileSize; mb ++)
-      logPagePtr.p->logPageWord[((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
+      {
-                   (twmoLoop * ZFD_PART_SIZE)) + twmoIndex] = 
+        jam();
-            wmoLogFilePtr.p->logMaxGciCompleted[twmoIndex];
+        Uint32 pos0 = pos + fd * (ZFD_MBYTE_SIZE * clogFileSize) + mb;
-      arrGuard((((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) +
+        Uint32 pos1 = pos0 + clogFileSize;
-                 (twmoLoop * ZFD_PART_SIZE)) + ZNO_MBYTES_IN_FILE) +
+        Uint32 pos2 = pos1 + clogFileSize;
-                  twmoIndex, ZPAGE_SIZE);
+        arrGuard(pos0, ZPAGE_SIZE);
-      logPagePtr.p->logPageWord[(((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
+        arrGuard(pos1, ZPAGE_SIZE);
-           (twmoLoop * ZFD_PART_SIZE)) + ZNO_MBYTES_IN_FILE) + twmoIndex] = 
+        arrGuard(pos2, ZPAGE_SIZE);
-         wmoLogFilePtr.p->logMaxGciStarted[twmoIndex];
+        logPagePtr.p->logPageWord[pos0] = filePtr.p->logMaxGciCompleted[mb];
-      arrGuard((((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) +
+        logPagePtr.p->logPageWord[pos1] = filePtr.p->logMaxGciStarted[mb];
-        (twmoLoop * ZFD_PART_SIZE)) + (2 * ZNO_MBYTES_IN_FILE)) +
+        logPagePtr.p->logPageWord[pos2] = filePtr.p->logLastPrepRef[mb];
-         twmoIndex, ZPAGE_SIZE);
+      }
-      logPagePtr.p->logPageWord[(((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
+      filePtr.i = filePtr.p->prevLogFile;
-        (twmoLoop * ZFD_PART_SIZE)) + (2 * ZNO_MBYTES_IN_FILE)) + twmoIndex] = 
+    }
-          wmoLogFilePtr.p->logLastPrepRef[twmoIndex];
+    pos += (twmoNoLogDescriptors * ZFD_MBYTE_SIZE * clogFileSize);
-    }//for
+    arrGuard(pos, ZPAGE_SIZE);
-    wmoLogFilePtr.i = wmoLogFilePtr.p->prevLogFile;
+    logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX] = pos;
-    twmoLoop = twmoLoop + 1;
+    logPagePtr.p->logPageWord[pos] = ZNEXT_LOG_RECORD_TYPE;
-    goto WMO_LOOP;
+  }
-  }//if
+
-  logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX] = 
-    (ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) +
-    (ZFD_PART_SIZE * twmoNoLogDescriptors);
-  arrGuard(logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX], ZPAGE_SIZE);
-  logPagePtr.p->logPageWord[logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX]] = 
-       ZNEXT_LOG_RECORD_TYPE;
 /* ------------------------------------------------------- */
 /*       THIS IS A SPECIAL WRITE OF THE FIRST PAGE IN THE  */
 /*       LOG FILE. THIS HAS SPECIAL SIGNIFANCE TO FIND     */
@@ -13701,9 +13754,9 @@ void Dblqh::openSrLastFileLab(Signal* signal)
 void Dblqh::readSrLastFileLab(Signal* signal) 
 {
   logPartPtr.p->logLap = logPagePtr.p->logPageWord[ZPOS_LOG_LAP];
-  if (logPartPtr.p->noLogFiles > ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+  if (logPartPtr.p->noLogFiles > cmaxLogFilesInPageZero) {
     jam();
-    initGciInLogFileRec(signal, ZMAX_LOG_FILES_IN_PAGE_ZERO);
+    initGciInLogFileRec(signal, cmaxLogFilesInPageZero);
   } else {
     jam();
     initGciInLogFileRec(signal, logPartPtr.p->noLogFiles);
@@ -13728,7 +13781,7 @@ void Dblqh::readSrLastMbyteLab(Signal* signal)
       logPartPtr.p->lastMbyte = logFilePtr.p->currentMbyte - 1;
     }//if
   }//if
-  arrGuard(logFilePtr.p->currentMbyte, 16);
+  arrGuard(logFilePtr.p->currentMbyte, clogFileSize);
   logFilePtr.p->logMaxGciCompleted[logFilePtr.p->currentMbyte] = 
     logPagePtr.p->logPageWord[ZPOS_MAX_GCI_COMPLETED];
   logFilePtr.p->logMaxGciStarted[logFilePtr.p->currentMbyte] = 
@@ -13736,7 +13789,7 @@ void Dblqh::readSrLastMbyteLab(Signal* signal)
   logFilePtr.p->logLastPrepRef[logFilePtr.p->currentMbyte] = 
     logPagePtr.p->logPageWord[ZLAST_LOG_PREP_REF];
   releaseLogpage(signal);
-  if (logFilePtr.p->currentMbyte < (ZNO_MBYTES_IN_FILE - 1)) {
+  if (logFilePtr.p->currentMbyte < (clogFileSize - 1)) {
     jam();
     logFilePtr.p->currentMbyte++;
     readSinglePage(signal, ZPAGES_IN_MBYTE * logFilePtr.p->currentMbyte);
@@ -13750,21 +13803,21 @@ void Dblqh::readSrLastMbyteLab(Signal* signal)
      * ---------------------------------------------------------------------- */
     if (logPartPtr.p->lastMbyte == ZNIL) {
       jam();
-      logPartPtr.p->lastMbyte = ZNO_MBYTES_IN_FILE - 1;
+      logPartPtr.p->lastMbyte = clogFileSize - 1;
     }//if
   }//if
   logFilePtr.p->logFileStatus = LogFileRecord::CLOSING_SR;
   closeFile(signal, logFilePtr, __LINE__);
-  if (logPartPtr.p->noLogFiles > ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+  if (logPartPtr.p->noLogFiles > cmaxLogFilesInPageZero) {
     Uint32 fileNo;
-    if (logFilePtr.p->fileNo >= ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+    if (logFilePtr.p->fileNo >= cmaxLogFilesInPageZero) {
       jam();
-      fileNo = logFilePtr.p->fileNo - ZMAX_LOG_FILES_IN_PAGE_ZERO;
+      fileNo = logFilePtr.p->fileNo - cmaxLogFilesInPageZero;
     } else {
       jam();
       fileNo = 
 	(logPartPtr.p->noLogFiles + logFilePtr.p->fileNo) - 
-	ZMAX_LOG_FILES_IN_PAGE_ZERO;
+	cmaxLogFilesInPageZero;
     }//if
     if (fileNo == 0) {
       jam();
@@ -13774,11 +13827,11 @@ void Dblqh::readSrLastMbyteLab(Signal* signal)
        * -------------------------------------------------------------------- */
       fileNo = 1;
       logPartPtr.p->srRemainingFiles = 
-	logPartPtr.p->noLogFiles - (ZMAX_LOG_FILES_IN_PAGE_ZERO - 1);
+	logPartPtr.p->noLogFiles - (cmaxLogFilesInPageZero - 1);
     } else {
       jam();
       logPartPtr.p->srRemainingFiles = 
-	logPartPtr.p->noLogFiles - ZMAX_LOG_FILES_IN_PAGE_ZERO;
+	logPartPtr.p->noLogFiles - cmaxLogFilesInPageZero;
     }//if
     LogFileRecordPtr locLogFilePtr;
     findLogfile(signal, fileNo, logPartPtr, &locLogFilePtr);
@@ -13803,9 +13856,9 @@ void Dblqh::openSrNextFileLab(Signal* signal)
 
 void Dblqh::readSrNextFileLab(Signal* signal) 
 {
-  if (logPartPtr.p->srRemainingFiles > ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+  if (logPartPtr.p->srRemainingFiles > cmaxLogFilesInPageZero) {
     jam();
-    initGciInLogFileRec(signal, ZMAX_LOG_FILES_IN_PAGE_ZERO);
+    initGciInLogFileRec(signal, cmaxLogFilesInPageZero);
   } else {
     jam();
     initGciInLogFileRec(signal, logPartPtr.p->srRemainingFiles);
@@ -13813,16 +13866,16 @@ void Dblqh::readSrNextFileLab(Signal* signal)
   releaseLogpage(signal);
   logFilePtr.p->logFileStatus = LogFileRecord::CLOSING_SR;
   closeFile(signal, logFilePtr, __LINE__);
-  if (logPartPtr.p->srRemainingFiles > ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+  if (logPartPtr.p->srRemainingFiles > cmaxLogFilesInPageZero) {
     Uint32 fileNo;
-    if (logFilePtr.p->fileNo >= ZMAX_LOG_FILES_IN_PAGE_ZERO) {
+    if (logFilePtr.p->fileNo >= cmaxLogFilesInPageZero) {
       jam();
-      fileNo = logFilePtr.p->fileNo - ZMAX_LOG_FILES_IN_PAGE_ZERO;
+      fileNo = logFilePtr.p->fileNo - cmaxLogFilesInPageZero;
     } else {
       jam();
       fileNo = 
 	(logPartPtr.p->noLogFiles + logFilePtr.p->fileNo) - 
-	ZMAX_LOG_FILES_IN_PAGE_ZERO;
+	cmaxLogFilesInPageZero;
     }//if
     if (fileNo == 0) {
       jam();
@@ -13831,11 +13884,11 @@ void Dblqh::readSrNextFileLab(Signal* signal)
        * -------------------------------------------------------------------- */
       fileNo = 1;
       logPartPtr.p->srRemainingFiles = 
-	logPartPtr.p->srRemainingFiles - (ZMAX_LOG_FILES_IN_PAGE_ZERO - 1);
+	logPartPtr.p->srRemainingFiles - (cmaxLogFilesInPageZero - 1);
     } else {
       jam();
       logPartPtr.p->srRemainingFiles = 
-	logPartPtr.p->srRemainingFiles - ZMAX_LOG_FILES_IN_PAGE_ZERO;
+	logPartPtr.p->srRemainingFiles - cmaxLogFilesInPageZero;
     }//if
     LogFileRecordPtr locLogFilePtr;
     findLogfile(signal, fileNo, logPartPtr, &locLogFilePtr);
@@ -14706,7 +14759,7 @@ void Dblqh::srLogLimits(Signal* signal)
    *   EXECUTED.
    * ----------------------------------------------------------------------- */
   while(true) {
-    ndbrequire(tmbyte < 16);
+    ndbrequire(tmbyte < clogFileSize);
     if (logPartPtr.p->logExecState == LogPartRecord::LES_SEARCH_STOP) {
       if (logFilePtr.p->logMaxGciCompleted[tmbyte] < logPartPtr.p->logLastGci) {
         jam();
@@ -14747,7 +14800,7 @@ void Dblqh::srLogLimits(Signal* signal)
     if (logPartPtr.p->logExecState != LogPartRecord::LES_EXEC_LOG) {
       if (tmbyte == 0) {
         jam();
-        tmbyte = ZNO_MBYTES_IN_FILE - 1;
+        tmbyte = clogFileSize - 1;
         logFilePtr.i = logFilePtr.p->prevLogFile;
         ptrCheckGuard(logFilePtr, clogFileFileSize, logFileRecord);
       } else {
@@ -15141,7 +15194,7 @@ void Dblqh::execSr(Signal* signal)
 	  logPagePtr.p->logPageWord[ZPAGE_HEADER_SIZE + ZPOS_NO_FD];
           logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX] = 
 	      (ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
-	      (noFdDescriptors * ZFD_PART_SIZE);
+	      (noFdDescriptors * ZFD_MBYTE_SIZE * clogFileSize);
       }
       break;
 /* ========================================================================= */
@@ -15181,11 +15234,11 @@ void Dblqh::execSr(Signal* signal)
 /*---------------------------------------------------------------------------*/
 /* START EXECUTION OF A NEW MBYTE IN THE LOG.                                */
 /*---------------------------------------------------------------------------*/
-      if (logFilePtr.p->currentMbyte < (ZNO_MBYTES_IN_FILE - 1)) {
+      if (logFilePtr.p->currentMbyte < (clogFileSize - 1)) {
         jam();
         logPartPtr.p->logExecState = LogPartRecord::LES_EXEC_LOG_NEW_MBYTE;
       } else {
-        ndbrequire(logFilePtr.p->currentMbyte == (ZNO_MBYTES_IN_FILE - 1));
+        ndbrequire(logFilePtr.p->currentMbyte == (clogFileSize - 1));
         jam();
 /*---------------------------------------------------------------------------*/
 /* WE HAVE TO CHANGE FILE. CLOSE THIS ONE AND THEN OPEN THE NEXT.            */
@@ -15380,7 +15433,7 @@ void Dblqh::invalidateLogAfterLastGCI(Signal* signal) {
     jam();
     releaseLfo(signal);
     releaseLogpage(signal); 
-    if (logPartPtr.p->invalidatePageNo < (ZNO_MBYTES_IN_FILE * ZPAGES_IN_MBYTE - 1)) {
+    if (logPartPtr.p->invalidatePageNo < (clogFileSize * ZPAGES_IN_MBYTE - 1)) {
       // We continue in this file.
       logPartPtr.p->invalidatePageNo++;
     } else {
@@ -16721,6 +16774,22 @@ void Dblqh::initialiseLogFile(Signal* signal)
       ptrAss(logFilePtr, logFileRecord);
       logFilePtr.p->nextLogFile = logFilePtr.i + 1;
       logFilePtr.p->logFileStatus = LogFileRecord::LFS_IDLE;
+
+      logFilePtr.p->logLastPrepRef = new Uint32[clogFileSize];
+      logFilePtr.p->logMaxGciCompleted = new Uint32[clogFileSize];
+      logFilePtr.p->logMaxGciStarted = new Uint32[clogFileSize];
+
+      if (logFilePtr.p->logLastPrepRef == 0 ||
+          logFilePtr.p->logMaxGciCompleted == 0 ||
+          logFilePtr.p->logMaxGciStarted == 0)
+      {
+        char buf[256];
+        BaseString::snprintf(buf, sizeof(buf),
+                             "Failed to alloc mbyte(%u) arrays for logfile %u",
+                             clogFileSize, logFilePtr.i);
+        progError(__LINE__, NDBD_EXIT_MEMALLOC, buf);
+      }
+
     }//for
     logFilePtr.i = clogFileFileSize - 1;
     ptrAss(logFilePtr, logFileRecord);
@@ -17049,41 +17118,31 @@ void Dblqh::initFragrec(Signal* signal,
  * ========================================================================= */
 void Dblqh::initGciInLogFileRec(Signal* signal, Uint32 noFdDescriptors) 
 {
-  LogFileRecordPtr iglLogFilePtr;
+  LogFileRecordPtr filePtr = logFilePtr;
-  UintR tiglLoop;
+  Uint32 pos = ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE;
-  UintR tiglIndex;
+  for (Uint32 fd = 0; fd < noFdDescriptors; fd++)
-
+  {
-  tiglLoop = 0;
-  iglLogFilePtr.i = logFilePtr.i;
-  iglLogFilePtr.p = logFilePtr.p;
-IGL_LOOP:
-  for (tiglIndex = 0; tiglIndex <= ZNO_MBYTES_IN_FILE - 1; tiglIndex++) {
-    arrGuard(((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
-	      (tiglLoop * ZFD_PART_SIZE)) + tiglIndex, ZPAGE_SIZE);
-    iglLogFilePtr.p->logMaxGciCompleted[tiglIndex] = 
-      logPagePtr.p->logPageWord[((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
-				 (tiglLoop * ZFD_PART_SIZE)) + tiglIndex];
-    arrGuard((((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + ZNO_MBYTES_IN_FILE) + 
-	      (tiglLoop * ZFD_PART_SIZE)) + tiglIndex, ZPAGE_SIZE);
-    iglLogFilePtr.p->logMaxGciStarted[tiglIndex] = 
-      logPagePtr.p->logPageWord[(((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
-				  ZNO_MBYTES_IN_FILE) + 
-				 (tiglLoop * ZFD_PART_SIZE)) + tiglIndex];
-    arrGuard((((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
-	       (2 * ZNO_MBYTES_IN_FILE)) + (tiglLoop * ZFD_PART_SIZE)) + 
-	     tiglIndex, ZPAGE_SIZE);
-    iglLogFilePtr.p->logLastPrepRef[tiglIndex] = 
-      logPagePtr.p->logPageWord[(((ZPAGE_HEADER_SIZE + ZFD_HEADER_SIZE) + 
-				  (2 * ZNO_MBYTES_IN_FILE)) + 
-				 (tiglLoop * ZFD_PART_SIZE)) + tiglIndex];
-  }//for
-  tiglLoop = tiglLoop + 1;
-  if (tiglLoop < noFdDescriptors) {
     jam();
-    iglLogFilePtr.i = iglLogFilePtr.p->prevLogFile;
+    for (Uint32 mb = 0; mb < clogFileSize; mb++)
-    ptrCheckGuard(iglLogFilePtr, clogFileFileSize, logFileRecord);
+    {
-    goto IGL_LOOP;
+      jam();
-  }//if
+      Uint32 pos0 = pos + fd * (ZFD_MBYTE_SIZE * clogFileSize) + mb;
+      Uint32 pos1 = pos0 + clogFileSize;
+      Uint32 pos2 = pos1 + clogFileSize;
+      arrGuard(pos0, ZPAGE_SIZE);
+      arrGuard(pos1, ZPAGE_SIZE);
+      arrGuard(pos2, ZPAGE_SIZE);
+      filePtr.p->logMaxGciCompleted[mb] = logPagePtr.p->logPageWord[pos0];
+      filePtr.p->logMaxGciStarted[mb] = logPagePtr.p->logPageWord[pos1];
+      filePtr.p->logLastPrepRef[mb] = logPagePtr.p->logPageWord[pos2];
+    }
+    if (fd + 1 < noFdDescriptors)
+    {
+      jam();
+      filePtr.i = filePtr.p->prevLogFile;
+      ptrCheckGuard(filePtr, clogFileFileSize, logFileRecord);
+    }
+  }
 }//Dblqh::initGciInLogFileRec()
 
 /* ========================================================================== 
@@ -18336,7 +18395,7 @@ void Dblqh::writeNextLog(Signal* signal)
   ndbrequire(logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX] < ZPAGE_SIZE);
   logPagePtr.p->logPageWord[logPagePtr.p->logPageWord[ZCURR_PAGE_INDEX]] = 
     ZNEXT_MBYTE_TYPE;
-  if (logFilePtr.p->currentMbyte == (ZNO_MBYTES_IN_FILE - 1)) {
+  if (logFilePtr.p->currentMbyte == (clogFileSize - 1)) {
     jam();
 /* -------------------------------------------------- */
 /*       CALCULATE THE NEW REMAINING WORDS WHEN       */
@@ -18425,7 +18484,7 @@ void Dblqh::writeNextLog(Signal* signal)
       systemError(signal, __LINE__);
     }//if
   }//if
-  if (logFilePtr.p->currentMbyte == (ZNO_MBYTES_IN_FILE - 1)) {
+  if (logFilePtr.p->currentMbyte == (clogFileSize - 1)) {
     jam();
     twnlNextMbyte = 0;
     if (logFilePtr.p->fileChangeState != LogFileRecord::NOT_ONGOING) {

storage/ndb/src/kernel/blocks/dblqh/Makefile.am

@@ -16,7 +16,7 @@
 EXTRA_PROGRAMS = ndbd_redo_log_reader
 
 ndbd_redo_log_reader_SOURCES = redoLogReader/records.cpp \
-  redoLogReader/redoLogFileReader.cpp
+  redoLogReader/reader.cpp
 
 include $(top_srcdir)/storage/ndb/config/common.mk.am
 include $(top_srcdir)/storage/ndb/config/type_kernel.mk.am

storage/ndb/src/kernel/blocks/dbtup/Dbtup.hpp

@@ -2700,6 +2700,7 @@ private:
 
   ArrayPool<Page> c_page_pool;
   Uint32 cnoOfAllocatedPages;
+  Uint32 m_max_allocate_pages;
   
   Tablerec *tablerec;
   Uint32 cnoOfTablerec;

storage/ndb/src/kernel/blocks/dbtup/DbtupDebug.cpp

@@ -74,6 +74,10 @@ Dbtup::reportMemoryUsage(Signal* signal, int incDec){
   sendSignal(CMVMI_REF, GSN_EVENT_REP, signal, 6, JBB);
 }
 
+#ifdef VM_TRACE
+extern Uint32 fc_left, fc_right, fc_remove;
+#endif
+
 void
 Dbtup::execDUMP_STATE_ORD(Signal* signal)
 {
@@ -155,12 +159,20 @@ Dbtup::execDUMP_STATE_ORD(Signal* signal)
     return;
   }//if
 #endif
-#if defined VM_TRACE && 0
+#if defined VM_TRACE
-  if (type == 1211){
+  if (type == 1211 || type == 1212 || type == 1213){
-    ndbout_c("Startar modul test av Page Manager");
+    Uint32 seed = time(0);
+    if (signal->getLength() > 1)
+      seed = signal->theData[1];
+    ndbout_c("Startar modul test av Page Manager (seed: 0x%x)", seed);
+    srand(seed);
 
     Vector<Chunk> chunks;
     const Uint32 LOOPS = 1000;
+    Uint32 sum_req = 0;
+    Uint32 sum_conf = 0;
+    Uint32 sum_loop = 0;
+    Uint32 max_loop = 0;
     for(Uint32 i = 0; i<LOOPS; i++){
 
       // Case
@@ -178,7 +190,14 @@ Dbtup::execDUMP_STATE_ORD(Signal* signal)
 	c = 1 + rand() % 2;
       }
       
-      ndbout_c("loop=%d case=%d free=%d alloc=%d", i, c, free, alloc);
+      if (type == 1211)
+        ndbout_c("loop=%d case=%d free=%d alloc=%d", i, c, free, alloc);
+
+      if (type == 1213)
+      {
+        c = 1;
+        alloc = 2 + (sum_conf >> 3) + (sum_conf >> 4);
+      }
       switch(c){ 
       case 0:{ // Release
 	const int ch = rand() % chunks.size();
@@ -190,23 +209,33 @@ Dbtup::execDUMP_STATE_ORD(Signal* signal)
       case 2: { // Seize(n) - fail
 	alloc += free;
 	// Fall through
+        sum_req += free;
+        goto doalloc;
       }
       case 1: { // Seize(n) (success)
-
+        sum_req += alloc;
+    doalloc:
 	Chunk chunk;
 	allocConsPages(alloc, chunk.pageCount, chunk.pageId);
 	ndbrequire(chunk.pageCount <= alloc);
 	if(chunk.pageCount != 0){
 	  chunks.push_back(chunk);
 	  if(chunk.pageCount != alloc) {
-	    ndbout_c("  Tried to allocate %d - only allocated %d - free: %d",
+	    if (type == 1211)
-		     alloc, chunk.pageCount, free);
+              ndbout_c("  Tried to allocate %d - only allocated %d - free: %d",
+                       alloc, chunk.pageCount, free);
 	  }
 	} else {
 	  ndbout_c("  Failed to alloc %d pages with %d pages free",
 		   alloc, free);
 	}
 	
+        sum_conf += chunk.pageCount;
+        Uint32 tot = fc_left + fc_right + fc_remove;
+        sum_loop += tot;
+        if (tot > max_loop)
+          max_loop = tot;
+
 	for(Uint32 i = 0; i<chunk.pageCount; i++){
 	  PagePtr pagePtr;
 	  pagePtr.i = chunk.pageId + i;
@@ -225,6 +254,10 @@ Dbtup::execDUMP_STATE_ORD(Signal* signal)
       returnCommonArea(chunk.pageId, chunk.pageCount);      
       chunks.erase(chunks.size() - 1);
     }
+
+    ndbout_c("Got %u%% of requested allocs, loops : %u 100*avg: %u max: %u",
+             (100 * sum_conf) / sum_req, sum_loop, 100*sum_loop / LOOPS,
+             max_loop);
   }
 #endif
 }//Dbtup::execDUMP_STATE_ORD()

storage/ndb/src/kernel/blocks/dbtup/DbtupGen.cpp

@@ -305,6 +305,12 @@ void Dbtup::execREAD_CONFIG_REQ(Signal* signal)
   Uint32 noOfTriggers= 0;
   
   Uint32 tmp= 0;
+
+  if (ndb_mgm_get_int_parameter(p, CFG_DB_MAX_ALLOCATE, &tmp))
+    tmp = 32 * 1024 * 1024;
+  m_max_allocate_pages = (tmp + GLOBAL_PAGE_SIZE - 1) / GLOBAL_PAGE_SIZE;
+
+  tmp = 0;
   ndbrequire(!ndb_mgm_get_int_parameter(p, CFG_TUP_PAGE_RANGE, &tmp));
   initPageRangeSize(tmp);
   ndbrequire(!ndb_mgm_get_int_parameter(p, CFG_TUP_TABLE, &cnoOfTablerec));

storage/ndb/src/kernel/blocks/dbtup/DbtupPagMan.cpp

@@ -146,10 +146,17 @@ void Dbtup::initializePage()
   cnoOfAllocatedPages = tmp; // Is updated by returnCommonArea
 }//Dbtup::initializePage()
 
+#ifdef VM_TRACE
+Uint32 fc_left, fc_right, fc_remove;
+#endif
+
 void Dbtup::allocConsPages(Uint32 noOfPagesToAllocate,
                            Uint32& noOfPagesAllocated,
                            Uint32& allocPageRef)
 {
+#ifdef VM_TRACE
+  fc_left = fc_right = fc_remove = 0;
+#endif
   if (noOfPagesToAllocate == 0){ 
     jam();
     noOfPagesAllocated = 0;
@@ -228,7 +235,10 @@ void Dbtup::findFreeLeftNeighbours(Uint32& allocPageRef,
 {
   PagePtr pageFirstPtr, pageLastPtr;
   Uint32 remainAllocate = noOfPagesToAllocate - noPagesAllocated;
-  while (allocPageRef > 0) {
+  Uint32 loop = 0;
+  while (allocPageRef > 0 && 
+         ++loop < 16) 
+  {
     jam();
     pageLastPtr.i = allocPageRef - 1;
     c_page_pool.getPtr(pageLastPtr);
@@ -256,6 +266,9 @@ void Dbtup::findFreeLeftNeighbours(Uint32& allocPageRef,
         remainAllocate -= listSize;
       }//if
     }//if
+#ifdef VM_TRACE
+    fc_left++;
+#endif
   }//while
 }//Dbtup::findFreeLeftNeighbours()
 
@@ -269,7 +282,10 @@ void Dbtup::findFreeRightNeighbours(Uint32& allocPageRef,
     jam();
     return;
   }//if
-  while ((allocPageRef + noPagesAllocated) < c_page_pool.getSize()) {
+  Uint32 loop = 0;
+  while ((allocPageRef + noPagesAllocated) < c_page_pool.getSize() &&
+         ++loop < 16) 
+  {
     jam();
     pageFirstPtr.i = allocPageRef + noPagesAllocated;
     c_page_pool.getPtr(pageFirstPtr);
@@ -296,24 +312,37 @@ void Dbtup::findFreeRightNeighbours(Uint32& allocPageRef,
         remainAllocate -= listSize;
       }//if
     }//if
+#ifdef VM_TRACE
+    fc_right++;
+#endif
   }//while
 }//Dbtup::findFreeRightNeighbours()
 
 void Dbtup::insertCommonArea(Uint32 insPageRef, Uint32 insList) 
 {
   cnoOfAllocatedPages -= (1 << insList);
-  PagePtr pageLastPtr, pageInsPtr;
+  PagePtr pageLastPtr, pageInsPtr, pageHeadPtr;
 
+  pageHeadPtr.i = cfreepageList[insList];
   c_page_pool.getPtr(pageInsPtr, insPageRef);
   ndbrequire(insList < 16);
   pageLastPtr.i = (pageInsPtr.i + (1 << insList)) - 1;
 
-  pageInsPtr.p->next_cluster_page = cfreepageList[insList];
+  pageInsPtr.p->page_state = ZFREE_COMMON;
+  pageInsPtr.p->next_cluster_page = pageHeadPtr.i;
   pageInsPtr.p->prev_cluster_page = RNIL;
   pageInsPtr.p->last_cluster_page = pageLastPtr.i;
   cfreepageList[insList] = pageInsPtr.i;
 
+  if (pageHeadPtr.i != RNIL)
+  {
+    jam();
+    c_page_pool.getPtr(pageHeadPtr);
+    pageHeadPtr.p->prev_cluster_page = pageInsPtr.i;
+  }
+  
   c_page_pool.getPtr(pageLastPtr);
+  pageLastPtr.p->page_state = ZFREE_COMMON;
   pageLastPtr.p->first_cluster_page = pageInsPtr.i;
   pageLastPtr.p->next_page = RNIL;
 }//Dbtup::insertCommonArea()
@@ -321,12 +350,13 @@ void Dbtup::insertCommonArea(Uint32 insPageRef, Uint32 insList)
 void Dbtup::removeCommonArea(Uint32 remPageRef, Uint32 list) 
 {
   cnoOfAllocatedPages += (1 << list);  
-  PagePtr pagePrevPtr, pageNextPtr, pageLastPtr, pageSearchPtr, remPagePtr;
+  PagePtr pagePrevPtr, pageNextPtr, pageLastPtr, remPagePtr;
 
   c_page_pool.getPtr(remPagePtr, remPageRef);
   ndbrequire(list < 16);
   if (cfreepageList[list] == remPagePtr.i) {
     jam();
+    ndbassert(remPagePtr.p->prev_cluster_page == RNIL);
     cfreepageList[list] = remPagePtr.p->next_cluster_page;
     pageNextPtr.i = cfreepageList[list];
     if (pageNextPtr.i != RNIL) {
@@ -335,30 +365,25 @@ void Dbtup::removeCommonArea(Uint32 remPageRef, Uint32 list)
       pageNextPtr.p->prev_cluster_page = RNIL;
     }//if
   } else {
-    pageSearchPtr.i = cfreepageList[list];
+    pagePrevPtr.i = remPagePtr.p->prev_cluster_page;
-    while (true) {
-      jam();
-      c_page_pool.getPtr(pageSearchPtr);
-      pagePrevPtr = pageSearchPtr;
-      pageSearchPtr.i = pageSearchPtr.p->next_cluster_page;
-      if (pageSearchPtr.i == remPagePtr.i) {
-        jam();
-        break;
-      }//if
-    }//while
     pageNextPtr.i = remPagePtr.p->next_cluster_page;
+    c_page_pool.getPtr(pagePrevPtr);
     pagePrevPtr.p->next_cluster_page = pageNextPtr.i;
-    if (pageNextPtr.i != RNIL) {
+    if (pageNextPtr.i != RNIL)
+    {
       jam();
       c_page_pool.getPtr(pageNextPtr);
       pageNextPtr.p->prev_cluster_page = pagePrevPtr.i;
-    }//if
+    }
   }//if
   remPagePtr.p->next_cluster_page= RNIL;
   remPagePtr.p->last_cluster_page= RNIL;
   remPagePtr.p->prev_cluster_page= RNIL;
+  remPagePtr.p->page_state = ~ZFREE_COMMON;
 
   pageLastPtr.i = (remPagePtr.i + (1 << list)) - 1;
   c_page_pool.getPtr(pageLastPtr);
   pageLastPtr.p->first_cluster_page= RNIL;
+  pageLastPtr.p->page_state = ~ZFREE_COMMON;
+
 }//Dbtup::removeCommonArea()

storage/ndb/src/kernel/blocks/dbtup/DbtupPageMap.cpp

@@ -432,6 +432,11 @@ void Dbtup::allocMoreFragPages(Fragrecord* const regFragPtr)
 // We will grow by 18.75% plus two more additional pages to grow
 // a little bit quicker in the beginning.
 /* -----------------------------------------------------------------*/
+
+  if (noAllocPages > m_max_allocate_pages)
+  {
+    noAllocPages = m_max_allocate_pages;
+  }
   Uint32 allocated = allocFragPages(regFragPtr, noAllocPages);
   regFragPtr->noOfPagesToGrow += allocated;
 }//Dbtup::allocMoreFragPages()

storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp

@@ -277,6 +277,14 @@ void Ndbcntr::execSTTOR(Signal* signal)
     break;
   case ZSTART_PHASE_1:
     jam();
+    {
+      Uint32 db_watchdog_interval = 0;
+      const ndb_mgm_configuration_iterator * p = 
+        m_ctx.m_config.getOwnConfigIterator();
+      ndb_mgm_get_int_parameter(p, CFG_DB_WATCHDOG_INTERVAL, &db_watchdog_interval);
+      ndbrequire(db_watchdog_interval);
+      update_watch_dog_timer(db_watchdog_interval);
+    }
     startPhase1Lab(signal);
     break;
   case ZSTART_PHASE_2:
@@ -1410,6 +1418,13 @@ void Ndbcntr::execNODE_FAILREP(Signal* signal)
 {
   jamEntry();
 
+  if (ERROR_INSERTED(1001))
+  {
+    sendSignalWithDelay(reference(), GSN_NODE_FAILREP, signal, 100, 
+                        signal->getLength());
+    return;
+  }
+  
   const NodeFailRep * nodeFail = (NodeFailRep *)&signal->theData[0];
   NdbNodeBitmask allFailed; 
   allFailed.assign(NdbNodeBitmask::Size, nodeFail->theNodes);
@@ -2734,16 +2749,34 @@ void Ndbcntr::execSTART_ORD(Signal* signal){
   c_missra.execSTART_ORD(signal);
 }
 
+#define CLEAR_DX 13
+#define CLEAR_LCP 3
+
 void
-Ndbcntr::clearFilesystem(Signal* signal){
+Ndbcntr::clearFilesystem(Signal* signal)
+{
+  const Uint32 lcp = c_fsRemoveCount >= CLEAR_DX;
+  
   FsRemoveReq * req  = (FsRemoveReq *)signal->getDataPtrSend();
   req->userReference = reference();
   req->userPointer   = 0;
   req->directory     = 1;
   req->ownDirectory  = 1;
-  FsOpenReq::setVersion(req->fileNumber, 3);
+
-  FsOpenReq::setSuffix(req->fileNumber, FsOpenReq::S_CTL); // Can by any...
+  if (lcp == 0)
-  FsOpenReq::v1_setDisk(req->fileNumber, c_fsRemoveCount);
+  {
+    FsOpenReq::setVersion(req->fileNumber, 3);
+    FsOpenReq::setSuffix(req->fileNumber, FsOpenReq::S_CTL); // Can by any...
+    FsOpenReq::v1_setDisk(req->fileNumber, c_fsRemoveCount);
+  }
+  else
+  {
+    FsOpenReq::setVersion(req->fileNumber, 5);
+    FsOpenReq::setSuffix(req->fileNumber, FsOpenReq::S_DATA);
+    FsOpenReq::v5_setLcpNo(req->fileNumber, c_fsRemoveCount - CLEAR_DX);
+    FsOpenReq::v5_setTableId(req->fileNumber, 0);
+    FsOpenReq::v5_setFragmentId(req->fileNumber, 0);
+  }
   sendSignal(NDBFS_REF, GSN_FSREMOVEREQ, signal, 
              FsRemoveReq::SignalLength, JBA);
   c_fsRemoveCount++;
@@ -2752,12 +2785,12 @@ Ndbcntr::clearFilesystem(Signal* signal){
 void
 Ndbcntr::execFSREMOVECONF(Signal* signal){
   jamEntry();
-  if(c_fsRemoveCount == 13){
+  if(c_fsRemoveCount == CLEAR_DX + CLEAR_LCP){
     jam();
     sendSttorry(signal);
   } else {
     jam();
-    ndbrequire(c_fsRemoveCount < 13);
+    ndbrequire(c_fsRemoveCount < CLEAR_DX + CLEAR_LCP);
     clearFilesystem(signal);
   }//if
 }

storage/ndb/src/kernel/blocks/ndbfs/AsyncFile.cpp

@@ -163,7 +163,12 @@ AsyncFile::run()
   theStartFlag = true;
   // Create write buffer for bigger writes
   theWriteBufferSize = WRITEBUFFERSIZE;
-  theWriteBuffer = (char *) ndbd_malloc(theWriteBufferSize); 
+  theWriteBufferUnaligned = (char *) ndbd_malloc(theWriteBufferSize +
+                                                 NDB_O_DIRECT_WRITE_ALIGNMENT-1);
+  theWriteBuffer = (char *)
+    (((UintPtr)theWriteBufferUnaligned + NDB_O_DIRECT_WRITE_ALIGNMENT - 1) &
+     ~(UintPtr)(NDB_O_DIRECT_WRITE_ALIGNMENT - 1));
+
   NdbMutex_Unlock(theStartMutexPtr);
   NdbCondition_Signal(theStartConditionPtr);
   
@@ -247,6 +252,78 @@ AsyncFile::run()
 static char g_odirect_readbuf[2*GLOBAL_PAGE_SIZE -1];
 #endif
 
+int
+AsyncFile::check_odirect_write(Uint32 flags, int& new_flags, int mode)
+{
+  assert(new_flags & (O_CREAT | O_TRUNC));
+#ifdef O_DIRECT
+  int ret;
+  char * bufptr = (char*)((UintPtr(g_odirect_readbuf)+(GLOBAL_PAGE_SIZE - 1)) & ~(GLOBAL_PAGE_SIZE - 1));
+  while (((ret = ::write(theFd, bufptr, GLOBAL_PAGE_SIZE)) == -1) && 
+         (errno == EINTR));
+  if (ret == -1)
+  {
+    new_flags &= ~O_DIRECT;
+    ndbout_c("%s Failed to write using O_DIRECT, disabling", 
+             theFileName.c_str());
+  }
+  
+  close(theFd);
+  theFd = ::open(theFileName.c_str(), new_flags, mode);
+  if (theFd == -1)
+    return errno;
+#endif
+
+  return 0;
+}
+
+int
+AsyncFile::check_odirect_read(Uint32 flags, int &new_flags, int mode)
+{
+#ifdef O_DIRECT
+  int ret;
+  char * bufptr = (char*)((UintPtr(g_odirect_readbuf)+(GLOBAL_PAGE_SIZE - 1)) & ~(GLOBAL_PAGE_SIZE - 1));
+  while (((ret = ::read(theFd, bufptr, GLOBAL_PAGE_SIZE)) == -1) && 
+         (errno == EINTR));
+  if (ret == -1)
+  {
+    ndbout_c("%s Failed to read using O_DIRECT, disabling", 
+             theFileName.c_str());
+    goto reopen;
+  }
+  
+  if(lseek(theFd, 0, SEEK_SET) != 0)
+  {
+    return errno;
+  }
+  
+  if ((flags & FsOpenReq::OM_CHECK_SIZE) == 0)
+  {
+    struct stat buf;
+    if ((fstat(theFd, &buf) == -1))
+    {
+      return errno;
+    } 
+    else if ((buf.st_size % GLOBAL_PAGE_SIZE) != 0)
+    {
+      ndbout_c("%s filesize not a multiple of %d, disabling O_DIRECT", 
+               theFileName.c_str(), GLOBAL_PAGE_SIZE);
+      goto reopen;
+    }
+  }
+  
+  return 0;
+  
+reopen:
+  close(theFd);
+  new_flags &= ~O_DIRECT;
+  theFd = ::open(theFileName.c_str(), new_flags, mode);
+  if (theFd == -1)
+    return errno;  
+#endif
+  return 0;
+}
+
 void AsyncFile::openReq(Request* request)
 {  
   m_auto_sync_freq = 0;
@@ -312,7 +389,7 @@ void AsyncFile::openReq(Request* request)
   }
 #else
   Uint32 flags = request->par.open.flags;
-  Uint32 new_flags = 0;
+  int new_flags = 0;
 
   // Convert file open flags from Solaris to Liux
   if (flags & FsOpenReq::OM_CREATE)
@@ -343,10 +420,6 @@ void AsyncFile::openReq(Request* request)
   {
     new_flags |= O_DIRECT;
   }
-#elif defined O_SYNC
-  {
-    flags |= FsOpenReq::OM_SYNC;
-  }
 #endif
   
   if ((flags & FsOpenReq::OM_SYNC) && ! (flags & FsOpenReq::OM_INIT))
@@ -356,14 +429,18 @@ void AsyncFile::openReq(Request* request)
 #endif
   }
 
+  const char * rw = "";
   switch(flags & 0x3){
   case FsOpenReq::OM_READONLY:
+    rw = "r";
     new_flags |= O_RDONLY;
     break;
   case FsOpenReq::OM_WRITEONLY:
+    rw = "w";
     new_flags |= O_WRONLY;
     break;
   case FsOpenReq::OM_READWRITE:
+    rw = "rw";
     new_flags |= O_RDWR;
     break;
   default:
@@ -404,11 +481,6 @@ no_odirect:
 	if (new_flags & O_DIRECT)
 	{
 	  new_flags &= ~O_DIRECT;
-	  flags |= FsOpenReq::OM_SYNC;
-#ifdef O_SYNC
-	  if (! (flags & FsOpenReq::OM_INIT))
-	    new_flags |= O_SYNC;
-#endif
 	  goto no_odirect;
 	}
 #endif
@@ -421,11 +493,6 @@ no_odirect:
     else if (new_flags & O_DIRECT)
     {
       new_flags &= ~O_DIRECT;
-      flags |= FsOpenReq::OM_SYNC;
-#ifdef O_SYNC
-      if (! (flags & FsOpenReq::OM_INIT))
-	new_flags |= O_SYNC;
-#endif
       goto no_odirect;
     }
 #endif
@@ -512,7 +579,6 @@ no_odirect:
 	{
 	  ndbout_c("error on first write(%d), disable O_DIRECT", err);
 	  new_flags &= ~O_DIRECT;
-	  flags |= FsOpenReq::OM_SYNC;
 	  close(theFd);
 	  theFd = ::open(theFileName.c_str(), new_flags, mode);
 	  if (theFd != -1)
@@ -532,26 +598,32 @@ no_odirect:
   else if (flags & FsOpenReq::OM_DIRECT)
   {
 #ifdef O_DIRECT
-    do {
+    if (flags & (FsOpenReq::OM_TRUNCATE | FsOpenReq::OM_CREATE))
-      int ret;
+    {
-      char * bufptr = (char*)((UintPtr(g_odirect_readbuf)+(GLOBAL_PAGE_SIZE - 1)) & ~(GLOBAL_PAGE_SIZE - 1));
+      request->error = check_odirect_write(flags, new_flags, mode);
-      while (((ret = ::read(theFd, bufptr, GLOBAL_PAGE_SIZE)) == -1) && (errno == EINTR));
+    }
-      if (ret == -1)
+    else
-      {
+    {
-	ndbout_c("%s Failed to read using O_DIRECT, disabling", theFileName.c_str());
+      request->error = check_odirect_read(flags, new_flags, mode);
-	flags |= FsOpenReq::OM_SYNC;
+    }
-	flags |= FsOpenReq::OM_INIT;
+    
-	break;
+    if (request->error)
-      }
+      return;
-      if(lseek(theFd, 0, SEEK_SET) != 0)
-      {
-	request->error = errno;
-	return;
-      }
-    } while (0);
 #endif
   }
-
+#ifdef VM_TRACE
+  if (flags & FsOpenReq::OM_DIRECT)
+  {
+#ifdef O_DIRECT
+    ndbout_c("%s %s O_DIRECT: %d",
+             theFileName.c_str(), rw,
+             !!(new_flags & O_DIRECT));
+#else
+    ndbout_c("%s %s O_DIRECT: 0",
+             theFileName.c_str(), rw);
+#endif
+  }
+#endif  
   if ((flags & FsOpenReq::OM_SYNC) && (flags & FsOpenReq::OM_INIT))
   {
 #ifdef O_SYNC
@@ -562,6 +634,10 @@ no_odirect:
     new_flags &= ~(O_CREAT | O_TRUNC);
     new_flags |= O_SYNC;
     theFd = ::open(theFileName.c_str(), new_flags, mode);
+    if (theFd == -1)
+    {
+      request->error = errno;
+    }
 #endif
   }
 #endif
@@ -1079,7 +1155,8 @@ AsyncFile::rmrfReq(Request * request, char * path, bool removePath){
 void AsyncFile::endReq()
 {
   // Thread is ended with return
-  if (theWriteBuffer) ndbd_free(theWriteBuffer, theWriteBufferSize);
+  if (theWriteBufferUnaligned)
+    ndbd_free(theWriteBufferUnaligned, theWriteBufferSize);
 }
 
 

storage/ndb/src/kernel/blocks/ndbfs/AsyncFile.hpp

@@ -234,9 +234,13 @@ private:
   bool   theStartFlag;
   int theWriteBufferSize;
   char* theWriteBuffer;
+  void* theWriteBufferUnaligned;
   
   size_t m_write_wo_sync;  // Writes wo/ sync
   size_t m_auto_sync_freq; // Auto sync freq in bytes
+
+  int check_odirect_read(Uint32 flags, int&new_flags, int mode);
+  int check_odirect_write(Uint32 flags, int&new_flags, int mode);
 public:
   SimulatedBlock& m_fs;
   Ptr<GlobalPage> m_page_ptr;

storage/ndb/src/kernel/blocks/ndbfs/Ndbfs.cpp

@@ -652,7 +652,7 @@ AsyncFile*
 Ndbfs::createAsyncFile(){
 
   // Check limit of open files
-  if (m_maxFiles !=0 && theFiles.size()+1 ==  m_maxFiles) {
+  if (m_maxFiles !=0 && theFiles.size() ==  m_maxFiles) {
     // Print info about all open files
     for (unsigned i = 0; i < theFiles.size(); i++){
       AsyncFile* file = theFiles[i];

storage/ndb/src/kernel/blocks/pgman.cpp

@@ -668,6 +668,7 @@ Pgman::lirs_reference(Ptr<Page_entry> ptr)
         jam();
         move_cleanup_ptr(ptr);
         pl_queue.remove(ptr);
+        state &= ~ Page_entry::ONQUEUE;
       }
       if (state & Page_entry::BOUND)
       {
@@ -698,6 +699,12 @@ Pgman::lirs_reference(Ptr<Page_entry> ptr)
     pl_stack.add(ptr);
     state |= Page_entry::ONSTACK;
     state |= Page_entry::HOT;
+    // it could be on queue already
+    if (state & Page_entry::ONQUEUE) {
+      jam();
+      pl_queue.remove(ptr);
+      state &= ~Page_entry::ONQUEUE;
+    }
   }
 
   set_page_state(ptr, state);

storage/ndb/src/kernel/blocks/restore.cpp

@@ -557,6 +557,9 @@ Restore::restore_next(Signal* signal, FilePtr file_ptr)
       case BackupFormat::GCP_ENTRY:
 	parse_gcp_entry(signal, file_ptr, data, len);
 	break;
+      case BackupFormat::EMPTY_ENTRY:
+        // skip
+        break;
       case 0x4e444242: // 'NDBB'
 	if (check_file_version(signal, ntohl(* (data+2))) == 0)
 	{

storage/ndb/src/kernel/vm/Configuration.cpp

@@ -443,6 +443,11 @@ Configuration::setupConfiguration(){
 	      "TimeBetweenWatchDogCheck missing");
   }
 
+  if(iter.get(CFG_DB_WATCHDOG_INTERVAL_INITIAL, &_timeBetweenWatchDogCheckInitial)){
+    ERROR_SET(fatal, NDBD_EXIT_INVALID_CONFIG, "Invalid configuration fetched", 
+	      "TimeBetweenWatchDogCheckInitial missing");
+  }
+
   /**
    * Get paths
    */  
@@ -462,9 +467,12 @@ Configuration::setupConfiguration(){
    * Create the watch dog thread
    */
   { 
-    Uint32 t = _timeBetweenWatchDogCheck;
+    if (_timeBetweenWatchDogCheckInitial < _timeBetweenWatchDogCheck)
+      _timeBetweenWatchDogCheckInitial = _timeBetweenWatchDogCheck;
+
+    Uint32 t = _timeBetweenWatchDogCheckInitial;
     t = globalEmulatorData.theWatchDog ->setCheckInterval(t);
-    _timeBetweenWatchDogCheck = t;
+    _timeBetweenWatchDogCheckInitial = t;
   }
   
   ConfigValues* cf = ConfigValuesFactory::extractCurrentSection(iter.m_config);

storage/ndb/src/kernel/vm/Configuration.hpp

@@ -84,6 +84,7 @@ private:
   Uint32 _maxErrorLogs;
   Uint32 _lockPagesInMainMemory;
   Uint32 _timeBetweenWatchDogCheck;
+  Uint32 _timeBetweenWatchDogCheckInitial;
 
   ndb_mgm_configuration * m_ownConfig;
   ndb_mgm_configuration * m_clusterConfig;

storage/ndb/src/kernel/vm/SimulatedBlock.cpp

@@ -19,6 +19,7 @@
 #include <NdbOut.hpp>
 #include <GlobalData.hpp>
 #include <Emulator.hpp>
+#include <WatchDog.hpp>
 #include <ErrorHandlingMacros.hpp>
 #include <TimeQueue.hpp>
 #include <TransporterRegistry.hpp>
@@ -38,6 +39,9 @@
 #include <AttributeDescriptor.hpp>
 #include <NdbSqlUtil.hpp>
 
+#include <EventLogger.hpp>
+extern EventLogger g_eventLogger;
+
 #define ljamEntry() jamEntryLine(30000 + __LINE__)
 #define ljam() jamLine(30000 + __LINE__)
 
@@ -655,14 +659,20 @@ SimulatedBlock::getBatSize(Uint16 blockNo){
   return sb->theBATSize;
 }
 
+void* SimulatedBlock::allocRecord(const char * type, size_t s, size_t n, bool clear, Uint32 paramId)
+{
+  return allocRecordAligned(type, s, n, 0, 0, clear, paramId);
+}
+
 void* 
-SimulatedBlock::allocRecord(const char * type, size_t s, size_t n, bool clear, Uint32 paramId) 
+SimulatedBlock::allocRecordAligned(const char * type, size_t s, size_t n, void **unaligned_buffer, Uint32 align, bool clear, Uint32 paramId)
 {
 
   void * p = NULL;
-  size_t size = n*s;
+  Uint32 over_alloc = unaligned_buffer ? (align - 1) : 0;
-  Uint64 real_size = (Uint64)((Uint64)n)*((Uint64)s);
+  size_t size = n*s + over_alloc;
-  refresh_watch_dog(); 
+  Uint64 real_size = (Uint64)((Uint64)n)*((Uint64)s) + over_alloc;
+  refresh_watch_dog(9);
   if (real_size > 0){
 #ifdef VM_TRACE_MEM
     ndbout_c("%s::allocRecord(%s, %u, %u) = %llu bytes", 
@@ -696,14 +706,24 @@ SimulatedBlock::allocRecord(const char * type, size_t s, size_t n, bool clear, U
       char * ptr = (char*)p;
       const Uint32 chunk = 128 * 1024;
       while(size > chunk){
-	refresh_watch_dog(); 
+	refresh_watch_dog(9);
 	memset(ptr, 0, chunk);
 	ptr += chunk;
 	size -= chunk;
       }
-      refresh_watch_dog(); 
+      refresh_watch_dog(9);
       memset(ptr, 0, size);
     }
+    if (unaligned_buffer)
+    {
+      *unaligned_buffer = p;
+      p = (void *)(((UintPtr)p + over_alloc) & ~(UintPtr)(over_alloc));
+#ifdef VM_TRACE
+      g_eventLogger.info("'%s' (%u) %llu %llu, alignment correction %u bytes",
+                         type, align, (Uint64)p, (Uint64)p+n*s,
+                         (Uint32)((UintPtr)p - (UintPtr)*unaligned_buffer));
+#endif
+    }
   }
   return p;
 }
@@ -720,9 +740,16 @@ SimulatedBlock::deallocRecord(void ** ptr,
 }
 
 void
-SimulatedBlock::refresh_watch_dog()
+SimulatedBlock::refresh_watch_dog(Uint32 place)
 {
-  globalData.incrementWatchDogCounter(1);
+  globalData.incrementWatchDogCounter(place);
+}
+
+void
+SimulatedBlock::update_watch_dog_timer(Uint32 interval)
+{
+  extern EmulatorData globalEmulatorData;
+  globalEmulatorData.theWatchDog->setCheckInterval(interval);
 }
 
 void

storage/ndb/src/kernel/vm/SimulatedBlock.hpp

@@ -334,7 +334,8 @@ protected:
    * Refresh Watch Dog in initialising code
    *
    */
-  void refresh_watch_dog();
+  void refresh_watch_dog(Uint32 place = 1);
+  void update_watch_dog_timer(Uint32 interval);
 
   /**
    * Prog error
@@ -377,6 +378,7 @@ protected:
    *
    */
   void* allocRecord(const char * type, size_t s, size_t n, bool clear = true, Uint32 paramId = 0);
+  void* allocRecordAligned(const char * type, size_t s, size_t n, void **unaligned_buffer, Uint32 align = NDB_O_DIRECT_WRITE_ALIGNMENT, bool clear = true, Uint32 paramId = 0);
   
   /**
    * Deallocate record

storage/ndb/src/kernel/vm/WatchDog.cpp

@@ -16,6 +16,7 @@
 
 #include <ndb_global.h>
 #include <my_pthread.h>
+#include <sys/times.h>
 
 #include "WatchDog.hpp"
 #include "GlobalData.hpp"
@@ -24,6 +25,8 @@
 #include <ErrorHandlingMacros.hpp>
 #include <EventLogger.hpp>
 
+#include <NdbTick.h>
+
 extern EventLogger g_eventLogger;
 
 extern "C" 
@@ -71,66 +74,115 @@ WatchDog::doStop(){
   }
 }
 
+const char *get_action(Uint32 IPValue)
+{
+  const char *action;
+  switch (IPValue) {
+  case 1:
+    action = "Job Handling";
+    break;
+  case 2:
+    action = "Scanning Timers";
+    break;
+  case 3:
+    action = "External I/O";
+    break;
+  case 4:
+    action = "Print Job Buffers at crash";
+    break;
+  case 5:
+    action = "Checking connections";
+    break;
+  case 6:
+    action = "Performing Send";
+    break;
+  case 7:
+    action = "Polling for Receive";
+    break;
+  case 8:
+    action = "Performing Receive";
+    break;
+  case 9:
+    action = "Allocating memory";
+    break;
+  default:
+    action = "Unknown place";
+    break;
+  }//switch
+  return action;
+}
+
 void 
-WatchDog::run(){
+WatchDog::run()
-  unsigned int anIPValue;
+{
-  unsigned int alerts = 0;
+  unsigned int anIPValue, sleep_time;
   unsigned int oldIPValue = 0;
+  unsigned int theIntervalCheck = theInterval;
+  struct MicroSecondTimer start_time, last_time, now;
+  NdbTick_getMicroTimer(&start_time);
+  last_time = start_time;
 
   // WatchDog for the single threaded NDB
-  while(!theStop){
+  while (!theStop)
-    Uint32 tmp  = theInterval / 500;
+  {
-    tmp= (tmp ? tmp : 1);
+    sleep_time= 100;
-    
-    while(!theStop && tmp > 0){
-      NdbSleep_MilliSleep(500);
-      tmp--;
-    }
 
+    NdbSleep_MilliSleep(sleep_time);
     if(theStop)
       break;
 
+    NdbTick_getMicroTimer(&now);
+    if (NdbTick_getMicrosPassed(last_time, now)/1000 > sleep_time*2)
+    {
+      struct tms my_tms;
+      times(&my_tms);
+      g_eventLogger.info("Watchdog: User time: %llu  System time: %llu",
+                         (Uint64)my_tms.tms_utime,
+                         (Uint64)my_tms.tms_stime);
+      g_eventLogger.warning("Watchdog: Warning overslept %u ms, expected %u ms.",
+                            NdbTick_getMicrosPassed(last_time, now)/1000,
+                            sleep_time);
+    }
+    last_time = now;
+
     // Verify that the IP thread is not stuck in a loop
     anIPValue = *theIPValue;
-    if(anIPValue != 0) {
+    if (anIPValue != 0)
+    {
       oldIPValue = anIPValue;
       globalData.incrementWatchDogCounter(0);
-      alerts = 0;
+      NdbTick_getMicroTimer(&start_time);
-    } else {
+      theIntervalCheck = theInterval;
-      const char *last_stuck_action;
+    }
-      alerts++;
+    else
-      switch (oldIPValue) {
+    {
-      case 1:
+      int warn = 1;
-        last_stuck_action = "Job Handling";
+      Uint32 elapsed = NdbTick_getMicrosPassed(start_time, now)/1000;
-        break;
+      /*
-      case 2:
+        oldIPValue == 9 indicates malloc going on, this can take some time
-        last_stuck_action = "Scanning Timers";
+        so only warn if we pass the watchdog interval
-        break;
+      */
-      case 3:
+      if (oldIPValue == 9)
-        last_stuck_action = "External I/O";
+        if (elapsed < theIntervalCheck)
-        break;
+          warn = 0;
-      case 4:
+        else
-        last_stuck_action = "Print Job Buffers at crash";
+          theIntervalCheck += theInterval;
-        break;
+
-      case 5:
+      if (warn)
-        last_stuck_action = "Checking connections";
+      {
-        break;
+        const char *last_stuck_action = get_action(oldIPValue);
-      case 6:
+        g_eventLogger.warning("Ndb kernel is stuck in: %s", last_stuck_action);
-        last_stuck_action = "Performing Send";
+        {
-        break;
+          struct tms my_tms;
-      case 7:
+          times(&my_tms);
-        last_stuck_action = "Polling for Receive";
+          g_eventLogger.info("Watchdog: User time: %llu  System time: %llu",
-        break;
+                             (Uint64)my_tms.tms_utime,
-      case 8:
+                             (Uint64)my_tms.tms_stime);
-        last_stuck_action = "Performing Receive";
+        }
-        break;
+        if (elapsed > 3 * theInterval)
-      default:
+        {
-        last_stuck_action = "Unknown place";
+          shutdownSystem(last_stuck_action);
-        break;
+        }
-      }//switch
-      g_eventLogger.warning("Ndb kernel is stuck in: %s", last_stuck_action);
-      if(alerts == 3){
-	shutdownSystem(last_stuck_action);
       }
     }
   }

storage/ndb/src/mgmapi/mgmapi.cpp

@@ -524,7 +524,7 @@ ndb_mgm_connect(NdbMgmHandle handle, int no_retries,
   NDB_SOCKET_TYPE sockfd= NDB_INVALID_SOCKET;
   Uint32 i;
   SocketClient s(0, 0);
-  s.set_connect_timeout(handle->timeout);
+  s.set_connect_timeout((handle->timeout+999)/1000);
   if (!s.init())
   {
     fprintf(handle->errstream, 

storage/ndb/src/mgmclient/CommandInterpreter.cpp

@@ -18,6 +18,7 @@
 #include <Vector.hpp>
 #include <mgmapi.h>
 #include <util/BaseString.hpp>
+#include <ndbd_exit_codes.h>
 
 class MgmtSrvr;
 
@@ -704,6 +705,113 @@ CommandInterpreter::printError()
   }
 }
 
+/*
+ * print log event from mgmsrv to console screen
+ */
+static void
+printLogEvent(struct ndb_logevent* event)
+{
+  switch (event->type) {
+    /** 
+     * NDB_MGM_EVENT_CATEGORY_BACKUP
+     */
+    case NDB_LE_BackupStarted:
+      ndbout_c("Backup %d started from node %d", 
+               event->BackupStarted.backup_id, event->BackupStarted.starting_node);
+      break;
+    case NDB_LE_BackupFailedToStart:
+      ndbout_c("Backup request from %d failed to start. Error: %d", 
+               event->BackupFailedToStart.starting_node, event->BackupFailedToStart.error);
+      break;
+    case NDB_LE_BackupCompleted:
+      ndbout_c("Backup %u started from node %u completed\n" 
+               " StartGCP: %u StopGCP: %u\n" 
+               " #Records: %u #LogRecords: %u\n" 
+               " Data: %u bytes Log: %u bytes",
+               event->BackupCompleted.backup_id, event->BackupCompleted.starting_node,
+               event->BackupCompleted.start_gci, event->BackupCompleted.stop_gci,
+               event->BackupCompleted.n_records, event->BackupCompleted.n_log_records,
+               event->BackupCompleted.n_bytes, event->BackupCompleted.n_log_bytes);
+      break;
+    case NDB_LE_BackupAborted:
+      ndbout_c("Backup %d started from %d has been aborted. Error: %d",
+               event->BackupAborted.backup_id, event->BackupAborted.starting_node, 
+               event->BackupAborted.error);
+      break;
+    /** 
+     * NDB_MGM_EVENT_CATEGORY_STARTUP
+     */ 
+    case NDB_LE_NDBStartStarted:
+      ndbout_c("Start initiated (version %d.%d.%d)", 
+               getMajor(event->NDBStartStarted.version),
+               getMinor(event->NDBStartStarted.version),
+               getBuild(event->NDBStartStarted.version));
+      break;
+    case NDB_LE_NDBStartCompleted:
+      ndbout_c("Started (version %d.%d.%d)", 
+               getMajor(event->NDBStartCompleted.version),
+               getMinor(event->NDBStartCompleted.version),
+               getBuild(event->NDBStartCompleted.version));
+      break;
+    case NDB_LE_NDBStopStarted:
+      ndbout_c("%s shutdown initiated", 
+               (event->NDBStopStarted.stoptype == 1 ? "Cluster" : "Node"));
+      break;
+    case NDB_LE_NDBStopCompleted:
+      {
+        BaseString action_str("");
+        BaseString signum_str("");
+        getRestartAction(event->NDBStopCompleted.action, action_str);
+        if (event->NDBStopCompleted.signum)
+          signum_str.appfmt(" Initiated by signal %d.", 
+                            event->NDBStopCompleted.signum);
+        ndbout_c("Node shutdown completed%s.%s", 
+                 action_str.c_str(), 
+                 signum_str.c_str());
+      }
+      break;
+    case NDB_LE_NDBStopForced:
+      {
+        BaseString action_str("");
+        BaseString reason_str("");
+        BaseString sphase_str("");
+        int signum = event->NDBStopForced.signum;
+        int error = event->NDBStopForced.error; 
+        int sphase = event->NDBStopForced.sphase; 
+        int extra = event->NDBStopForced.extra; 
+        getRestartAction(event->NDBStopForced.action, action_str);
+        if (signum)
+          reason_str.appfmt(" Initiated by signal %d.", signum);
+        if (error)
+        {
+          ndbd_exit_classification cl;
+          ndbd_exit_status st;
+          const char *msg = ndbd_exit_message(error, &cl);
+          const char *cl_msg = ndbd_exit_classification_message(cl, &st);
+          const char *st_msg = ndbd_exit_status_message(st);
+          reason_str.appfmt(" Caused by error %d: \'%s(%s). %s\'.", 
+                            error, msg, cl_msg, st_msg);
+          if (extra != 0)
+            reason_str.appfmt(" (extra info %d)", extra);
+        }
+        if (sphase < 255)
+          sphase_str.appfmt(" Occured during startphase %u.", sphase);
+        ndbout_c("Forced node shutdown completed%s.%s%s", 
+                 action_str.c_str(), sphase_str.c_str(), 
+                 reason_str.c_str());
+      }
+      break;
+    case NDB_LE_NDBStopAborted:
+      ndbout_c("Node shutdown aborted");
+      break;
+    /** 
+     * default nothing to print
+     */ 
+    default:
+      break;
+  }
+}
+
 //*****************************************************************************
 //*****************************************************************************
 
@@ -720,30 +828,21 @@ event_thread_run(void* p)
   int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_BACKUP,
 		   1, NDB_MGM_EVENT_CATEGORY_STARTUP,
 		   0 };
-  int fd = ndb_mgm_listen_event(handle, filter);
+
-  if (fd != NDB_INVALID_SOCKET)
+  NdbLogEventHandle log_handle= NULL;
+  struct ndb_logevent log_event;
+
+  log_handle= ndb_mgm_create_logevent_handle(handle, filter);
+  if (log_handle) 
   {
     do_event_thread= 1;
-    char *tmp= 0;
-    char buf[1024];
     do {
-      SocketInputStream in(fd,2000);
+      if (ndb_logevent_get_next(log_handle, &log_event, 2000) <= 0)
-      if((tmp = in.gets(buf, sizeof(buf))))
+        continue;
-      {
+      Guard g(printmutex);
-	const char ping_token[]= "<PING>";
+      printLogEvent(&log_event);
-	if (memcmp(ping_token,tmp,sizeof(ping_token)-1))
-	  if(tmp && strlen(tmp))
-          {
-            Guard g(printmutex);
-            ndbout << tmp;
-          }
-      }
-      else if(in.timedout() && ndb_mgm_check_connection(handle)<0)
-      {
-        break;
-      }
     } while(do_event_thread);
-    NDB_CLOSE_SOCKET(fd);
+    ndb_mgm_destroy_logevent_handle(&log_handle);
   }
   else
   {
@@ -1007,6 +1106,7 @@ CommandInterpreter::execute_impl(const char *_line, bool interactive)
     DBUG_RETURN(true);
   }                
   else if(strcasecmp(firstToken, "ENTER") == 0 &&
+	  allAfterFirstToken != NULL &&
 	  allAfterFirstToken != NULL &&
 	  strncasecmp(allAfterFirstToken, "SINGLE USER MODE ", 
 		  sizeof("SINGLE USER MODE") - 1) == 0){
@@ -2476,7 +2576,6 @@ CommandInterpreter::executeStartBackup(char* parameters, bool interactive)
 {
   struct ndb_mgm_reply reply;
   unsigned int backupId;
-  int fd = -1;
 
   Vector<BaseString> args;
   {
@@ -2494,8 +2593,6 @@ CommandInterpreter::executeStartBackup(char* parameters, bool interactive)
   if (sz == 2 && args[1] == "NOWAIT")
   {
     flags = 0;
-    result = ndb_mgm_start_backup(m_mgmsrv, 0, &backupId, &reply);
-    goto END_BACKUP;
   }
   else if (sz == 1 || (sz == 3 && args[1] == "WAIT" && args[2] == "COMPLETED"))
   {
@@ -2513,62 +2610,74 @@ CommandInterpreter::executeStartBackup(char* parameters, bool interactive)
     return -1;
   }
 
-  /**
+  NdbLogEventHandle log_handle= NULL;
-   * If interactive...event listner is already running
+  struct ndb_logevent log_event;
-   */
   if (flags == 2 && !interactive)
   {
     int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_BACKUP, 0, 0 };
-    fd = ndb_mgm_listen_event(m_mgmsrv, filter);
+    log_handle = ndb_mgm_create_logevent_handle(m_mgmsrv, filter);
-    if (fd < 0)
+    if (!log_handle)
     {
       ndbout << "Initializing start of backup failed" << endl;
       printError();
-      return fd;
+      return -1;
     }
   }
   result = ndb_mgm_start_backup(m_mgmsrv, flags, &backupId, &reply);
 
-END_BACKUP:
   if (result != 0) {
     ndbout << "Backup failed" << endl;
     printError();
 
-    if (fd >= 0) 
+    if (log_handle) 
-      close(fd);
+      ndb_mgm_destroy_logevent_handle(&log_handle);
     return result;
   }
 
-  if (fd >= 0)
+  /**
+   * If interactive, event listner thread is already running
+   */
+  if (log_handle && !interactive)
   {
-    char *tmp;
+    int count = 0;
-    char buf[1024];
+    int retry = 0;
-    {
-      SocketInputStream in(fd);
-      int count = 0;
-      do {
-	tmp = in.gets(buf, 1024);
-	if(tmp)
-	{
-	  ndbout << tmp;
-	  unsigned int id;
-	  if(sscanf(tmp, "%*[^:]: Backup %d ", &id) == 1 && id == backupId){
-	    count++;
-	  }
-	}
-      } while(count < 2);
-    }
-    
-    SocketInputStream in(fd, 10);
     do {
-      tmp = in.gets(buf, 1024);
+      if (ndb_logevent_get_next(log_handle, &log_event, 60000) > 0)
-      if(tmp && tmp[0] != 0)
       {
-	ndbout << tmp;
+        int print = 0;
+        switch (log_event.type) {
+          case NDB_LE_BackupStarted:
+            if (log_event.BackupStarted.backup_id == backupId)
+              print = 1;
+            break;
+          case NDB_LE_BackupCompleted:
+            if (log_event.BackupCompleted.backup_id == backupId)
+              print = 1;
+            break;
+          case NDB_LE_BackupAborted:
+            if (log_event.BackupAborted.backup_id == backupId)
+              print = 1;
+            break;
+          default:
+            break;
+        }
+        if (print)
+        {
+          Guard g(m_print_mutex);
+          printLogEvent(&log_event);
+          count++;
+        }
       }
-    } while(tmp && tmp[0] != 0);
+      else
+      {
+        retry++;
+      }
+    } while(count < 2 && retry < 3);
 
-    close(fd);
+    if (retry >= 3)
+      ndbout << "get backup event failed for " << retry << " times" << endl;
+
+    ndb_mgm_destroy_logevent_handle(&log_handle);
   }
 
   return 0;

storage/ndb/src/mgmclient/Makefile.am

@@ -21,7 +21,8 @@ libndbmgmclient_la_LIBADD = ../mgmapi/libmgmapi.la \
 			    ../common/logger/liblogger.la \
 			    ../common/portlib/libportlib.la \
 			    ../common/util/libgeneral.la \
-			    ../common/portlib/libportlib.la
+			    ../common/portlib/libportlib.la \
+			    ../common/debugger/libtrace.la
 
 
 ndb_mgm_SOURCES = main.cpp

storage/ndb/src/mgmclient/main.cpp

@@ -23,6 +23,8 @@ extern "C" {
 #elif !defined(__NETWARE__)
 #include <readline/readline.h>
 extern "C" int add_history(const char *command); /* From readline directory */
+extern "C" int read_history(const char *command);
+extern "C" int write_history(const char *command);
 #define HAVE_READLINE
 #endif
 }
@@ -155,10 +157,35 @@ int main(int argc, char** argv){
   signal(SIGPIPE, handler);
   com = new Ndb_mgmclient(opt_connect_str,1);
   int ret= 0;
+  BaseString histfile;
   if (!opt_execute_str)
   {
+#ifdef HAVE_READLINE
+    char *histfile_env= getenv("NDB_MGM_HISTFILE");
+    if (histfile_env)
+      histfile.assign(histfile_env,strlen(histfile_env));
+    else if(getenv("HOME"))
+    {
+      histfile.assign(getenv("HOME"),strlen(getenv("HOME")));
+      histfile.append("/.ndb_mgm_history");
+    }
+    if (histfile.length())
+      read_history(histfile.c_str());
+#endif
+
     ndbout << "-- NDB Cluster -- Management Client --" << endl;
     while(read_and_execute(_try_reconnect));
+
+#ifdef HAVE_READLINE
+    if (histfile.length())
+    {
+      BaseString histfile_tmp;
+      histfile_tmp.assign(histfile);
+      histfile_tmp.append(".TMP");
+      if(!write_history(histfile_tmp.c_str()))
+        my_rename(histfile_tmp.c_str(), histfile.c_str(), MYF(MY_WME));
+    }
+#endif
   }
   else
   {

storage/ndb/src/mgmsrv/ConfigInfo.cpp

@@ -579,6 +579,18 @@ const ConfigInfo::ParamInfo ConfigInfo::m_ParamInfo[] = {
     "70",
     STR_VALUE(MAX_INT_RNIL) },
 
+  {
+    CFG_DB_WATCHDOG_INTERVAL_INITIAL,
+    "TimeBetweenWatchDogCheckInitial",
+    DB_TOKEN,
+    "Time between execution checks inside a database node in the early start phases when memory is allocated",
+    ConfigInfo::CI_USED,
+    true,
+    ConfigInfo::CI_INT,
+    "6000",
+    "70",
+    STR_VALUE(MAX_INT_RNIL) },
+
   {
     CFG_DB_STOP_ON_ERROR,
     "StopOnError",
@@ -879,6 +891,18 @@ const ConfigInfo::ParamInfo ConfigInfo::m_ParamInfo[] = {
     "3",
     STR_VALUE(MAX_INT_RNIL) },
 
+  {
+    CFG_DB_REDOLOG_FILE_SIZE,
+    "FragmentLogFileSize",
+    DB_TOKEN,
+    "Size of each Redo log file",
+    ConfigInfo::CI_USED,
+    false,
+    ConfigInfo::CI_INT,
+    "16M",
+    "4M",
+    "1G" },
+
   {
     CFG_DB_MAX_OPEN_FILES,
     "MaxNoOfOpenFiles",
@@ -1297,6 +1321,18 @@ const ConfigInfo::ParamInfo ConfigInfo::m_ParamInfo[] = {
     "0",
     STR_VALUE(MAX_INT_RNIL) },
 
+  { 
+    CFG_DB_MAX_ALLOCATE,
+    "MaxAllocate",
+    DB_TOKEN,
+    "Maximum size of allocation to use when allocating memory for tables",
+    ConfigInfo::CI_USED,
+    false,
+    ConfigInfo::CI_INT,
+    "32M",
+    "1M",
+    "1G" },
+
   { 
     CFG_DB_MEMREPORT_FREQUENCY,
     "MemReportFrequency",
@@ -1309,6 +1345,18 @@ const ConfigInfo::ParamInfo ConfigInfo::m_ParamInfo[] = {
     "0",
     STR_VALUE(MAX_INT_RNIL) },
   
+  {
+    CFG_DB_O_DIRECT,
+    "ODirect",
+    DB_TOKEN,
+    "Use O_DIRECT file write/read when possible",
+    ConfigInfo::CI_USED,
+    true,
+    ConfigInfo::CI_BOOL,
+    "false",
+    "false",
+    "true"},
+
   /***************************************************************************
    * API
    ***************************************************************************/

storage/ndb/src/mgmsrv/MgmtSrvr.cpp

@@ -18,6 +18,7 @@
 
 #include "MgmtSrvr.hpp"
 #include "MgmtErrorReporter.hpp"
+#include "ndb_mgmd_error.h"
 #include <ConfigRetriever.hpp>
 
 #include <NdbOut.hpp>
@@ -239,13 +240,6 @@ MgmtSrvr::stopEventLog()
   // Nothing yet
 }
 
-class ErrorItem 
-{
-public:
-  int _errorCode;
-  const char * _errorText;
-};
-
 bool
 MgmtSrvr::setEventLogFilter(int severity, int enable)
 {
@@ -268,62 +262,6 @@ MgmtSrvr::isEventLogFilterEnabled(int severity)
   return g_eventLogger.isEnable((Logger::LoggerLevel)severity);
 }
 
-static ErrorItem errorTable[] = 
-{
-  {MgmtSrvr::NO_CONTACT_WITH_PROCESS, "No contact with the process (dead ?)."},
-  {MgmtSrvr::PROCESS_NOT_CONFIGURED, "The process is not configured."},
-  {MgmtSrvr::WRONG_PROCESS_TYPE, 
-   "The process has wrong type. Expected a DB process."},
-  {MgmtSrvr::COULD_NOT_ALLOCATE_MEMORY, "Could not allocate memory."},
-  {MgmtSrvr::SEND_OR_RECEIVE_FAILED, "Send to process or receive failed."},
-  {MgmtSrvr::INVALID_LEVEL, "Invalid level. Should be between 1 and 30."},
-  {MgmtSrvr::INVALID_ERROR_NUMBER, "Invalid error number. Should be >= 0."},
-  {MgmtSrvr::INVALID_TRACE_NUMBER, "Invalid trace number."},
-  {MgmtSrvr::NOT_IMPLEMENTED, "Not implemented."},
-  {MgmtSrvr::INVALID_BLOCK_NAME, "Invalid block name"},
-
-  {MgmtSrvr::CONFIG_PARAM_NOT_EXIST, 
-   "The configuration parameter does not exist for the process type."},
-  {MgmtSrvr::CONFIG_PARAM_NOT_UPDATEABLE, 
-   "The configuration parameter is not possible to update."},
-  {MgmtSrvr::VALUE_WRONG_FORMAT_INT_EXPECTED, 
-   "Incorrect value. Expected integer."},
-  {MgmtSrvr::VALUE_TOO_LOW, "Value is too low."},
-  {MgmtSrvr::VALUE_TOO_HIGH, "Value is too high."},
-  {MgmtSrvr::VALUE_WRONG_FORMAT_BOOL_EXPECTED, 
-   "Incorrect value. Expected TRUE or FALSE."},
-
-  {MgmtSrvr::CONFIG_FILE_OPEN_WRITE_ERROR, 
-   "Could not open configuration file for writing."},
-  {MgmtSrvr::CONFIG_FILE_OPEN_READ_ERROR, 
-   "Could not open configuration file for reading."},
-  {MgmtSrvr::CONFIG_FILE_WRITE_ERROR, 
-   "Write error when writing configuration file."},
-  {MgmtSrvr::CONFIG_FILE_READ_ERROR, 
-   "Read error when reading configuration file."},
-  {MgmtSrvr::CONFIG_FILE_CLOSE_ERROR, "Could not close configuration file."},
-
-  {MgmtSrvr::CONFIG_CHANGE_REFUSED_BY_RECEIVER, 
-   "The change was refused by the receiving process."},
-  {MgmtSrvr::COULD_NOT_SYNC_CONFIG_CHANGE_AGAINST_PHYSICAL_MEDIUM, 
-   "The change could not be synced against physical medium."},
-  {MgmtSrvr::CONFIG_FILE_CHECKSUM_ERROR, 
-   "The config file is corrupt. Checksum error."},
-  {MgmtSrvr::NOT_POSSIBLE_TO_SEND_CONFIG_UPDATE_TO_PROCESS_TYPE, 
-   "It is not possible to send an update of a configuration variable "
-   "to this kind of process."},
-  {MgmtSrvr::NODE_SHUTDOWN_IN_PROGESS, "Node shutdown in progress" },
-  {MgmtSrvr::SYSTEM_SHUTDOWN_IN_PROGRESS, "System shutdown in progress" },
-  {MgmtSrvr::NODE_SHUTDOWN_WOULD_CAUSE_SYSTEM_CRASH,
-   "Node shutdown would cause system crash" },
-  {MgmtSrvr::UNSUPPORTED_NODE_SHUTDOWN,
-   "Unsupported multi node shutdown. Abort option required." },
-  {MgmtSrvr::NODE_NOT_API_NODE, "The specified node is not an API node." },
-  {MgmtSrvr::OPERATION_NOT_ALLOWED_START_STOP, 
-   "Operation not allowed while nodes are starting or stopping."},
-  {MgmtSrvr::NO_CONTACT_WITH_DB_NODES, "No contact with database nodes" }
-};
-
 int MgmtSrvr::translateStopRef(Uint32 errCode)
 {
   switch(errCode){
@@ -343,8 +281,6 @@ int MgmtSrvr::translateStopRef(Uint32 errCode)
   return 4999;
 }
 
-static int noOfErrorCodes = sizeof(errorTable) / sizeof(ErrorItem);
-
 int 
 MgmtSrvr::getNodeCount(enum ndb_mgm_node_type type) const 
 {
@@ -1969,18 +1905,8 @@ MgmtSrvr::dumpState(int nodeId, const Uint32 args[], Uint32 no)
 
 const char* MgmtSrvr::getErrorText(int errorCode, char *buf, int buf_sz)
 {
-
-  for (int i = 0; i < noOfErrorCodes; ++i) {
-    if (errorCode == errorTable[i]._errorCode) {
-      BaseString::snprintf(buf, buf_sz, errorTable[i]._errorText);
-      buf[buf_sz-1]= 0;
-      return buf;
-    }
-  }
-
   ndb_error_string(errorCode, buf, buf_sz);
   buf[buf_sz-1]= 0;
-
   return buf;
 }
 

storage/ndb/src/mgmsrv/MgmtSrvr.hpp

@@ -148,45 +148,6 @@ public:
    */
   bool isEventLogFilterEnabled(int severity);
 
-  STATIC_CONST( NO_CONTACT_WITH_PROCESS = 5000 );
-  STATIC_CONST( PROCESS_NOT_CONFIGURED = 5001 );
-  STATIC_CONST( WRONG_PROCESS_TYPE = 5002 );
-  STATIC_CONST( COULD_NOT_ALLOCATE_MEMORY = 5003 );
-  STATIC_CONST( SEND_OR_RECEIVE_FAILED = 5005 );
-  STATIC_CONST( INVALID_LEVEL = 5006 );
-  STATIC_CONST( INVALID_ERROR_NUMBER = 5007 );
-  STATIC_CONST( INVALID_TRACE_NUMBER = 5008 );
-  STATIC_CONST( NOT_IMPLEMENTED = 5009 );
-  STATIC_CONST( INVALID_BLOCK_NAME = 5010 );
-
-  STATIC_CONST( CONFIG_PARAM_NOT_EXIST = 5011 );
-  STATIC_CONST( CONFIG_PARAM_NOT_UPDATEABLE = 5012 );
-  STATIC_CONST( VALUE_WRONG_FORMAT_INT_EXPECTED = 5013 );
-  STATIC_CONST( VALUE_TOO_LOW = 5014 );
-  STATIC_CONST( VALUE_TOO_HIGH = 5015 );
-  STATIC_CONST( VALUE_WRONG_FORMAT_BOOL_EXPECTED = 5016 );
-
-  STATIC_CONST( CONFIG_FILE_OPEN_WRITE_ERROR = 5017 );
-  STATIC_CONST( CONFIG_FILE_OPEN_READ_ERROR = 5018 );
-  STATIC_CONST( CONFIG_FILE_WRITE_ERROR = 5019 );
-  STATIC_CONST( CONFIG_FILE_READ_ERROR = 5020 );
-  STATIC_CONST( CONFIG_FILE_CLOSE_ERROR = 5021 );
-
-  STATIC_CONST( CONFIG_CHANGE_REFUSED_BY_RECEIVER = 5022 );
-  STATIC_CONST( COULD_NOT_SYNC_CONFIG_CHANGE_AGAINST_PHYSICAL_MEDIUM = 5023 );
-  STATIC_CONST( CONFIG_FILE_CHECKSUM_ERROR = 5024 );
-  STATIC_CONST( NOT_POSSIBLE_TO_SEND_CONFIG_UPDATE_TO_PROCESS_TYPE = 5025 );
-
-  STATIC_CONST( NODE_SHUTDOWN_IN_PROGESS = 5026 );
-  STATIC_CONST( SYSTEM_SHUTDOWN_IN_PROGRESS = 5027 );
-  STATIC_CONST( NODE_SHUTDOWN_WOULD_CAUSE_SYSTEM_CRASH = 5028 );
-
-  STATIC_CONST( NO_CONTACT_WITH_DB_NODES = 5030 );
-  STATIC_CONST( UNSUPPORTED_NODE_SHUTDOWN = 5031 );
-
-  STATIC_CONST( NODE_NOT_API_NODE = 5062 );
-  STATIC_CONST( OPERATION_NOT_ALLOWED_START_STOP = 5063 );
-
   /**
    *   This enum specifies the different signal loggig modes possible to set 
    *   with the setSignalLoggingMode method.

storage/ndb/src/mgmsrv/ndb_mgmd_error.h

@@ -0,0 +1,33 @@
+/* Copyright (C) 2007 MySQL AB
+
+   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
+   the Free Software Foundation; version 2 of the License.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
+
+#ifndef NDB_MGMD_ERROR_H
+#define NDB_MGMD_ERROR_H
+
+#define NO_CONTACT_WITH_PROCESS 5000
+#define WRONG_PROCESS_TYPE 5002
+#define SEND_OR_RECEIVE_FAILED 5005
+#define INVALID_ERROR_NUMBER 5007
+#define INVALID_TRACE_NUMBER 5008
+#define INVALID_BLOCK_NAME 5010
+#define NODE_SHUTDOWN_IN_PROGESS 5026
+#define SYSTEM_SHUTDOWN_IN_PROGRESS 5027
+#define NODE_SHUTDOWN_WOULD_CAUSE_SYSTEM_CRASH 5028
+#define NO_CONTACT_WITH_DB_NODES 5030
+#define UNSUPPORTED_NODE_SHUTDOWN 5031
+#define NODE_NOT_API_NODE 5062
+#define OPERATION_NOT_ALLOWED_START_STOP 5063
+
+#endif

storage/ndb/src/ndbapi/NdbBlob.cpp

@@ -1261,6 +1261,7 @@ NdbBlob::deletePartsUnknown(Uint32 part)
         DBUG_RETURN(-1);
       }
       tOp->m_abortOption= NdbOperation::AO_IgnoreError;
+      tOp->m_noErrorPropagation = true;
       n++;
     }
     DBUG_PRINT("info", ("bat=%u", bat));
@@ -1597,6 +1598,7 @@ NdbBlob::preExecute(NdbTransaction::ExecType anExecType, bool& batch)
       }
       if (isWriteOp()) {
         tOp->m_abortOption = NdbOperation::AO_IgnoreError;
+        tOp->m_noErrorPropagation = true;
       }
       theHeadInlineReadOp = tOp;
       // execute immediately
@@ -1643,6 +1645,7 @@ NdbBlob::preExecute(NdbTransaction::ExecType anExecType, bool& batch)
       }
       if (isWriteOp()) {
         tOp->m_abortOption = NdbOperation::AO_IgnoreError;
+        tOp->m_noErrorPropagation = true;
       }
       theHeadInlineReadOp = tOp;
       // execute immediately

storage/ndb/src/ndbapi/NdbEventOperationImpl.cpp

@@ -41,6 +41,7 @@
 #include <NdbEventOperation.hpp>
 #include "NdbEventOperationImpl.hpp"
 #include <signaldata/AlterTable.hpp>
+#include "ndb_internal.hpp"
 
 #include <EventLogger.hpp>
 extern EventLogger g_eventLogger;
@@ -2838,7 +2839,7 @@ send_report:
   data[5]= apply_gci >> 32;
   data[6]= latest_gci & ~(Uint32)0;
   data[7]= latest_gci >> 32;
-  m_ndb->theImpl->send_event_report(data,8);
+  Ndb_internal::send_event_report(m_ndb, data,8);
 #ifdef VM_TRACE
   assert(m_total_alloc >= m_free_data_sz);
 #endif

storage/ndb/src/ndbapi/NdbOperation.cpp

@@ -76,7 +76,8 @@ NdbOperation::NdbOperation(Ndb* aNdb, NdbOperation::Type aType) :
   m_keyInfoGSN(GSN_KEYINFO),
   m_attrInfoGSN(GSN_ATTRINFO),
   theBlobList(NULL),
-  m_abortOption(-1)
+  m_abortOption(-1),
+  m_noErrorPropagation(false)
 {
   theReceiver.init(NdbReceiver::NDB_OPERATION, this);
   theError.code = 0;
@@ -101,7 +102,8 @@ NdbOperation::setErrorCode(int anErrorCode)
   theError.code = anErrorCode;
   theNdbCon->theErrorLine = theErrorLine;
   theNdbCon->theErrorOperation = this;
-  theNdbCon->setOperationErrorCode(anErrorCode);
+  if (!(m_abortOption == AO_IgnoreError && m_noErrorPropagation))
+    theNdbCon->setOperationErrorCode(anErrorCode);
 }
 
 /******************************************************************************
@@ -116,6 +118,7 @@ NdbOperation::setErrorCodeAbort(int anErrorCode)
   theError.code = anErrorCode;
   theNdbCon->theErrorLine = theErrorLine;
   theNdbCon->theErrorOperation = this;
+  // ignore m_noErrorPropagation
   theNdbCon->setOperationErrorCodeAbort(anErrorCode);
 }
 
@@ -161,6 +164,7 @@ NdbOperation::init(const NdbTableImpl* tab, NdbTransaction* myConnection){
   theMagicNumber        = 0xABCDEF01;
   theBlobList = NULL;
   m_abortOption = -1;
+  m_noErrorPropagation = false;
   m_no_disk_flag = 1;
 
   tSignal = theNdb->getSignal();

storage/ndb/src/ndbapi/ndb_internal.hpp

@@ -0,0 +1,26 @@
+/* Copyright (C) 2007 MySQL AB
+
+   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
+   the Free Software Foundation; version 2 of the License.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
+
+#include "NdbImpl.hpp"
+
+class Ndb_internal
+{
+private:
+  friend class NdbEventBuffer;
+  Ndb_internal() {}
+  virtual ~Ndb_internal() {}
+  static int send_event_report(Ndb *ndb, Uint32 *data, Uint32 length)
+    { return ndb->theImpl->send_event_report(data, length); }
+};

storage/ndb/src/ndbapi/ndberror.c

@@ -19,6 +19,9 @@
 #include <ndberror.h>
 #include <m_string.h>
 
+#include "../mgmsrv/ndb_mgmd_error.h"
+
+
 typedef struct ErrorBundle {
   int code;
   int mysql_code;
@@ -179,7 +182,7 @@ ErrorBundle ErrorCodes[] = {
   { 873,  DMEC, TR, "Out of attrinfo records for scan in tuple manager" },
   { 899,  DMEC, TR, "Rowid already allocated" },
   { 1217, DMEC, TR, "Out of operation records in local data manager (increase MaxNoOfLocalOperations)" },
-  { 1220, DMEC, TR, "REDO log files overloaded, consult online manual (decrease TimeBetweenLocalCheckpoints, and|or increase NoOfFragmentLogFiles)" },
+  { 1220, DMEC, TR, "REDO log files overloaded, consult online manual (increase FragmentLogFileSize)" },
   { 1222, DMEC, TR, "Out of transaction markers in LQH" },
   { 4021, DMEC, TR, "Out of Send Buffer space in NDB API" },
   { 4022, DMEC, TR, "Out of Send Buffer space in NDB API" },
@@ -619,6 +622,33 @@ ErrorBundle ErrorCodes[] = {
   { 4273, DMEC, IE, "No blob table in dict cache" },
   { 4274, DMEC, IE, "Corrupted main table PK in blob operation" },
   { 4275, DMEC, AE, "The blob method is incompatible with operation type or lock mode" },
+
+  { NO_CONTACT_WITH_PROCESS, DMEC, AE,
+    "No contact with the process (dead ?)."},
+  { WRONG_PROCESS_TYPE, DMEC, AE,
+   "The process has wrong type. Expected a DB process."},
+  { SEND_OR_RECEIVE_FAILED, DMEC, AE,
+    "Send to process or receive failed."},
+  { INVALID_ERROR_NUMBER, DMEC, AE,
+    "Invalid error number. Should be >= 0."},
+  { INVALID_TRACE_NUMBER, DMEC, AE,
+    "Invalid trace number."},
+  { INVALID_BLOCK_NAME, DMEC, AE,
+    "Invalid block name"},
+  { NODE_SHUTDOWN_IN_PROGESS, DMEC, AE,
+    "Node shutdown in progress" },
+  { SYSTEM_SHUTDOWN_IN_PROGRESS, DMEC, AE,
+    "System shutdown in progress" },
+  { NODE_SHUTDOWN_WOULD_CAUSE_SYSTEM_CRASH, DMEC, AE,
+   "Node shutdown would cause system crash" },
+  { UNSUPPORTED_NODE_SHUTDOWN, DMEC, AE,
+   "Unsupported multi node shutdown. Abort option required." },
+  { NODE_NOT_API_NODE, DMEC, AE,
+    "The specified node is not an API node." },
+  { OPERATION_NOT_ALLOWED_START_STOP, DMEC, AE,
+   "Operation not allowed while nodes are starting or stopping."},
+  { NO_CONTACT_WITH_DB_NODES, DMEC, AE,
+    "No contact with database nodes" }
 };
 
 static

storage/ndb/test/ndbapi/testMgm.cpp

@@ -212,6 +212,76 @@ int runTestApiSession(NDBT_Context* ctx, NDBT_Step* step)
   }
 }
 
+int runTestApiConnectTimeout(NDBT_Context* ctx, NDBT_Step* step)
+{
+  char *mgm= ctx->getRemoteMgm();
+  int result= NDBT_FAILED;
+  int cc= 0;
+  int mgmd_nodeid= 0;
+  ndb_mgm_reply reply;
+
+  NdbMgmHandle h;
+  h= ndb_mgm_create_handle();
+  ndb_mgm_set_connectstring(h, mgm);
+
+  ndbout << "TEST connect timeout" << endl;
+
+  ndb_mgm_set_timeout(h, 3000);
+
+  struct timeval tstart, tend;
+  int secs;
+  timerclear(&tstart);
+  timerclear(&tend);
+  gettimeofday(&tstart,NULL);
+
+  ndb_mgm_connect(h,0,0,0);
+
+  gettimeofday(&tend,NULL);
+
+  secs= tend.tv_sec - tstart.tv_sec;
+  ndbout << "Took about: " << secs <<" seconds"<<endl;
+
+  if(secs < 4)
+    result= NDBT_OK;
+  else
+    goto done;
+
+  ndb_mgm_set_connectstring(h, mgm);
+
+  ndbout << "TEST connect timeout" << endl;
+
+  ndb_mgm_destroy_handle(&h);
+
+  h= ndb_mgm_create_handle();
+  ndb_mgm_set_connectstring(h, "1.1.1.1");
+
+  ndbout << "TEST connect timeout (invalid host)" << endl;
+
+  ndb_mgm_set_timeout(h, 3000);
+
+  timerclear(&tstart);
+  timerclear(&tend);
+  gettimeofday(&tstart,NULL);
+
+  ndb_mgm_connect(h,0,0,0);
+
+  gettimeofday(&tend,NULL);
+
+  secs= tend.tv_sec - tstart.tv_sec;
+  ndbout << "Took about: " << secs <<" seconds"<<endl;
+
+  if(secs < 4)
+    result= NDBT_OK;
+  else
+    result= NDBT_FAILED;
+
+done:
+  ndb_mgm_disconnect(h);
+  ndb_mgm_destroy_handle(&h);
+
+  return result;
+}
+
 int runTestApiTimeoutBasic(NDBT_Context* ctx, NDBT_Step* step)
 {
   char *mgm= ctx->getRemoteMgm();
@@ -727,6 +797,11 @@ TESTCASE("ApiSessionFailure",
 	 "Test failures in MGMAPI session"){
   INITIALIZER(runTestApiSession);
 
+}
+TESTCASE("ApiConnectTimeout",
+	 "Connect timeout tests for MGMAPI"){
+  INITIALIZER(runTestApiConnectTimeout);
+
 }
 TESTCASE("ApiTimeoutBasic",
 	 "Basic timeout tests for MGMAPI"){

storage/ndb/test/ndbapi/testNodeRestart.cpp

@@ -1633,6 +1633,85 @@ runBug28023(NDBT_Context* ctx, NDBT_Step* step)
   return NDBT_OK;
 }
 
+
+int
+runBug28717(NDBT_Context* ctx, NDBT_Step* step)
+{
+  int result = NDBT_OK;
+  int loops = ctx->getNumLoops();
+  int records = ctx->getNumRecords();
+  Ndb* pNdb = GETNDB(step);
+  NdbRestarter res;
+
+  if (res.getNumDbNodes() < 4)
+  {
+    return NDBT_OK;
+  }
+
+  int master = res.getMasterNodeId();
+  int node0 = res.getRandomNodeOtherNodeGroup(master, rand());
+  int node1 = res.getRandomNodeSameNodeGroup(node0, rand());
+  
+  ndbout_c("master: %d node0: %d node1: %d", master, node0, node1);
+  
+  if (res.restartOneDbNode(node0, false, true, true))
+  {
+    return NDBT_FAILED;
+  }
+
+  {
+    int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_CHECKPOINT, 0 };
+    NdbLogEventHandle handle = 
+      ndb_mgm_create_logevent_handle(res.handle, filter);
+    
+
+    int dump[] = { DumpStateOrd::DihStartLcpImmediately };
+    struct ndb_logevent event;
+    
+    for (Uint32 i = 0; i<3; i++)
+    {
+      res.dumpStateOneNode(master, dump, 1);
+      while(ndb_logevent_get_next(handle, &event, 0) >= 0 &&
+            event.type != NDB_LE_LocalCheckpointStarted);
+      while(ndb_logevent_get_next(handle, &event, 0) >= 0 &&
+            event.type != NDB_LE_LocalCheckpointCompleted);
+    } 
+  }
+  
+  if (res.waitNodesNoStart(&node0, 1))
+    return NDBT_FAILED;
+  
+  int val2[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 1 };
+  
+  if (res.dumpStateOneNode(node0, val2, 2))
+    return NDBT_FAILED;
+  
+  if (res.insertErrorInNode(node0, 5010))
+    return NDBT_FAILED;
+  
+  if (res.insertErrorInNode(node1, 1001))
+    return NDBT_FAILED;
+  
+  if (res.startNodes(&node0, 1))
+    return NDBT_FAILED;
+  
+  NdbSleep_SecSleep(3);
+
+  if (res.insertErrorInNode(node1, 0))
+    return NDBT_FAILED;
+
+  if (res.waitNodesNoStart(&node0, 1))
+    return NDBT_FAILED;
+
+  if (res.startNodes(&node0, 1))
+    return NDBT_FAILED;
+
+  if (res.waitClusterStarted())
+    return NDBT_FAILED;
+  
+  return NDBT_OK;
+}
+
 NDBT_TESTSUITE(testNodeRestart);
 TESTCASE("NoLoad", 
 	 "Test that one node at a time can be stopped and then restarted "\
@@ -1993,6 +2072,9 @@ TESTCASE("Bug27466", ""){
 TESTCASE("Bug28023", ""){
   INITIALIZER(runBug28023);
 }
+TESTCASE("Bug28717", ""){
+  INITIALIZER(runBug28717);
+}
 NDBT_TESTSUITE_END(testNodeRestart);
 
 int main(int argc, const char** argv){

storage/ndb/test/run-test/conf-dl145a.cnf

@@ -21,3 +21,6 @@ BackupMemory = 64M
 MaxNoOfConcurrentScans = 100
 MaxNoOfSavedMessages= 1000
 SendBufferMemory = 2M
+NoOfFragmentLogFiles = 4
+FragmentLogFileSize = 64M
+

storage/ndb/test/run-test/daily-basic-tests.txt

@@ -567,6 +567,10 @@ max-time: 1500
 cmd: testDict
 args: -n CreateAndDrop 
 
+max-time: 1000
+cmd: testNodeRestart
+args: -n Bug28717 T1
+
 max-time: 1500
 cmd: testDict
 args: -n CreateAndDropAtRandom -l 200 T1
@@ -706,7 +710,7 @@ args: -n ExecuteAsynch T1
 
 max-time: 1000
 cmd: testNdbApi
-args: -n BugBug28443
+args: -n Bug28443
 
 #max-time: 500
 #cmd: testInterpreter
@@ -894,6 +898,10 @@ max-time: 120
 cmd: testMgm
 args: -n ApiSessionFailure T1
 
+max-time: 15
+cmd: testMgm
+args: -n ApiConnectTimeout T1
+
 max-time: 120
 cmd: testMgm
 args: -n ApiTimeoutBasic T1

storage/ndb/tools/ndb_error_reporter

@@ -62,13 +62,13 @@ foreach my $node (@nodes)
 	(($config_get_fs)?" with filesystem":"").
 	"\n\n";
     my $recurse= ($config_get_fs)?'-r ':'';
-    system 'scp '.$recurse.$config_username.config($node,'host').
+    system 'scp -p '.$recurse.$config_username.config($node,'host').
 	':'.config($node,'datadir')."/ndb_".$node."* ".
 	"$reportdir/\n";
 }
 
 print "\n\n Copying configuration file...\n\n\t$config_file\n\n";
-system "cp $config_file $reportdir/";
+system "cp -p $config_file $reportdir/";
 
 my $r = system 'bzip2 2>&1 > /dev/null  < /dev/null';
 my $outfile;

storage/ndb/tools/restore/Restore.cpp

@@ -873,13 +873,32 @@ bool RestoreDataIterator::readFragmentHeader(int & ret, Uint32 *fragmentId)
   
   debug << "RestoreDataIterator::getNextFragment" << endl;
   
-  if (buffer_read(&Header, sizeof(Header), 1) != 1){
+  while (1)
+  {
+    /* read first part of header */
+    if (buffer_read(&Header, 8, 1) != 1)
+    {
+      ret = 0;
+      return false;
+    } // if
+
+    /* skip if EMPTY_ENTRY */
+    Header.SectionType  = ntohl(Header.SectionType);
+    Header.SectionLength  = ntohl(Header.SectionLength);
+    if (Header.SectionType == BackupFormat::EMPTY_ENTRY)
+    {
+      void *tmp;
+      buffer_get_ptr(&tmp, Header.SectionLength*4-8, 1);
+      continue;
+    }
+    break;
+  }
+  /* read rest of header */
+  if (buffer_read(((char*)&Header)+8, sizeof(Header)-8, 1) != 1)
+  {
     ret = 0;
     return false;
-  } // if
+  }
-  
-  Header.SectionType  = ntohl(Header.SectionType);
-  Header.SectionLength  = ntohl(Header.SectionLength);
   Header.TableId  = ntohl(Header.TableId);
   Header.FragmentNo  = ntohl(Header.FragmentNo);
   Header.ChecksumType  = ntohl(Header.ChecksumType);

storage/ndb/tools/restore/consumer_restore.cpp

@@ -1158,19 +1158,22 @@ void BackupRestore::tuple_a(restore_callback_t *cb)
 	char * dataPtr = attr_data->string_value;
 	Uint32 length = 0;
        
-        const unsigned char * src = (const unsigned char *)dataPtr;
+        if (!attr_data->null)
-        switch(attr_desc->m_column->getType()){
+        {
-        case NdbDictionary::Column::Varchar:
+          const unsigned char * src = (const unsigned char *)dataPtr;
-        case NdbDictionary::Column::Varbinary:
+          switch(attr_desc->m_column->getType()){
-          length = src[0] + 1;
+          case NdbDictionary::Column::Varchar:
-          break;
+          case NdbDictionary::Column::Varbinary:
-        case NdbDictionary::Column::Longvarchar:
+            length = src[0] + 1;
-        case NdbDictionary::Column::Longvarbinary:
+            break;
-          length = src[0] + (src[1] << 8) + 2;
+          case NdbDictionary::Column::Longvarchar:
-          break;
+          case NdbDictionary::Column::Longvarbinary:
-        default:
+            length = src[0] + (src[1] << 8) + 2;
-          length = attr_data->size;
+            break;
-          break;
+          default:
+            length = attr_data->size;
+            break;
+          }
         }
 	if (j == 0 && tup.getTable()->have_auto_inc(i))
 	  tup.getTable()->update_max_auto_val(dataPtr,size*arraySize);