Commit 87d9388e authored by monty@donna.mysql.com's avatar monty@donna.mysql.com
Browse files

Only write full transactions to binary log 

A lot of new functions for BDB tables
Fix for DROP DATABASE on windows
Default server_id variables
parent a5c75df3

Docs/manual.texi

+154 −48
Original line number Diff line number Diff line
@@ -712,7 +712,7 @@ Solving some common problems with MySQL 
* Log Replication::             Database replication with  update log

* Backup::                      Database backups

* Update log::                  The update log

* Binary log::

* Binary log::			The binary log

* Slow query log::              Log of slow queries

* Multiple servers::            Running multiple @strong{MySQL} servers on the same machine








@@ -9117,6 +9117,32 @@ bin\mysqld-nt --remove                # remove MySQL as a service







By invoking @code{mysqld} directly.









@end itemize

















When the @code{mysqld} daemon starts up, it changes directory to the









data directory.  This is where it expects to write log files and the pid









(process ID) file, and where it expects to find databases.

















The data directory location is hardwired in when the distribution is









compiled.  However, if @code{mysqld} expects to find the data directory









somewhere other than where it really is on your system, it will not work









properly.  If you have problems with incorrect paths, you can find out









what options @code{mysqld} allows and what the default path settings are by









invoking @code{mysqld} with the @code{--help} option.  You can override the









defaults by specifying the correct pathnames as command-line arguments to









@code{mysqld}.  (These options can be used with @code{safe_mysqld} as well.)

















Normally you should need to tell @code{mysqld} only the base directory under









which @strong{MySQL} is installed.  You can do this with the @code{--basedir}









option.  You can also use @code{--help} to check the effect of changing path









options (note that @code{--help} @emph{must} be the final option of the









@code{mysqld} command).  For example:

















@example









shell> EXECDIR/mysqld --basedir=/usr/local --help









@end example

















Once you determine the path settings you want, start the server without









the @code{--help} option.

















Whichever method you use to start the server, if it fails to start up









correctly, check the log file to see if you can find out why.  Log files









are located in the data directory (typically








@@ -9146,32 +9172,6 @@ the old Berkeley DB log file from the database directory to some other







place, where you can later examine these.  The log files are named









@file{log.0000000001}, where the number will increase over time.

















When the @code{mysqld} daemon starts up, it changes directory to the









data directory.  This is where it expects to write log files and the pid









(process ID) file, and where it expects to find databases.

















The data directory location is hardwired in when the distribution is









compiled.  However, if @code{mysqld} expects to find the data directory









somewhere other than where it really is on your system, it will not work









properly.  If you have problems with incorrect paths, you can find out









what options @code{mysqld} allows and what the default path settings are by









invoking @code{mysqld} with the @code{--help} option.  You can override the









defaults by specifying the correct pathnames as command-line arguments to









@code{mysqld}.  (These options can be used with @code{safe_mysqld} as well.)

















Normally you should need to tell @code{mysqld} only the base directory under









which @strong{MySQL} is installed.  You can do this with the @code{--basedir}









option.  You can also use @code{--help} to check the effect of changing path









options (note that @code{--help} @emph{must} be the final option of the









@code{mysqld} command).  For example:

















@example









shell> EXECDIR/mysqld --basedir=/usr/local --help









@end example

















Once you determine the path settings you want, start the server without









the @code{--help} option.

















If you get the following error, it means that some other program (or another









@code{mysqld} server) is already using the TCP/IP port or socket









@code{mysqld} is trying to use:








@@ -9222,6 +9222,10 @@ This will not run in the background and it should also write a trace in







@file{\mysqld.trace}, which may help you determine the source of your









problems. @xref{Windows}.

















If you are using BDB (Berkeley DB) tables, you should familiarize









yourself with the different BDB specific startup options. @xref{BDB start}.

























@node Automatic start, Command-line options, Starting server, Post-installation









@subsection Starting and Stopping MySQL Automatically









@cindex starting, the server automatically








@@ -9747,6 +9751,10 @@ Version 3.23:















@itemize @bullet









@item









If you do a @code{DROP DATABASE} on a symbolic linked database, both the









link and the original database is deleted.  (This didn't happen in 3.22









because configure didn't detect the @code{readlink} system call).









@item









@code{OPTIMIZE TABLE} now only works for @strong{MyISAM} tables.









For other table types, you can use @code{ALTER TABLE} to optimize the table.









During @code{OPTIMIZE TABLE} the table is now locked from other threads.








@@ -17464,7 +17472,9 @@ DROP DATABASE [IF EXISTS] db_name







@end example

















@code{DROP DATABASE} drops all tables in the database and deletes the









database.  @strong{Be VERY careful with this command!}









database.  If you do a @code{DROP DATABASE} on a symbolic linked









database, both the link and the original database is deleted. @strong{Be









VERY careful with this command!}

















@code{DROP DATABASE} returns the number of files that were removed from









the database directory.  Normally, this is three times the number of








@@ -18261,10 +18271,13 @@ Deleted records are maintained in a linked list and subsequent @code{INSERT}







operations reuse old record positions. You can use @code{OPTIMIZE TABLE} to









reclaim the unused space and to defragment the data file.

















For the moment @code{OPTIMIZE TABLE} only works on @strong{MyISAM}









tables.  You can get optimize table to work on other table types by









starting @code{mysqld} with @code{--skip-new} or @code{--safe-mode}, but in









this case @code{OPTIMIZE TABLE} is just mapped to @code{ALTER TABLE}.









For the moment @code{OPTIMIZE TABLE} only works on @strong{MyISAM} and









@code{BDB} tables. For @code{BDB} tables, @code{OPTIMIZE TABLE} is









currently mapped to @code{ANALYZE TABLE}. @xref{ANALYZE TABLE}.

















You can get optimize table to work on other table types by starting









@code{mysqld} with @code{--skip-new} or @code{--safe-mode}, but in this









case @code{OPTIMIZE TABLE} is just mapped to @code{ALTER TABLE}.

















@code{OPTIMIZE TABLE} works the following way:









@itemize @bullet








@@ -18277,7 +18290,7 @@ If the statistics are not up to date (and the repair couldn't be done







by sorting the index), update them.









@end itemize

















@code{OPTIMIZE TABLE} is equvialent of running









@code{OPTIMIZE TABLE} for @code{MyISAM} tables is equvialent of running









@code{myisamchk --quick --check-changed-tables --sort-index --analyze}









on the table.
















@@ -18294,11 +18307,12 @@ CHECK TABLE tbl_name[,tbl_name...] [option [option...]]







option = QUICK | FAST | EXTEND | CHANGED









@end example

















@code{CHECK TABLE} only works on @code{MyISAM} tables and is the same thing









as running @code{myisamchk -m table_name} on the table.









@code{CHECK TABLE} only works on @code{MyISAM} and @code{BDB} tables. On









@code{MyISAM} tables it's the same thing as running @code{myisamchk -m









table_name} on the table.

















Check the table(s) for errors and update the key statistics for the table.









The command returns a table with the following columns:









Checks the table(s) for errors. For @code{MyISAM} tables the key statistics









is updated. The command returns a table with the following columns:

















@multitable @columnfractions .35 .65









@item @strong{Column} @tab @strong{Value}








@@ -18325,6 +18339,9 @@ The different check types stand for the following:







@item @code{EXTENDED} @tab Do a full key lookup for all keys for each row.  This ensures that the table is 100 % consistent, but will take a long time!









@end multitable

















Note that for BDB tables the different check options doesn't affect the









check in any way!

















You can combine check options as in:

















@example








@@ -18423,7 +18440,9 @@ ANALYZE TABLE tbl_name[,tbl_name...]







@end example

















Analyze and store the key distribution for the table.  During the









analyze the table is locked with a read lock.









analyze the table is locked with a read lock.  This works on









@code{MyISAM} and @code{BDB} tables.

















This is equivalent to running @code{myisamchk -a} on the table.

















@strong{MySQL} uses the stored key distribution to decide in which order








@@ -20108,16 +20127,15 @@ If @code{key_reads} is big, then your @code{key_cache} is probably too







small.  The cache hit rate can be calculated with









@code{key_reads}/@code{key_read_requests}.









@item









If @code{Handler_read_rnd} is big, then you probably have a lot of queries









that require @strong{MySQL} to scan whole tables or you have joins that don't use









keys properly.









If @code{Handler_read_rnd} is big, then you probably have a lot of









queries that require @strong{MySQL} to scan whole tables or you have









joins that don't use keys properly.









@item









If @code{Created_tmp_tables} or @code{Sort_merge_passes} are high then









your @code{mysqld} @code{sort_buffer} variables is probably too small.









@item









@code{Created_tmp_files} doesn't count the files needed to handle temporary









tables.









@item









@end itemize

















@node SHOW VARIABLES, SHOW PROCESSLIST, SHOW STATUS, SHOW








@@ -20143,6 +20161,7 @@ differ somewhat:







| bdb_home                | /usr/local/mysql/data/          |









| bdb_logdir              |                                 |









| bdb_tmpdir              | /tmp/                           |









| binlog_cache_size       | 32768                           |









| character_set           | latin1                          |









| character_sets          | latin1                          |









| connect_timeout         | 5                               |








@@ -20239,7 +20258,7 @@ cache.







@item @code{bdb_home}









The value of the @code{--bdb-home} option.

















@item @code{bdb_lock_max}









@item @code{bdb_max_lock}









The maximum number of locks (1000 by default) you can have active on a









BDB table. You should increase this if you get errors of type @code{bdb:









Lock table is out of available locks} or @code{Got error 12 from ...}








@@ -20249,9 +20268,17 @@ a lot of rows to calculate the query.







@item @code{bdb_logdir}









The value of the @code{--bdb-logdir} option.

















@item @code{bdb_shared_data}









Is @code{ON} if you are using @code{--bdb-shared-data}.

















@item @code{bdb_tmpdir}









The value of the @code{--bdb-tmpdir} option.

















@item @code{binlog_cache_size}.  The size of the cache to hold the SQL









statements for the binary log during a transaction.  If you often use









big, multi-statement transactions you can increase this to get more









performance. @xref{COMMIT}.

















@item @code{character_set}









The default character set.
















@@ -20390,6 +20417,11 @@ wrong) packets.  You must increase this value if you are using big







@code{BLOB} columns. It should be as big as the biggest @code{BLOB} you want









to use.

















@item @code{max_binlog_cache_size}.  If a multi-statement transaction









requires more than this amount of memory, one will get the error









"Multi-statement transaction required more than 'max_binlog_cache_size'









bytes of storage".

















@item @code{max_connections}









The number of simultaneous clients allowed. Increasing this value increases









the number of file descriptors that @code{mysqld} requires.  See below for








@@ -21014,6 +21046,21 @@ table you will get an error (@code{ER_WARNING_NOT_COMPLETE_ROLLBACK}) as







a warning.  All transactional safe tables will be restored but any









non-transactional table will not change.

















If you are using @code{BEGIN} or @code{SET AUTO_COMMIT=0}, you









should use the @strong{MySQL} binary log for backups instead of the









old update log;  The transaction is stored in the binary log









in one chunk, during @code{COMMIT}, the to ensure and @code{ROLLBACK}:ed









transactions are not stored. @xref{Binary log}.

















The following commands automaticly ends an transaction (as if you had done









a @code{COMMIT} before executing the command):

















@multitable @columnfractions .33 .33 .33









@item @code{ALTER TABLE} @tab @code{BEGIN} @tab @code{CREATE INDEX}









@item @code{DROP DATABASE} @tab @code{DROP TABLE} @tab @code{RENAME TABLE}









@item @code{TRUNCATE} 









@end multitable

















@findex LOCK TABLES









@findex UNLOCK TABLES









@node LOCK TABLES, SET OPTION, COMMIT, Reference








@@ -22511,11 +22558,12 @@ BDB tables:







@item @code{--bdb-home=directory} @tab  Base directory for BDB tables. This should be the same directory you use for --datadir.









@item @code{--bdb-lock-detect=#} @tab  Berkeley lock detect. One of (DEFAULT, OLDEST, RANDOM, or YOUNGEST).









@item @code{--bdb-logdir=directory} @tab Berkeley DB log file directory.









@item @code{--bdb-nosync} @tab Don't synchronously flush logs.









@item @code{--bdb-no-sync} @tab Don't synchronously flush logs.









@item @code{--bdb-recover} @tab Start Berkeley DB in recover mode.









@item @code{--bdb-shared-data} @tab Start Berkeley DB in multi-process mode (Don't use @code{DB_PRIVATE} when initializing Berkeley DB)









@item @code{--bdb-tmpdir=directory} @tab Berkeley DB tempfile name.









@item @code{--skip-bdb} @tab Don't use berkeley db.









@item @code{-O bdb_lock_max=1000} @tab Set the maximum number of locks possible. @xref{SHOW VARIABLES}.









@item @code{-O bdb_max_lock=1000} @tab Set the maximum number of locks possible. @xref{SHOW VARIABLES}.









@end multitable

















If you use @code{--skip-bdb}, @strong{MySQL} will not initialize the








@@ -22526,13 +22574,17 @@ Normally you should start mysqld with @code{--bdb-recover} if you intend







to use BDB tables.  This may, however, give you problems when you try to









start mysqld if the BDB log files are corrupted. @xref{Starting server}.

















With @code{bdb_lock_max} you can specify the maximum number of locks









With @code{bdb_max_lock} you can specify the maximum number of locks









(1000 by default) you can have active on a BDB table. You should









increase this if you get errors of type @code{bdb: Lock table is out of









available locks} or @code{Got error 12 from ...}  when you have do long









transactions or when @code{mysqld} has to examine a lot of rows to









calculate the query.

















You may also want to change @code{binlog_cache_size} and









@code{max_binlog_cache_size} if you are using big multi-line transactions.









@xref{COMMIT}.

















@node BDB characteristic, BDB TODO, BDB start, BDB









@subsection Some characteristic of @code{BDB} tables:
















@@ -22578,6 +22630,10 @@ tables. In other words, the key information will take a little more







space in @code{BDB} tables compared to MyISAM tables which don't use









@code{PACK_KEYS=0}.









@item









There is often holes in the BDB table to allow you to insert new rows









between different keys.  This makes BDB tables somewhat larger than









MyISAM tables.









@item









@strong{MySQL} performs a checkpoint each time a new Berkeley DB log









file is started, and removes any log files that are not needed for









current transactions.  One can also run @code{FLUSH LOGS} at any time








@@ -22585,6 +22641,17 @@ to checkpoint the Berkeley DB tables.















For disaster recovery, one should use table backups plus MySQL's binary









log. @xref{Backup}.









@item









The optimizer needs to know an approximation of the number of rows in









the table.  @strong{MySQL} solves this by counting inserts and









maintaining this in a separate segment in each BDB table.  If you don't









do a lot of @code{DELETE} or @code{ROLLBACK}:s this number should be









accurate enough for the @strong{MySQL} optimizer, but as @strong{MySQL}









only store the number on close, it may be wrong if @strong{MySQL} dies









unexpectedly. It should not be fatal even if this number is not 100 %









correct.  One can update the number of rows by executing @code{ANALYZE









TABLE} or @code{OPTIMIZE TABLE}. @xref{ANALYZE TABLE} . @xref{OPTIMIZE









TABLE}.









@end itemize

















@node BDB TODO, BDB errors, BDB characteristic, BDB








@@ -25367,7 +25434,8 @@ server-id=<some unique number between 1 and 2^32-1>







@end example

















@code{server-id} must be different for each server participating in









replication.









replication.  If you don't specify a server-id, it will be set to









1 if you have not defined @code{master-host}, else it will be set to 2.

















@item Restart the slave(s).
















@@ -26341,6 +26409,7 @@ like this:







Possible variables for option --set-variable (-O) are:









back_log              current value: 5









bdb_cache_size        current value: 1048540









binlog_cache_size     current_value: 32768









connect_timeout       current value: 5









delayed_insert_timeout  current value: 300









delayed_insert_limit  current value: 100








@@ -26352,6 +26421,7 @@ key_buffer_size       current value: 1048540







lower_case_table_names  current value: 0









long_query_time       current value: 10









max_allowed_packet    current value: 1048576









max_binlog_cache_size current_value: 4294967295









max_connections       current value: 100









max_connect_errors    current value: 10









max_delayed_threads   current value: 20








@@ -33323,7 +33393,8 @@ and the crash.







@node Binary log, Slow query log, Update log, Common problems









@section The Binary Log

















In the future we expect the binary log to replace the update log!









In the future the binary log will replace the update log, so we









recommend you to switch to this log format as soon as possible!

















The binary log contains all information that is available in the update









log in a more efficient format. It also contains information about how long








@@ -33369,6 +33440,20 @@ direct from a remote mysql server!







@code{mysqlbinlog --help} will give you more information of how to use









this program!

















If you are using @code{BEGIN} or @code{SET AUTO_COMMIT=0}, you must use









the @strong{MySQL} binary log for backups instead of the old update log.

















All updates (@code{UPDATE}, @code{DELETE} or @code{INSERT}) that changes









a transactional table (like BDB tables) is cached until a @code{COMMIT}.









Any updates to a not transactional table is stored in the binary log at









once.  Every thread will on start allocate a buffer of









@code{binlog_cache_size} to buffer queries.  If a query is bigger than









this, the thread will open a temporary file to handle the bigger cache.









The temporary file will be deleted when the thread ends.

















The @code{max_binlog_cache_size} can be used to restrict the total size used









to cache a multi-transaction query.

















@cindex slow query log









@cindex files, slow query log









@node Slow query log, Multiple servers, Binary log, Common problems








@@ -39275,6 +39360,27 @@ though, so Version 3.23 is not released as a stable version yet.







@appendixsubsec Changes in release 3.23.29









@itemize @bullet









@item









Renamed variable @code{bdb_lock_max} to @code{bdb_max_lock}.









@item









Changed the default server-id to 1 for masters and 2 for slaves









to make it easier to use the binary log.









@item









Added @code{CHECK}, @code{ANALYZE} and @code{OPTIMIZE} of BDB tables.









@item









Store in BDB tables the number of rows; This helps to optimize queries









when we need an approximation of the number of row.









@item









Made @code{DROP TABLE}, @code{RENAME TABLE}, @code{CREATE INDEX} and









@code{DROP INDEX} are now transaction endpoints.









@item









Added option @code{--bdb-shared-data} to @code{mysqld}.









@item









Added variables @code{binlog_cache_size} and @code{max_binlog_cache_size} to









@code{mysqld}.









@item









If you do a @code{DROP DATABASE} on a symbolic linked database, both









the link and the original database is deleted.









@item









Fixed that @code{DROP DATABASE} works on OS/2.









@item









Fixed bug when doing a @code{SELECT DISTINCT ... table1 LEFT JOIN

























configure.in



+1
−1






























Original line number
Diff line number
Diff line







@@ -1239,7 +1239,7 @@ AC_CHECK_FUNCS(alarm bmove \







 chsize ftruncate rint finite fpsetmask fpresetsticky\









 cuserid fcntl fconvert poll \









 getrusage getpwuid getcwd getrlimit getwd index stpcpy locking longjmp \









 perror pread realpath rename \









 perror pread realpath readlink rename \









 socket strnlen madvise mkstemp \









 strtol strtoul strtoull snprintf tempnam thr_setconcurrency \









 gethostbyaddr_r gethostbyname_r getpwnam \
































include/mysqld_error.h



+2
−1






























Original line number
Diff line number
Diff line







@@ -197,4 +197,5 @@







#define ER_CRASHED_ON_USAGE 1194









#define ER_CRASHED_ON_REPAIR 1195









#define ER_WARNING_NOT_COMPLETE_ROLLBACK 1196









#define ER_ERROR_MESSAGES 197








#define ER_TRANS_CACHE_FULL 1197









#define ER_ERROR_MESSAGES 198
























sql/ha_berkeley.cc



+327
−64




























File changed.
Preview size limit exceeded, changes collapsed.
























sql/ha_berkeley.h



+18
−8






























Original line number
Diff line number
Diff line







@@ -27,11 +27,13 @@

















typedef struct st_berkeley_share {









  ulonglong auto_ident;









  ha_rows rows, org_rows, *rec_per_key;









  THR_LOCK lock;









  pthread_mutex_t mutex;









  char *table_name;









  DB *status_block;









  uint table_name_length,use_count;









  bool primary_key_inited;









  uint status,version;









} BDB_SHARE;




























@@ -49,7 +51,8 @@ class ha_berkeley: public handler







  BDB_SHARE *share;









  ulong int_option_flag;









  ulong alloced_rec_buff_length;









  uint primary_key,last_dup_key, hidden_primary_key;









  ulong changed_rows;









  uint primary_key,last_dup_key, hidden_primary_key, version;









  bool fixed_length_row, fixed_length_primary_key, key_read;









  bool  fix_rec_buff_for_blob(ulong length);









  byte current_ident[BDB_HIDDEN_PRIMARY_KEY_LENGTH];








@@ -58,7 +61,8 @@ class ha_berkeley: public handler







  int pack_row(DBT *row,const  byte *record, bool new_row);









  void unpack_row(char *record, DBT *row);









  void ha_berkeley::unpack_key(char *record, DBT *key, uint index);









  DBT *pack_key(DBT *key, uint keynr, char *buff, const byte *record);









  DBT *create_key(DBT *key, uint keynr, char *buff, const byte *record,









		  int key_length = MAX_KEY_LENGTH);









  DBT *pack_key(DBT *key, uint keynr, char *buff, const byte *key_ptr,









		uint key_length);









  int remove_key(DB_TXN *trans, uint keynr, const byte *record,








@@ -79,8 +83,9 @@ class ha_berkeley: public handler







		    HA_KEYPOS_TO_RNDPOS | HA_READ_ORDER | HA_LASTKEY_ORDER |









		    HA_LONGLONG_KEYS | HA_NULL_KEY | HA_HAVE_KEY_READ_ONLY |









		    HA_BLOB_KEY | HA_NOT_EXACT_COUNT | 









		    HA_PRIMARY_KEY_IN_READ_INDEX | HA_DROP_BEFORE_CREATE),









    last_dup_key((uint) -1)









		    HA_PRIMARY_KEY_IN_READ_INDEX | HA_DROP_BEFORE_CREATE |









		    HA_AUTO_PART_KEY),









    last_dup_key((uint) -1),version(0)









  {









  }









  ~ha_berkeley() {}








@@ -123,6 +128,10 @@ class ha_berkeley: public handler







  int reset(void);









  int external_lock(THD *thd, int lock_type);









  void position(byte *record);









  int analyze(THD* thd,HA_CHECK_OPT* check_opt);









  int optimize(THD* thd, HA_CHECK_OPT* check_opt);









  int check(THD* thd, HA_CHECK_OPT* check_opt);



















  ha_rows records_in_range(int inx,









			   const byte *start_key,uint start_key_len,









			   enum ha_rkey_function start_search_flag,








@@ -135,7 +144,7 @@ class ha_berkeley: public handler







  THR_LOCK_DATA **store_lock(THD *thd, THR_LOCK_DATA **to,









			     enum thr_lock_type lock_type);



















  void update_auto_primary_key();









  void get_status();









  inline void get_auto_primary_key(byte *to)









  {









    ulonglong tmp;








@@ -144,11 +153,12 @@ class ha_berkeley: public handler







    int5store(to,share->auto_ident);









    pthread_mutex_unlock(&share->mutex);









  }









  longlong ha_berkeley::get_auto_increment();









};



















extern bool berkeley_skip;









extern bool berkeley_skip, berkeley_shared_data;









extern u_int32_t berkeley_init_flags,berkeley_lock_type,berkeley_lock_types[];









extern ulong berkeley_cache_size, berkeley_lock_max;









extern ulong berkeley_cache_size, berkeley_max_lock;









extern char *berkeley_home, *berkeley_tmpdir, *berkeley_logdir;









extern long berkeley_lock_scan_time;









extern TYPELIB berkeley_lock_typelib;