merge

monty@hundin.mysql.fi

monty@work.mysql.com

mwagner@evoq.mwagner.org

paul@central.snake.net

sasha@mysql.sashanet.com

tonu@hundin.mysql.fi

* Compile and link options::    How compiling and linking affects the speed of MySQL

* Disk issues::                 Disk issues

* Symbolic links::              

* Server parameters::           Tuning server parameters

* Table cache::                 How MySQL opens and closes tables

* Creating many tables::        Drawbacks of creating large numbers of tables in the same database

* Table locking::               Table locking issues

* DNS::                         

Disk Issues

Using Symbolic Links

* Symbolic links::              Using symbolic links for databases and tables

* Symbolic links to database::  

* Symbolic links to tables::    

Speed of Queries that Access or Update Data

shell> ./bin/safe_mysqld --user=mysql &

@end example

For a binary distribution, do this:

For a binary distribution (not RPM or pkg packages), do this:

@example

shell> cd mysql_installation_directory

Don't use new, possible wrong routines.  Implies @code{--skip-delay-key-write}.

This will also set default table type to @code{ISAM}.  @xref{ISAM}.

@item --skip-stack-trace

Don't write stack traces.  This option is useful when you are running

@code{mysqld} under a debugger. @xref{Debugging server}.

@item --skip-symlinks

Don't delete or rename files that symlinks in the data directory points to.

@item --skip-safemalloc

If @strong{MySQL} is configured with @code{--with-debug=full}, all programs

Don't allow 'SHOW DATABASE' commands, unless the user has @strong{process}

privilege.

@item --skip-stack-trace

Don't write stack traces.  This option is useful when you are running

@code{mysqld} under a debugger. @xref{Debugging server}.

@item --skip-thread-priority

Disable using thread priorities for faster response time.

Normally this is done with the @code{su} command.  For more details, see

@ref{Changing MySQL user, , Changing @strong{MySQL} user}.

@item

Don't support symlinks to tables (This can be disabled with the

@code{--skip-symlinks} option. This is especially important if you run

@code{mysqld} as root as anyone that has write access to the mysqld data

directories could then delete any file in the system!

@xref{Symbolic links to tables}.

@item

If you put a password for the Unix @code{root} user in the @code{mysql.server}

script, make sure this script is readable only by @code{root}.

or      ROW_FORMAT= @{ default | dynamic | fixed | compressed @}

or	RAID_TYPE= @{1 | STRIPED | RAID0 @} RAID_CHUNKS=#  RAID_CHUNKSIZE=#

or	UNION = (table_name,[table_name...])

or      DATA DIRECTORY="directory"

or      INDEX DIRECTORY="directory"

select_statement:

	[IGNORE | REPLACE] SELECT ...  (Some legal select statement)

by all @code{UNIQUE} keys and then the normal keys.  This helps the

@strong{MySQL} optimizer to prioritize which key to use and also more quickly

detect duplicated @code{UNIQUE} keys.

@item

By using @code{DATA DIRECTORY="directory"} or @code{INDEX

DIRECTORY="directory"} you can specify where the table handler should

put it's table and index files.  This only works for @code{MyISAM} tables

in @code{MySQL} 4.0, when you are not using the @code{--skip-symlinks}

option. @xref{Symbolic links to tables}.

@end itemize

@cindex silent column changes

deleted blocks and by extending blocks if the next block is deleted.

@item

@code{myisampack} can pack @code{BLOB} and @code{VARCHAR} columns.

@item

You can use put the datafile and index file on different directories

to get more speed (with the @code{DATA/INDEX DIRECTORY="path"} option to

@code{CREATE TABLE}). @xref{CREATE TABLE}.

@end itemize

@code{MyISAM} also supports the following things, which @strong{MySQL}

@code{RAND()}. You can, for example, use @code{UNIX_TIMESTAMP()} for the

argument to @code{RAND()}.

@item

You have to use the same character set (@code{--default-character-set})

on the master and the slave. If not, you may get duplicate key errors on

the slave, because a key that is regarded as unique on the master may

not be that in the other character set.

@item

@code{LOAD DATA INFILE} will be handled properly as long as the file

still resides on the master server at the time of update

propagation. @code{LOAD LOCAL DATA INFILE} will be skipped.

@menu

* Compile and link options::    How compiling and linking affects the speed of MySQL

* Disk issues::                 Disk issues

* Symbolic links::              Using Symbolic Links

* Server parameters::           Tuning server parameters

* Table cache::                 How MySQL opens and closes tables

* Creating many tables::        Drawbacks of creating large numbers of tables in the same database

@cindex disk issues

@cindex performance, disk issues

@node Disk issues, Server parameters, Compile and link options, System

@node Disk issues, Symbolic links, Compile and link options, System

@subsection Disk Issues

@itemize @bullet

with the noatime flag.

@end itemize

@menu

* Symbolic links::              Using symbolic links for databases and tables

@end menu

@cindex symbolic links

@cindex links, symbolic

@cindex databases, symbolic links

@cindex tables, symbolic links

@node Symbolic links,  , Disk issues, Disk issues

@subsubsection Using Symbolic Links for Databases and Tables

@node Symbolic links, Server parameters, Disk issues, System

@subsection Using Symbolic Links

You can move tables and databases from the database directory to other

locations and replace them with symbolic links to the new locations.

You might want to do this, for example, to move a database to a file

system with more free space.

system with more free space or increase the speed of your system by

spreading your tables to different disk.

The recommended may to do this, is to just symlink databases to different

disk and only symlink tables as a last resort.

.

If @strong{MySQL} notices that a table is symbolically linked, it will

resolve the symlink and use the table it points to instead. This works

on all systems that support the @code{realpath()} call (at least Linux

and Solaris support @code{realpath()})!  On systems that don't support

@code{realpath()}, you should not access the table through the real path

and through the symlink at the same time!  If you do, the table will be

inconsistent after any update.

@cindex databases, symbolic links

@menu

* Symbolic links to database::  

* Symbolic links to tables::    

@end menu

@strong{MySQL} doesn't that you link one directory to multiple

@node Symbolic links to database, Symbolic links to tables, Symbolic links, Symbolic links

@subsubsection Using Symbolic Links for Databases

The way to symlink a database is to first create a directory on some

disk where you have free space and then create a symlink to it from

the @strong{MySQL} database directory.

@example

shell> mkdir /dr1/databases/test

shell> ln -s /dr1/databases/test mysqld-datadir

@end example

@strong{MySQL} doesn't support that you link one directory to multiple

databases.  Replacing a database directory with a symbolic link will

work fine as long as you don't make a symbolic link between databases.

Suppose you have a database @code{db1} under the @strong{MySQL} data

@strong{MySQL} with @code{-DUSE_SYMDIR}. This allows you to put different

databases on different disks. @xref{Windows symbolic links}.

@cindex databases, symbolic links

@node Symbolic links to tables,  , Symbolic links to database, Symbolic links

@subsubsection Using Symbolic Links for Tables

Before @strong{MySQL} 4.0 you should not symlink tables, if you are not

very carefully with them. The problem is that if you run @code{ALTER

TABLE}, @code{REPAIR TABLE} or @code{OPTIMIZE TABLE} on a symlinked

table, the symlinks will be removed and replaced by the original

files. This happens because the above command works by creating a

temporary file in the database directory and when the command is

complete, replace the original file with the temporary file.

You should not symlink tables on system that doesn't have a fully

working @code{realpath()} call. (At least Linux and Solaris support

@code{realpath()})

In @strong{MySQL} 4.0 symlinks is only fully supported for @code{MyISAM}

tables.  For other table types you will probably get strange problems

when doing any of the above mentioned commands.

The handling of symbolic links in @strong{MySQL} 4.0 works the following

way (this is mostly relevant only for @code{MyISAM} tables).

@itemize @bullet

@item

In the data directory you will always have the table definition file

and the data/index files.

@item

You can symlink the index file and the data file to different directories

independent of the other.

@item

The symlinking can be done from the operating system (if @code{mysqld} is

not running) or with the @code{INDEX/DATA DIRECTORY="path-to-dir"} command

in @code{CREATE TABLE}. @xref{CREATE TABLE}.

@item

@code{myisamchk} will not replace a symlink with the index/file but

work directly on the files the symlinks points to. Any temporary files

will be created in the same directory where the data/index file is.

@item

When you drop a table that is using symlinks, both the symlink and the

file the symlink points to is dropped.  This is a good reason to why you

should NOT run @code{mysqld} as root and not allow persons to have write

access to the @strong{MySQL} database directories.

@item

If you rename a table with @code{ALTER TABLE RENAME} and you don't change

database, the symlink in the database directory will be renamed to the new

name and the data/index file will be renamed accordingly.

@item

If you use @code{ALTER TABLE RENAME} to move a table to another database,

then the table will be moved to the other database directory and the old

symlinks and the files they pointed to will be deleted.

@item

If you are not using symlinks you should use the @code{--skip-symlinks}

option to @code{mysqld} to ensure that no one can drop or rename a file

outside of the @code{mysqld} data directory.

@end itemize

Things that are not yet fully supported:

@cindex TODO, symlinks

@itemize @bullet

@item

@code{ALTER TABLE} ignores all @code{INDEX/DATA DIRECTORY="path"} options.

@item

@code{CREATE TABLE} doesn't report if the table has symbolic links.

@item

@code{mysqldump} doesn't include the symbolic links information in the output.

@item

@code{BACKUP TABLE} and @code{RESTORE TABLE} doesn't use symbolic links.

@end itemize

@cindex parameters, server

@cindex @code{mysqld} server, buffer sizes

@cindex buffer sizes, @code{mysqld} server

@cindex startup parameters

@node Server parameters, Table cache, Disk issues, System

@node Server parameters, Table cache, Symbolic links, System

@subsection Tuning Server Parameters

You can get the default buffer sizes used by the @code{mysqld} server

@node ALTER TABLE problems, Change column order, No matching rows, Problems

@section Problems with @code{ALTER TABLE}.

@code{ALTER TABLE} changes a table to the current character set.

If you during @code{ALTER TABLE} get a duplicate key error, then the cause

is either that the new character sets maps to keys to the same value

or that the table is corrupted, in which case you should run

@code{REPAIR TABLE} on the table.

If @code{ALTER TABLE} dies with an error like this:

@example

Romanian error messages.

@item Peter Feher

Hungarian error messages.

@item Roberto M. Serqueira

Portugise error messages.

@item David Sacerdote @email{davids@@secnet.com}

Ideas for secure checking of DNS hostnames.

@item Wei-Jou Chen @email{jou@@nematic.ieo.nctu.edu.tw}

Ported (and extended) the benchmark suite to @code{DBI}/@code{DBD}. Have

been of great help with @code{crash-me} and running benchmarks. Some new

date functions. The mysql_setpermissions script.

@item Jay Flaherty @email{fty@@utk.edu}

@item Jay Flaherty @email{fty@@mediapulse.com}

Big parts of the Perl @code{DBI}/@code{DBD} section in the manual.

@item Paul Southworth @email{pauls@@etext.org}, Ray Loyzaga @email{yar@@cs.su.oz.au}

Proof-reading of the Reference Manual.

@item

Added @code{HANDLER} command.

@item

Added @code{SQL_CALC_FOUND_ROWS} and @code{FOUND_ROWS()}. This make it

possible to know how many rows a query would have returned if one hadn't

used @code{LIMIT}.

Added support for symbolic links to @code{MyISAM} tables.

@item

Added @code{SQL_CALC_FOUND_ROWS} and @code{FOUND_ROWS()}. This makes it

possible to know how many rows a query would have returned

without a @code{LIMIT} clause.

@item

Changed output format of @code{SHOW OPEN TABLES}.

@item

@item

Optimized queries of type:

@code{SELECT DISTINCT * from table_name ORDER by key_part1 LIMIT #}

@item

Added support for sym-linking of MyISAM tables.

@end itemize

@node News-3.23.x, News-3.22.x, News-4.0.x, News

@appendixsubsec Changes in release 3.23.39

@itemize @bullet

@item

Fixed problem that client 'hang' when @code{LOAD TABLE FROM MASTER} failed.

@item

Running @code{myisamchk --fast --force} will not anymore repair tables

that only had the open count wrong.

@item

Added functions to handle symbolic links to make life easier in 4.0.

@item

We are now using the @code{-lcma} thread library on HPUX 10.20 to

get @strong{MySQL} more stabile on HPUX.

@item

extern ulong	my_default_record_cache_size;

extern my_bool NEAR my_disable_locking,NEAR my_disable_async_io,

       NEAR my_disable_flush_key_blocks;

               NEAR my_disable_flush_key_blocks, NEAR my_disable_symlinks;

extern char	wild_many,wild_one,wild_prefix;

extern const char *charsets_dir;

extern char *defaults_extra_file;

  "Embedded server",

};

/* Start of code added by Roberto M. Serqueira - martinsc@uol.com.br - 05.24.2001 */

#elif defined PORTUGUESE

const char *client_errors[]=

{

  "Erro desconhecido do MySQL",

  "No pode criar 'UNIX socket' (%d)",

  "No pode se conectar ao servidor MySQL local atravs do 'socket' '%-.64s' (%d)", 

  "No pode se conectar ao servidor MySQL em '%-.64s' (%d)",

  "No pode criar 'socket TCP/IP' (%d)",

  "'Host' servidor MySQL '%-.64s' (%d) desconhecido", 

  "Servidor MySQL desapareceu",

  "Incompatibilidade de protocolos. Verso do Servidor: %d - Verso do Cliente: %d",

  "Cliente do MySQL com falta de memria",

  "Informao invlida de 'host'",

  "Localhost via 'UNIX socket'",

  "%-.64s via 'TCP/IP'",

  "Erro na negociao de acesso ao servidor",

  "Conexo perdida com servidor MySQL durante 'query'",

  "Comandos fora de sincronismo. Voc no pode executar este comando agora",

  "%-.64s via 'named pipe'",

  "No pode esperar pelo 'named pipe' para o 'host' %-.64s - 'pipe' %-.32s (%lu)",

  "No pode abrir 'named pipe' para o 'host' %-.64s - 'pipe' %-.32s (%lu)",

  "No pode estabelecer o estado do 'named pipe' para o 'host' %-.64s - 'pipe' %-.32s (%lu)",

  "No pode inicializar conjunto de caracteres %-.64s (caminho %-.64s)",

  "Obteve pacote maior do que 'max_allowed_packet'",

  "Embedded server"

};

#else /* ENGLISH */

const char *client_errors[]=

{

  "Can't set state of named pipe to host: %-.64s  pipe: %-.32s (%lu)",

  "Can't initialize character set %-.64s (path: %-.64s)",

  "Got packet bigger than 'max_allowed_packet'",

  "Embedded server",

  "Embedded server"

};

#endif

    mi_check_print_warning(param,

			   "%d clients is using or hasn't closed the table properly",

			   share->state.open_count);

    /* If this will be fixed by the check, forget the warning */

    if (param->testflag & T_UPDATE_STATE)

      param->warning_printed=save;

  }

  return 0;