mk-query-profiler

Langue: en

Version: 2008-12-29 (fedora - 04/07/09)

Section: 1 (Commandes utilisateur)

NAME

mk-query-profiler - Execute SQL statements and print statistics, or measure activity caused by other processes.

SYNOPSIS

mk-query-profiler can profile the (semicolon-terminated, blank-line separated) queries in a file:
    mk-query-profiler queries.sql
    cat queries.sql | mk-query-profiler
    mk-query-profiler -vv queries.sql
    mk-query-profiler -v --separate --only 2,5,6 queries.sql
    mk-query-profiler --tab queries.sql > results.csv
 
 

It can also just observe what happens in the server:

    mk-query-profiler --external
 
 

Or it can run shell commands from a file and measure the result:

    mk-query-profiler --external commands.txt
    mk-query-profiler --external - < commands.txt
 
 

Read ``HOW TO INTERPRET'' to learn what it all means.

DOWNLOADING

You can download Maatkit from Google Code at <http://code.google.com/p/maatkit/>, or you can get any of the tools easily with a command like the following:
    wget http://www.maatkit.org/get/toolname
    or
    wget http://www.maatkit.org/trunk/toolname
 
 

Where "toolname" can be replaced with the name (or fragment of a name) of any of the Maatkit tools. Once downloaded, they're ready to run; no installation is needed. The first URL gets the latest released version of the tool, and the second gets the latest trunk code from Subversion.

OPTIONS

--allowcache
short form: -a

Let MySQL query cache cache the queries executed.

By default this is disabled. When enabled, cache profiling information is added to the printout. See <http://dev.mysql.com/doc/en/query-cache.html> for more information about the query cache.

--askpass
Prompt for a password when connecting to MySQL.
--calibrate
short form: -c; negatable: yes; default: yes

Try to compensate for "SHOW STATUS".

Measure and compensate for the ``cost of observation'' caused by running SHOW STATUS. Only works reliably on a quiet server; on a busy server, other processes can cause the calibration to be wrong.

--charset
short form: -A; type: string

Default character set.

Enables character set settings in Perl and MySQL. If the value is "utf8", sets Perl's binmode on STDOUT to utf8, passes the "mysql_enable_utf8" option to DBD::mysql, and runs "SET NAMES UTF8" after connecting to MySQL. Any other value sets binmode on STDOUT without the utf8 layer, and runs "SET NAMES" after connecting to MySQL.

--database
short form: -D; type: string

Database to use for connection.

--defaults-file
short form: -F; type: string

Only read mysql options from the given file. You must give an absolute pathname.

--external
short form: -e

Calibrate, then pause while an external program runs.

This is typically useful while you run an external program. When you press [enter] mk-query-profiler will stop sleeping and take another measurement, then print statistics as usual.

When there is a filename on the command line, mk-query-profiler executes each line in the file as a shell command. If you give - as the filename, mk-query-profiler reads from STDIN.

Output from shell commands is printed to STDOUT and terminated with __BEGIN__, after which mk-query-profiler prints its own output.

--flush
short form: -f; cumulative: yes

Flush tables. Specify twice to do between every query.

Calls FLUSH TABLES before profiling. If you are executing queries from a batch file, specifying --flush twice will cause mk-query-profiler to call FLUSH TABLES between every query, not just once at the beginning. Default is not to flush at all. See <http://dev.mysql.com/doc/en/flush.html> for more information.

--host
short form: -h; type: string

Connect to host.

--innodb
short form: -i; negatable: yes; default: yes

Show InnoDB statistics.

--only
short form: -n; type: hash

Only show statistics for this comma-separated list of queries or commands.

--password
short form: -p; type: string

Password to use for connection.

--port
short form: -P; type: int

Port number to use for connection.

--separate
short form: -s

Print stats separately for each query.

The default is to show only the summary of the entire batch. See also ``--verbose''.

--session
negatable: yes; default: yes

Use session "SHOW STATUS" and "SHOW VARIABLES".

Disabled if the server version doesn't support it.

--setvars
type: string; default: wait_timeout=10000

Set these MySQL variables.

Specify any variables you want to be set immediately after connecting to MySQL. These will be included in a "SET" command.

--socket
short form: -S; type: string

Socket file to use for connection.

--tab
short form: -t

Print tab-separated values instead of whitespace-aligned columns.

--user
short form: -u; type: string

User for login if not current user.

--verbose
short form: -v; cumulative: yes; default: 0

Verbosity; specify multiple times for more detailed output.

When ``--tab'' is given, prints variables that don't change. Otherwise increasing the level of verbosity includes extra sections in the output.

--verify
short form: -r

Verify nothing else is accessing the server.

This is a weak verification; it simply calibrates twice (see ``--calibrate'') and verifies that the cost of observation remains constant.

DESCRIPTION

mk-query-profiler reads a file containing one or more SQL statements or shell commands, executes them, and analyzes the output of SHOW STATUS afterwards. It then prints statistics about how the batch performed. For example, it can show how many table scans the batch caused, how many page reads, how many temporary tables, and so forth.

All command-line arguments are optional, but you must either specify a file containing the batch to profile as the last argument, or specify that you're profiling an external program with the ``--external'' option, or provide input to STDIN.

If the file contains multiple statements, they must be separated by blank lines. If you don't do that, mk-query-profiler won't be able to split the file into individual queries, and MySQL will complain about syntax errors.

If the MySQL server version is before 5.0.2, you should make sure the server is completely unused before trying to profile a batch. Prior to this version, SHOW STATUS showed only global status variables, so other queries will interfere and produce false results. mk-query-profiler will try to detect if anything did interfere, but there can be no guarantees.

Prior to MySQL 5.0.2, InnoDB status variables are not available, and prior to version 5.0.3, InnoDB row lock status variables are not available. mk-query-profiler will omit any output related to these variables if they're not available.

For more information about SHOW STATUS, read the relevant section of the MySQL manual at <http://dev.mysql.com/doc/en/server-status-variables.html>

HOW TO INTERPRET


TAB-SEPARATED OUTPUT

If you specify ``--tab'', you will get the raw output of SHOW STATUS in tab-separated format, convenient for opening with a spreadsheet. This is not the default output, but it's so much easier to describe that I'll cover it first.

Most of the command-line options for controlling verbosity and such are ignored in --tab mode.
The variable names you see in MySQL, such as 'Com_select', are kept --- there are no euphimisms, so you have to know your MySQL variables.
The columns are Variable_name, Before, After1...AfterN, Calibration. The Variable_name column is just what it sounds like. Before is the result from the first run of SHOW STATUS. After1, After2, etc are the results of running SHOW STATUS after each query in the batch. Finally, the last column is the result of running SHOW STATUS just after the last AfterN column, so you can see how much work SHOW STATUS itself causes.
If you specify ``--verbose'', output includes every variable mk-query-profiler measures. If not (default) it only includes variables where there was some difference from one column to the next.

NORMAL OUTPUT

If you don't specify --tab, you'll get a report formatted for human readability. This is the default output format.

mk-query-profiler can output a lot of information, as you've seen if you ran the examples in the ``SYNOPSIS''. What does it all mean?

First, there are two basic groups of information you might see: per-query and summary. If your batch contains only one query, these will be the same and you'll only see the summary. You can recognize the difference by looking for centered, all-caps, boxed-in section headers. Externally profiled commands will have EXTERNAL, individually profiled queries will have QUERY, and summary will say SUMMARY.

Next, the information in each section is grouped into subsections, headed by an underlined title. Each of these sections has varying information in it. Which sections you see depends on command-line arguments and your MySQL version. I'll explain each section briefly. If you really want to know where the numbers come from, read <http://dev.mysql.com/doc/en/server-status-variables.html>.

You need to understand which numbers are insulated from other queries and which are not. This depends on your MySQL version. Version 5.0.2 introduced the concept of session status variables, so you can see information about only your own connection. However, many variables aren't session-ized, so when you have MySQL 5.0.2 or greater, you will actually see a mix of session and global variables. That means other queries happening at the same time will pollute some of your results. If you have MySQL versions older than 5.0.2, you won't have ANY connection-specific stats, so your results will be polluted by other queries no matter what. Because of the mixture of session and global variables, by far the best way to profile is on a completely quiet server where nothing else is interfering with your results.

While explaining the results in the sections that follow, I'll refer to a value as ``protected'' if it comes from a session-specific variable and can be relied upon to be accurate even on a busy server. Just keep in mind, if you're not using MySQL 5.0.2 or newer, your results will be inaccurate unless you're running against a totally quiet server, even if I label it as ``protected.''

Overall stats

This section shows the overall elapsed time for the query, as measured by Perl, and the optimizer cost as reported by MySQL.

If you're viewing separate query statistics, this is all you'll see. If you're looking at a summary, you'll also see a breakdown of the questions the queries asked the server.

The execution time is not totally reliable, as it includes network round-trip time, Perl's own execution time, and so on. However, on a low-latency network, this should be fairly negligible, giving you a reasonable measure of the query's time, especially for queries longer than a few tenths of a second.

The optimizer cost comes from the Last_query_cost variable, and is protected from other connections in MySQL 5.0.7 and greater. It is not available before 5.0.1.

The total number of questions is not protected, but the breakdown of individual question types is, because it comes from the Com_ status variables.

Table and index accesses

This section shows you information about the batch's table and index-level operations (as opposed to row-level operations, which will be in the next section). The ``Table locks acquired'' and ``Temp files'' values are unprotected, but everything else in this section is protected.

The ``Potential filesorts'' value is calculated as the number of times a query had both a scan sort (Sort_scan) and created a temporary table (Created_tmp_tables). There is no Sort_filesort or similar status value, so it's a best guess at whether a query did a filesort. It should be fairly accurate.

If you specified ``--allowcache'', you'll see statistics on the query cache. These are unprotected.

Row operations

These values are all about the row-level operations your batch caused. For example, how many rows were inserted, updated, or deleted. You'll also see row-level index access statistics, such as how many times the query sought and read the next entry in an index.

Depending on your MySQL version, you'll either see one or two columns of information in this section. The one headed ``Handler'' is all from the Handler_ variables, and those statistics are protected. If your MySQL version supports it, you'll also see a column headed ``InnoDB,'' which is unprotected.

I/O Operations

This section gives information on I/O operations your batch caused, both in memory and on disk. Unless you have MySQL 5.0.2 or greater, you'll only see information on the key cache. Otherwise, you'll see a lot of information on InnoDB's I/O operations as well, such as how many times the query was able to satisfy a read from the buffer pool and how many times it had to go to the disk.

None of the information in this section is protected.

InnoDB Data Operations

This section only appears when you're querying MySQL 5.0.2 or newer. None of the information is protected. You'll see statistics about how many pages were affected, how many operations took place, and how many bytes were affected.

ENVIRONMENT

The environment variable "MKDEBUG" enables verbose debugging output in all of the Maatkit tools:
    MKDEBUG=1 mk-....
 
 

BUGS

Please use Google Code Issues and Groups to report bugs or request support: <http://code.google.com/p/maatkit/>.

Please include the complete command-line used to reproduce the problem you are seeing, the version of all MySQL servers involved, the complete output of the tool when run with ``--version'', and if possible, debugging output produced by running with the "MKDEBUG=1" environment variable.

SYSTEM REQUIREMENTS

You need Perl, DBI, DBD::mysql, and some core modules.

SEE ALSO

See also mk-profile-compact.

LICENSE

This program is copyright 2007-2008 Baron Schwartz. Feedback and improvements are welcome.

THIS PROGRAM IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.

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; OR the Perl Artistic License. On UNIX and similar systems, you can issue `man perlgpl' or `man perlartistic' to read these licenses.

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.

AUTHOR

Baron Schwartz.

ACKNOWLEDGEMENTS

I was inspired by the wonderful mysqlreport utility available at <http://www.hackmysql.com/>.

Other contributors: Bart van Bragt.

Thanks to all who have helped.

VERSION

This manual page documents Ver 1.1.13 Distrib 2725 $Revision: 2311 $.