pt-upgrade - Verify that query results are identical on different servers.
pt-upgrade [OPTIONS] LOGS|RESULTS DSN [DSN]
pt-upgrade executes queries in the given MySQL
LOGS on each
compares the results, and reports any significant differences. The tool can
also save the results for later analyses.
LOGS can be slow, general,
binary, tcpdump, and “raw”.
Compare host2 to host1 using queries in
pt-upgrade h=host1 h=host2 slow.log
Compare host2 to saved results from host1:
pt-upgrade h=host1 --save-results host1_results/ slow.log pt-upgrade host1_results1/ h=host2
Percona Toolkit is mature, proven in the real world, and well tested, but all database tools can pose a risk to the system and the database server. Before using this tool, please:
pt-upgrade helps determine if it is safe to upgrade (or downgrade) to a new version of MySQL. A safe and conservative upgrade plan has several steps, one of which is ensuring that queries will produce identical results on the new version of MySQL.
pt-upgrade executes queries from slow, general, binary, tcpdump, and “raw” logs on two servers, compares many aspects of each query’s exeuction and results, and reports any signficant differences. The two servers are typically development servers, one running the current production version of MySQL and the other running the new version of MySQL.
pt-upgrade has two use cases. The first, canonical case is running “host to host”. A log file and two DSN are given on the command line, one for each MySQL server. See the first example in the “SYNOPSIS”. Queries are executed and compared on each server as the tool runs. Queries with differences are printed as the tool runs, or when it finishes (see “WHEN QUERIES ARE REPORTED”). Nothing is saved to disk, so this use case requires less hard disk space, but the queries must be executed on both servers if the tool is ran again, even if one of the servers hasn’t changed. If there are a lot of queries or executing them takes a long time, and one server doesn’t change, you may want to use the second use case.
The second use case is running “reference results to host”. Reference
results are the complete results from a single MySQL server, saved to
disk. In this case, you must first generate the reference results
--save-results, then run the tool a second time to compare
another MySQL server to the results. See the second example in the
“SYNOPSIS”. Results are typically generated for the current version
of MySQL which doesn’t change. This use case can require a lot of
disk space because the results (i.e. rows) for all queries must be saved,
plus other data about the queries. If you plan to do many comparisons
against a fixed version of MySQL, this use case is more efficient. Or if
you don’t have access to both servers at the same time, this use case
allows you to “execute now, compare later”.
Consistent environments and consistent data are crucial for obtaining an accurate report. pt-upgrade should never be ran on a production server or any active server because there is no easy way to ensure a synchronous read for each query. If data is changing on either server while pt-upgrade is running, the report could contain more false-positives than legitimate differences. ** pt-upgrade assumes that both MySQL servers are static, unchanging (except for any changes made by the tool if ran with ``–no-read-only``).** A read-only workload shouldn’t affect the tool, except maybe query times, so read-only slaves could be used.
In a host to host comparison, results from the first host establish the norm to which results from the second host are compared. In a reference results to host comparison, the reference results are the norm to which the host is compared. Comparative phrases like “smaller than”, “better than”, etc. mean compared to the norm.
For example, if the query time for an event is
0.01 on the first host
0.5 on the second host, that is a significant difference because
0.5 is worse than
0.1, and so the query will be reported.
By default, pt-upgrade only executes
(This does not include ‘SELECT...INTO’ statements, which do not return
rows but dump output to a file or variable.)
If you’re using recreatable test or development servers and wish to
compare write statements too (e.g.
--no-read-only. If using a binary log, you must
--no-read-only because binary logs don’t contain
The tool does not create its own transactions, but any transactions in
LOG are executed as-is. Since logs are serial, transactions
shouldn’t normally be an issue. If, however, you need to compare queries
that are somehow transactionally related (in which case you probably
also need to disable
--[no]read-only), then pt-upgrade probably
won’t do what you need because it’s not designed for this purpose.
pt-upgrade runs with
autocommit=1 by default.
pt-upgrade has no throttling options because the tool should only be ran on dedicated testing or development servers. Do not run pt-upgrade on production servers! Consequently, the tool is CPU, memory, disk, and network intensive. It executes queries as fast as possible.
Signficant query differences are determined by comparing these aspects of each query from both hosts:
The number of rows returned by the query should be the same. This is reported as “missing rows” under “Row diffs”.
The row data returned by the query should be the same. All differences are significant: whitespace, float-precision, etc.
The query should either not produce any errors or warnings, or produce the same errors or warnings.
A query rarely executes with a constant time, but its execution time should be within the same order of magnitude or smaller.
If a query causes a SQL error on only one host, this is reported as “Query errors”. Since the query works on one host, its syntax is probably valid, and the error is due to some condition unique to the other host.
If a query causes a SQL error on both hosts, this is reported as “SQL errors”. The SQL syntax of the query could be invalid.
As pt-upgrade runs, it prints queries with differences as soon as it can (see “WHEN QUERIES ARE REPORTED”). To prevent the report from becoming too long, queries are not reported individually but grouped by fingerprint into classes. A query fingerprint is the abstracted form of a query, created by removing literal values, normalizing whitespace, etc. So these queries belong to the same class:
SELECT c FROM t WHERE id = 1 SELECT c FROM t WHERE id=5 select c from t where id = 9
The fingerprint for those queries is:
select c from t where id=?
Each query class can have up to
--max-class-size unique queries
(1,000 by default). Up to
--max-examples are reported for each
type of difference, per query class. By virtue of being in the same class,
an example of one query’s difference is usually representative of all queries
with the same difference, so it’s not necessary to report every example.
The total number of queries in a class with a particular difference is
indicated in the report.
#----------------------------------------------------------------------- # Logs #----------------------------------------------------------------------- File: /opt/mysql/slow.log Size: 59700 #----------------------------------------------------------------------- # Hosts #----------------------------------------------------------------------- host1: DSN: h=127.1,P=12345 hostname: dev1 MySQL: MySQL 5.1.68 host2: DSN: h=127.1,P=12348 hostname: dev2 MySQL: MySQL 5.5.10 ######################################################################## # Query class AAD020567F8398EE ######################################################################## Reporting class because it has diffs, but hasn't been reported yet. Total queries 1 Unique queries 1 Discarded queries 0 insert into t (id, username) values(?+) ## ## Warning diffs: 1 ## -- 1. Code: 1265 Level: Warning Message: Data truncated for column 'username' at row 1 vs. No warning 1265 INSERT INTO t (id, username) VALUES (NULL, 'long_username') #----------------------------------------------------------------------- # Stats #----------------------------------------------------------------------- failed_queries 0 not_select 0 queries_filtered 0 queries_no_diffs 0 queries_read 1 queries_with_diffs 1 queries_with_errors 0
The “Query class <ID>” sections are the most important because they list “QUERY DIFFERENCES”. The first part of the section lists the reason why the query class was report, followed by counts of queries in the class, followed by the fingerprint which defines the class.
The rest of the query class section lists the “QUERY DIFFERENCES” that
caused the class to be reported. Each type of difference begins with
a double hash mark header that lists the type and total number of queries
in the class with the difference. Then up to
--max-examples are listed,
numbered “– 1.”, “— 2.”, etc. Each example lists the difference for
the first and second hosts (respective to the “Hosts” section), followed by
the first SQL statement that revealed the difference.
A query class is reported as soon as any one of the “QUERY DIFFERENCES”
or query errors has
--max-examples. Else, all queries with differences
are reported when the tool finishes.
For example, if two query time differences are found for a query class, it is not reported yet. Once a third query time diffence is found, the query class is reported, including any other differences that may have been found too. Queries for the class will continue to be executed, but the class will not be reported again.
The “REPORT” is printed to STDOUT as the tool runs. Internal warnings,
--progress are printed to STDERR. To keep the two separate,
run the tool like:
pt-upgrade ... 1>report 2>err &
tail -f err while the tool is running to track its
In general, the tool exits zero if it finishes normally and there were no internal warnings or errors, and no “QUERY DIFFERENCES” were found. Else the tool exits non-zero with one or more of the following codes:
There were too many internal errors or warnings; see STDERR. See also
There were “QUERY DIFFERENCES”; see the “REPORT”.
--run-timeexpired; the tool did not finish reading the logs or reference results.
Other exit codes indicate that the tool crashed or died unexpectedly. The error that caused this should have printed to STDERR.
To check for a particular exit code, logical
&) the final exit
status with the exit code. For example, exit status 5 implies codes 1 and 4
5 & 1 is true, and
5 & 4 is true.
This tool accepts additional command-line arguments. Refer to the “SYNOPSIS” and usage information for details.
Prompt for a password when connecting to MySQL.
short form: -A; type: string
Default character set. 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.
Read this comma-separated list of config files; if specified, this must be the first option on the command line.
Continue parsing even if there is an error. The tool will not continue forever: it stops after 100 errors, in which case there is probably a bug in the tool or the input is invalid.
Fork to the background and detach from the shell. POSIX operating systems only.
short form: -D; type: string
Default database when connecting to MySQL.
short form: -F; type: string
Only read MySQL options from the given file. You must give an absolute pathname.
SET SESSION query_cache_type = OFF to disable the query cache.
Run but do not execute or compare queries. This is useful for checking command line options, connections to MySQL, and log or reference results parsing.
Allow events for which this Perl code returns true.
See the same option in the documentation for pt-query-digest.
Show help and exit.
short form: -h; type: string
MySQL hostname or IP.
Ignore these MySQL warning codes when comparing warnings.
Print STDOUT and STDERR to this file when daemonized. This option
only takes affect when
--daemonize is specified. The file is created
if it doesn’t exist, else output is appended to it.
type: int; default: 1000
Max number of unique queries in each query class. See “REPORT”.
type: int; default: 3
Max number of examples to list for each “QUERY DIFFERENCES”. A query class is reported as soon as this many examples for any type of query difference are found.
Create the given PID file. The tool won’t start if the PID file already exists and the PID it contains is different than the current PID. However, if the PID file exists and the PID it contains is no longer running, the tool will overwrite the PID file with the current PID. The PID file is removed automatically when the tool exits.
short form: -P; type: int
MySQL port number.
type: array; default: time,30
Print progress reports to STDERR. The tool prints progress reports while reading logs or reference results, roughly estimating how long until it finishes.
The value is a comma-separated list with two parts. The first part can be percentage, time, or iterations; the second part specifies how often an update should be printed, in percentage, seconds, or number of iterations.
SET statements. If
specified, all queries are exeucted:
Even when running in default read-only mode, you should use a MySQL user
SELECT privileges to insure against bugs in the tool.
type: Hash; default: hosts, logs, queries, stats
Print these sections of the “REPORT”.
How long to run before exiting. By default, the tool runs until it finishes reading the logs or reference results.
Save reference results to this directory. This option works only when one DSN is specified, to generate reference results. When comparing a host to reference results, specify its results directory instead of its DSN. See the second example in the “SYNOPSIS”.
Reference results can use a lot of disk space.
Set the MySQL variables in this comma-separated list of
By default, the tool sets:
Variables specified on the command line override these defaults. For
--set-vars wait_timeout=500 overrides the defaultvalue of
The tool prints a warning and continues if a variable cannot be set.
short form: -S; type: string
Socket file to use for connection.
type: string; default: slowlog
Type of log files. Valid types are:
VALUE LOG TYPE ======= =========================================== slowlog MySQL slow log genlog MySQL general log binlog MySQL binary log (converted by mysqlbinlog) rawlog Custom log with one SQL statement per line
type: string; default: percona_schema.pt_upgrade
Use this table to clear warnings. To clear all warnings from previous
queries, pt-upgrade executes
SELECT * FROM --upgrade-table LIMIT 1
on each host before executing each query.
The table must be database-qualified. The database and table are
automatically created unless
--no-create-upgrade-table is specified
--[no]create-upgrade-table). If the table does not already
exist, it is created with this definition:
CREATE TABLE pt_upgrade ( id INT NOT NULL PRIMARY KEY )
short form: -u; type: string
MySQL user if not the current system user.
Show version and exit.
Check for the latest version of Percona Toolkit, MySQL, and other programs.
This is a standard “check for updates automatically” feature, with two additional features. First, the tool checks the version of other programs on the local system in addition to its own version. For example, it checks the version of every MySQL server it connects to, Perl, and the Perl module DBD::mysql. Second, it checks for and warns about versions with known problems. For example, MySQL 5.5.25 had a critical bug and was re-released as 5.5.25a.
Any updates or known problems are printed to STDOUT before the tool’s normal output. This feature should never interfere with the normal operation of the tool.
For more information, visit https://www.percona.com/version-check.
Parse only events for this IP:port for
--type tcpdump. All other
IP addresses are ignored. If not specified, pt-upgrade watches all servers
by looking for any IP address using port 3306 or “mysql”. If you’re watching
a server with a non-standard port, this won’t work, so you must specify the
IP address and port to watch.
If you want to watch a mix of servers, some running on standard port 3306 and some running on non-standard ports, you need to create separate tcpdump outputs for the non-standard port servers and then specify this option for each. At present pt-upgrade cannot auto-detect servers on port 3306 and also be told to watch a server on a non-standard port.
These DSN options are used to create a DSN. Each option is given like
option=value. The options are case-sensitive, so P and p are not the
same option. There cannot be whitespace before or after the
if the value contains whitespace it must be quoted. DSN options are
comma-separated. See the percona-toolkit manpage for full details.
dsn: charset; copy: yes
Default character set.
dsn: database; copy: yes
dsn: mysql_read_default_file; copy: yes
Only read default options from the given file
dsn: host; copy: yes
Connect to host.
Explicitly enable LOAD DATA LOCAL INFILE.
For some reason, some vendors compile libmysql without the –enable-local-infile option, which disables the statement. This can lead to weird situations, like the server allowing LOCAL INFILE, but the client throwing exceptions if it’s used.
However, as long as the server allows LOAD DATA, clients can easily re-enable it; See https://dev.mysql.com/doc/refman/5.0/en/load-data-local.html and http://search.cpan.org/~capttofu/DBD-mysql/lib/DBD/mysql.pm. This option does exactly that.
Although we’ve not found a case where turning this option leads to errors or differing behavior, to be on the safe side, this option is not on by default.
dsn: password; copy: yes
Password to use when connecting. If password contains commas they must be escaped with a backslash: “exam,ple”
dsn: port; copy: yes
Port number to use for connection.
dsn: mysql_socket; copy: yes
Socket file to use for connection.
dsn: user; copy: yes
User for login if not current user.
The environment variable
PTDEBUG enables verbose debugging output to STDERR.
To enable debugging and capture all output to a file, run the tool like:
PTDEBUG=1 pt-upgrade ... > FILE 2>&1
Be careful: debugging output is voluminous and can generate several megabytes of output.
You need Perl, DBI, DBD::mysql, and some core packages that ought to be installed in any reasonably new version of Perl.
For a list of known bugs, see http://www.percona.com/bugs/pt-upgrade.
Please report bugs at https://bugs.launchpad.net/percona-toolkit. Include the following information in your bug report:
If possible, include debugging output by running the tool with
Visit http://www.percona.com/software/percona-toolkit/ to download the latest release of Percona Toolkit. Or, get the latest release from the command line:
wget percona.com/get/percona-toolkit.tar.gz wget percona.com/get/percona-toolkit.rpm wget percona.com/get/percona-toolkit.deb
You can also get individual tools from the latest release:
TOOL with the name of any tool.
This tool is part of Percona Toolkit, a collection of advanced command-line tools for MySQL developed by Percona. Percona Toolkit was forked from two projects in June, 2011: Maatkit and Aspersa. Those projects were created by Baron Schwartz and primarily developed by him and Daniel Nichter. Visit http://www.percona.com/software/ to learn about other free, open-source software from Percona.
This program is copyright 2009-2017 Percona LLC and/or its affiliates. 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 MERCHANTABILITY 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.
For general inquiries, please send us your question and someone will contact you.