6.6.2 innochecksum — Offline InnoDB File Checksum Utility

innochecksum prints checksums for InnoDB files. This tool reads an InnoDB tablespace file, calculates the checksum for each page, compares the calculated checksum to the stored checksum, and reports mismatches, which indicate damaged pages. It was originally developed to speed up verifying the integrity of tablespace files after power outages but can also be used after file copies. Because checksum mismatches cause InnoDB to deliberately shut down a running server, it may be preferable to use this tool rather than waiting for an in-production server to encounter the damaged pages.

innochecksum cannot be used on tablespace files that the server already has open. For such files, you should use CHECK TABLE to check tables within the tablespace. Attempting to run innochecksum on a tablespace that the server already has open results in an Unable to lock file error.

If checksum mismatches are found, restore the tablespace from backup or start the server and attempt to use mysqldump to make a backup of the tables within the tablespace.

Invoke innochecksum like this:

innochecksum [options] file_name

innochecksum Options

innochecksum supports the following options. For options that refer to page numbers, the numbers are zero-based.

• --help, -?

  Command-Line Format	--help
  Type	Boolean
  Default Value	false

  Displays command line help. Example usage:

  innochecksum --help

• --info, -I

  Command-Line Format	--info
  Type	Boolean
  Default Value	false

  Synonym for --help. Displays command line help. Example usage:

  innochecksum --info

• --version, -V

  Command-Line Format	--version
  Type	Boolean
  Default Value	false

  Displays version information. Example usage:

  innochecksum --version

• --verbose, -v

  Command-Line Format	--verbose
  Type	Boolean
  Default Value	false

  Verbose mode; prints a progress indicator to the log file every five seconds. In order for the progress indicator to be printed, the log file must be specified using the --log option. To turn on verbose mode, run:

  innochecksum --verbose

  To turn off verbose mode, run:

  innochecksum --verbose=FALSE

  The --verbose option and --log option can be specified at the same time. For example:

  innochecksum --verbose --log=/var/lib/mysql/test/logtest.txt

  To locate the progress indicator information in the log file, you can perform the following search:

  cat ./logtest.txt | grep -i "okay"

  The progress indicator information in the log file appears similar to the following:

  page 1663 okay: 2.863% done
  page 8447 okay: 14.537% done
  page 13695 okay: 23.568% done
  page 18815 okay: 32.379% done
  page 23039 okay: 39.648% done
  page 28351 okay: 48.789% done
  page 33023 okay: 56.828% done
  page 37951 okay: 65.308% done
  page 44095 okay: 75.881% done
  page 49407 okay: 85.022% done
  page 54463 okay: 93.722% done
  ...

• --count, -c

  Command-Line Format	--count
  Type	Base name
  Default Value	true

  Print a count of the number of pages in the file and exit. Example usage:

  innochecksum --count ../data/test/tab1.ibd

• --start-page=num, -s num

  Command-Line Format	--start-page=#
  Type	Numeric
  Default Value	0

  Start at this page number. Example usage:

  innochecksum --start-page=600 ../data/test/tab1.ibd

  or:

  innochecksum -s 600 ../data/test/tab1.ibd

• --end-page=num, -e num

  Command-Line Format	--end-page=#
  Type	Numeric
  Default Value	0
  Minimum Value	0
  Maximum Value	18446744073709551615

  End at this page number. Example usage:

  innochecksum --end-page=700 ../data/test/tab1.ibd

  or:

  innochecksum --p 700 ../data/test/tab1.ibd

• --page=num, -p num

  Command-Line Format	--page=#
  Type	Integer
  Default Value	0

  Check only this page number. Example usage:

  innochecksum --page=701 ../data/test/tab1.ibd

• --strict-check, -C

  Command-Line Format	--strict-check=algorithm
  Type	Enumeration
  Default Value	crc32
  Valid Values	innodb crc32 none

  Specify a strict checksum algorithm. Options include innodb, crc32, and none.

  In this example, the innodb checksum algorithm is specified:

  innochecksum --strict-check=innodb ../data/test/tab1.ibd

  In this example, the crc32 checksum algorithm is specified:

  innochecksum -C crc32 ../data/test/tab1.ibd

  The following conditions apply:

  • If you do not specify the --strict-check option, innochecksum validates against innodb, crc32 and none.

  • If you specify the none option, only checksums generated by none are allowed.

  • If you specify the innodb option, only checksums generated by innodb are allowed.

  • If you specify the crc32 option, only checksums generated by crc32 are allowed.

• --no-check, -n

  Command-Line Format	--no-check
  Type	Boolean
  Default Value	false

  Ignore the checksum verification when rewriting a checksum. This option may only be used with the innochecksum --write option. If the --write option is not specified, innochecksum terminates.

  In this example, an innodb checksum is rewritten to replace an invalid checksum:

  innochecksum --no-check --write innodb ../data/test/tab1.ibd

• --allow-mismatches, -a

  Command-Line Format	--allow-mismatches=#
  Type	Integer
  Default Value	0
  Minimum Value	0
  Maximum Value	18446744073709551615

  The maximum number of checksum mismatches allowed before innochecksum terminates. The default setting is 0. If --allow-mismatches=N, where N>=0, N mismatches are permitted and innochecksum terminates at N+1. When --allow-mismatches is set to 0, innochecksum terminates on the first checksum mismatch.

  In this example, an existing innodb checksum is rewritten to set --allow-mismatches to 1.

  innochecksum --allow-mismatches=1 --write innodb ../data/test/tab1.ibd

  With --allow-mismatches set to 1, if there is a mismatch at page 600 and another at page 700 on a file with 1000 pages, the checksum is updated for pages 0-599 and 601-699. Because --allow-mismatches is set to 1, the checksum tolerates the first mismatch and terminates on the second mismatch, leaving page 600 and pages 700-999 unchanged.

• --write=name, -w num

  Command-Line Format	--write=algorithm
  Type	Enumeration
  Default Value	crc32
  Valid Values	innodb crc32 none

  Rewrite a checksum. When rewriting an invalid checksum, the --no-check option must be used together with the --write option. The --no-check option tells innochecksum to ignore verification of the invalid checksum. You do not have to specify the --no-check option if the current checksum is valid.

  An algorithm must be specified when using the --write option. Possible values for the --write option are:

  • innodb: A checksum calculated in software, using the original algorithm from InnoDB.

  • crc32: A checksum calculated using the crc32 algorithm, possibly done with a hardware assist.

  • none: A constant number.

  The --write option rewrites entire pages to disk. If the new checksum is identical to the existing checksum, the new checksum is not written to disk in order to minimize I/O.

  innochecksum obtains an exclusive lock when the --write option is used.

  In this example, a crc32 checksum is written for tab1.ibd:

  innochecksum -w crc32 ../data/test/tab1.ibd

  In this example, a crc32 checksum is rewritten to replace an invalid crc32 checksum:

  innochecksum --no-check --write crc32 ../data/test/tab1.ibd

• --page-type-summary, -S

  Command-Line Format	--page-type-summary
  Type	Boolean
  Default Value	false

  Display a count of each page type in a tablespace. Example usage:

  innochecksum --page-type-summary ../data/test/tab1.ibd

  Sample output for --page-type-summary:

  File::../data/test/tab1.ibd
  ================PAGE TYPE SUMMARY==============
  #PAGE_COUNT PAGE_TYPE
  ===============================================
         2        Index page
         0        Undo log page
         1        Inode page
         0        Insert buffer free list page
         2        Freshly allocated page
         1        Insert buffer bitmap
         0        System page
         0        Transaction system page
         1        File Space Header
         0        Extent descriptor page
         0        BLOB page
         0        Compressed BLOB page
         0        Other type of page
  ===============================================
  Additional information:
  Undo page type: 0 insert, 0 update, 0 other
  Undo page state: 0 active, 0 cached, 0 to_free, 0 to_purge, 0 prepared, 0 other

• --page-type-dump, -D

  Command-Line Format	--page-type-dump=name
  Type	String
  Default Value	[none]

  Dump the page type information for each page in a tablespace to stderr or stdout. Example usage:

  innochecksum --page-type-dump=/tmp/a.txt ../data/test/tab1.ibd

• --log, -l

  Command-Line Format	--log=path
  Type	File name
  Default Value	[none]

  Log output for the innochecksum tool. A log file name must be provided. Log output contains checksum values for each tablespace page. For uncompressed tables, LSN values are also provided. The --log replaces the --debug option, which was available in earlier releases. Example usage:

  innochecksum --log=/tmp/log.txt ../data/test/tab1.ibd

  or:

  innochecksum -l /tmp/log.txt ../data/test/tab1.ibd

• - option.

  Specify the - option to read from standard input. If the - option is missing when “read from standard in” is expected, innochecksum prints innochecksum usage information indicating that the “-” option was omitted. Example usages:

  cat t1.ibd | innochecksum -

  In this example, innochecksum writes the crc32 checksum algorithm to a.ibd without changing the original t1.ibd file.

  cat t1.ibd | innochecksum --write=crc32 - > a.ibd

Running innochecksum on Multiple User-defined Tablespace Files

The following examples demonstrate how to run innochecksum on multiple user-defined tablespace files (.ibd files).

Run innochecksum for all tablespace (.ibd) files in the “test” database:

innochecksum ./data/test/*.ibd

Run innochecksum for all tablespace files (.ibd files) that have a file name starting with “t”:

innochecksum ./data/test/t*.ibd

Run innochecksum for all tablespace files (.ibd files) in the data directory:

innochecksum ./data/*/*.ibd

Note

Running innochecksum on multiple user-defined tablespace files is not supported on Windows operating systems, as Windows shells such as cmd.exe do not support glob pattern expansion. On Windows systems, innochecksum must be run separately for each user-defined tablespace file. For example:

innochecksum.exe t1.ibd
innochecksum.exe t2.ibd
innochecksum.exe t3.ibd

Running innochecksum on Multiple System Tablespace Files

By default, there is only one InnoDB system tablespace file (ibdata1) but multiple files for the system tablespace can be defined using the innodb_data_file_path option. In the following example, three files for the system tablespace are defined using the innodb_data_file_path option: ibdata1, ibdata2, and ibdata3.

./bin/mysqld --no-defaults --innodb-data-file-path="ibdata1:10M;ibdata2:10M;ibdata3:10M:autoextend"

The three files (ibdata1, ibdata2, and ibdata3) form one logical system tablespace. To run innochecksum on multiple files that form one logical system tablespace, innochecksum requires the - option to read tablespace files in from standard input, which is equivalent to concatenating multiple files to create one single file. For the example provided above, the following innochecksum command would be used:

cat ibdata* | innochecksum -

Refer to the innochecksum options information for more information about the “-” option.

Note

Running innochecksum on multiple files in the same tablespace is not supported on Windows operating systems, as Windows shells such as cmd.exe do not support glob pattern expansion. On Windows systems, innochecksum must be run separately for each system tablespace file. For example:

innochecksum.exe ibdata1
innochecksum.exe ibdata2
innochecksum.exe ibdata3