| Bug #11527 | Setting myisam_repair_threads to >1 leads to corruption | ||
|---|---|---|---|
| Submitted: | 23 Jun 2005 12:33 | Modified: | 4 Mar 2006 17:20 |
| Reporter: | Geert Vanderkelen | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server: MyISAM storage engine | Severity: | S3 (Non-critical) |
| Version: | 4.1.12 | OS: | Any (*) |
| Assigned to: | Ingo Strüwing | CPU Architecture: | Any |
| Tags: | corruption, myisam, myisam_repair_threads | ||
[30 Jan 2006 14:00]
Ingo Strüwing
Funny that you got different results for different numbers of myisam_repair_threads. The current code (4.1 .. 5.1) uses the variable just to check if it is > 1. It starts one thread per index then.
[31 Jan 2006 9:49]
Ingo Strüwing
In this test case there is ony one index. I proved by tracing that one thread is created only. Thus I can exclude threading/locking problems.
[9 Feb 2006 13:30]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/2364
[16 Feb 2006 18:46]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/2745
[4 Mar 2006 15:17]
Ingo Strüwing
Pushed to 4.1.19, 5.0.20, and 5.1.8. No test case. It requires >4GB data to repeat. I tested with a 7GB customer load file.
[4 Mar 2006 17:20]
Paul DuBois
Noted in 4.1.19, 5.0.19, 5.1.8 changelogs. Setting the <literal>myisam_repair_threads</literal> system variable to a value larger than 1 could cause corruption of large <literal>MyISAM</literal> tables. (Bug #11527)
Description: Hi, Setting the myisam_repair_threads to another value than the default 1, can or will lead to table corruption when adding an index for example. It has been tested under Linux and Windows, versions 4.1.12 and 5.0.7-beta. Dual CPU machines give different results under Windows than under Linux. On Windows using a Dual (or more) CPU, a value of 2 gives also corruption, where as on a Dual Xeon under Linux it works with 2, but not with value 3 or 4 or 5. Regards, Geert How to repeat: Need alot of data, note the table MyISAM grows over 4Gb as it is Fixed Rows. CREATE TABLE `bigtable` ( `ModelpointId` int(10) unsigned NOT NULL default '0', `TimestepId` datetime NOT NULL default '0000-00-00 00:00:00', `Var1` float default NULL, `Var2` float default NULL, `Var3` float default NULL, `Var4` float default NULL, `Var5` float default NULL ) ENGINE=MyISAM DEFAULT CHARSET=latin1; mysql> show variables like 'myisam_repair_threads'; +-----------------------+-------+ | Variable_name | Value | +-----------------------+-------+ | myisam_repair_threads | 2 | +-----------------------+-------+ 1 row in set (0.00 sec) mysql> load data infile 'bigtable_csvfile.txt' INTO TABLE bigtable FIELDS TERMINATED BY ',' OPTIONALLY ENCLOSED BY '"' LINES TERMINATED BY '\n'; Query OK, 130771200 rows affected (11 min 46.32 sec) Records: 130771200 Deleted: 0 Skipped: 0 Warnings: 0 mysql> create index ix1 on bigtable (ModelPointId); Query OK, 130771200 rows affected (5 min 38.43 sec) Records: 130771200 Duplicates: 0 Warnings: 0 mysql> show table status like 'bigtable'\G *************************** 1. row *************************** Name: bigtable Engine: MyISAM Version: 9 Row_format: Fixed Rows: 620675 Avg_row_length: 33 Data_length: 20482275 Max_data_length: 141733920767 Index_length: 5062656 Data_free: 0 Auto_increment: NULL Create_time: 2005-06-21 21:48:50 Update_time: 2005-06-21 21:54:26 Check_time: 2005-06-21 21:54:28 Collation: latin1_swedish_ci Checksum: NULL Create_options: Check table results in corrupted table. Has been verfied by 3 parties on different machine using different Linux distro's and different Windows versions.