Bug #12187 | LOAD TABLE FROM MASTER unable to load a certain MyISAM table | ||
---|---|---|---|
Submitted: | 26 Jul 2005 21:30 | Modified: | 31 Aug 2006 11:50 |
Reporter: | Guilhem Bichot | Email Updates: | |
Status: | Won't fix | Impact on me: | |
Category: | MySQL Server: Replication | Severity: | S2 (Serious) |
Version: | 5.0 | OS: | Solaris (Solaris) |
Assigned to: | Andrei Elkin | CPU Architecture: | Any |
[26 Jul 2005 21:30]
Guilhem Bichot
[1 Feb 2006 22:38]
Richard Sonnen
We can replicate this on multiple slaves running FreeBSD 6/MySQL 5.0.18, and FreeBSD 4.11/MySQL 5.0.18 against a master running FreeBSD 4.6.2/MySQL 4.0.22. Of 5 replicated production tables, 2 load, then drop to zero length with a 'Found wrong stored record at 0' error in the logs. This occurs with either 'LOAD DATA FROM MASTER' or 'LOAD TABLE <tablename> FROM MASTER'. Does put a bit of a kink in the upgrade path.
[17 Mar 2006 22:09]
Patrick Feliciano
I've tested 5.0.18 to 5.0.18 and it has the same issue.
[18 Mar 2006 21:17]
Steve Slater
5.1.7 beta has the same bug. Using the mysql-compiled RPMs.
[27 Apr 2006 3:55]
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/5621
[27 Apr 2006 3:56]
Greg Lehey
Created test case (rpl_load_data_from_master). Under 5.0/5.0, it was not possible to reproduce this behaviour (i.e. the test case passed).
[1 May 2006 4:27]
Greg Lehey
Test run on FreeBSD as well, since that was the original platform. Again there was no problem.
[8 May 2006 9:34]
Paul Abel
We are experiencing this bug too using 5.0.18 as both master and slave, running on Solaris 9
[15 May 2006 10:55]
Michael Vychizhanin
Got the same issue during copy from 4.0.27 -master (x86 Solaris) to 5.0.19 -slave (FreeBSD)
[15 May 2006 10:57]
Michael Vychizhanin
The same one from 4.0.27 (x86 Solaris 9) to 5.0.21 (FreeBSD 5.4). =(
[22 May 2006 2:36]
Greg Lehey
It seems that this problem still appears in conjunction with Solaris, so I'm reopening this bug and changing the OS to Solaris. It was set to "Linux", which doesn't match the problem description.
[9 Jun 2006 11:41]
Julien Huang
I've tested this and experiencing the same issue. - Master (Debian Sarge 32bits) $ uname -a Linux ns4 2.4.31 #2 SMP Mon Jul 18 17:25:56 CEST 2005 i686 GNU/Linux $ mysql --version mysql Ver 14.7 Distrib 4.1.15, for pc-linux-gnu (i686) using readline 4.3 - Slave (Ubuntu 5.10 64bits) $ uname -a Linux gina 2.6.12-10-amd64-generic #1 Fri Apr 28 13:18:42 UTC 2006 x86_64 GNU/Linux $ mysql --version mysql Ver 14.12 Distrib 5.0.22, for unknown-linux-gnu (x86_64) using readline 5.0 Both MySQL servers are standard binaries. We have also slaves running MySQL 4.1 on an Ubuntu 5.10 64bits system without the above issue. $ uname -a Linux eva 2.6.12-10-amd64-k8-smp #1 SMP Fri Apr 28 13:28:18 UTC 2006 x86_64 GNU/Linux $ mysql --version mysql Ver 14.7 Distrib 4.1.15, for unknown-linux-gnu (x86_64) using readline 4.3 The problem seems to be in MySQL 5 branch.
[28 Jun 2006 7:29]
Paul Abel
I've tested this on Solaris 8, which mirrors our production systems. I can reproduce the problem using MySQL built Solaris 8 64bit binaries of 5.0.22 & 5.1.11beta. 4.1.20 works with no issues, so looks like the bug was introduced in version 5. I will try to capture the log output from the same versions but the debug builds.
[31 Aug 2006 19:05]
Trudy Pelzer
Since the current implementation of LOAD DATA FROM MASTER and LOAD TABLE FROM MASTER is very limited, these statements are deprecated in versions 4.1, 5.0 and 5.1 of MySQL. We will introduce a more advanced technique (online backup) in a future (>5.1) version, that will have the added advantage of working with more storage engines. For 5.1 and earlier, the recommended alternative solution to using LOAD DATA|TABLE FROM MASTER is mysqldump on the master piped (or output copied) to the mysql client on the slave. This also has the advantage of working for all storage engines. Another alternative is mysqlhotcopy, which is faster because it does a binary copy, but works only for MyISAM tables. A deprecation warning has been added for these statements in 4.1, 5.0 and 5.1.