Bug #37417 Master/Slave INSERT/SELECT results differ
Submitted: 16 Jun 2008 1:47 Modified: 7 Nov 2008 6:17
Reporter: Adam Dixon Email Updates:
Status: No Feedback Impact on me:
None 
Category:MySQL Server: Replication Severity:S3 (Non-critical)
Version:5.0.64 OS:Any
Assigned to: Assigned Account CPU Architecture:Any
Triage: D2 (Serious) / R3 (Medium) / E4 (High)

[16 Jun 2008 1:47] Adam Dixon
Description:
Found in a stored proc using sample dataset, the results of a SELECT statement on master & slave differ.

See private comments for a repeat with customer data. I have not been able to shrink this down.

Basically data is reaching the slave with different auto increment ids.

How to repeat:
See Private comments and attachment for repeatable case.

Set up master/slave replication and paste v2_rundown_1.txt into command line. (ensure source commands import correctly etc)

Then do (on bother master and slave);
select * from v2guide.guide; 

Notice differences (same data gets different auto increment ids etc), this is because the select produces results in a different order.
[7 Oct 2008 6:17] Susanne Ebrecht
Adam, 

after I got deeper into analysing this problem I can't reproduce this.

I can reproduce only when time zone tables of mysql database aren't filled.

Please follow the instruction of:
http://dev.mysql.com/doc/refman/5.0/en/time-zone-support.html

Use on master:

./bin/mysql_tzinfo_to_sql /usr/share/zoneinfo | ./bin/mysql -h <masterhost> -P<masterport> -u <superuser> mysql

and on slave:
./bin/mysql_tzinfo_to_sql <slave>/usr/share/zoneinfo | ./bin/mysql -h <slavehost> -P<slaveport> -u <superuser> mysql

This will fill the time zone tables of mysql database.

After this time zone tables are filled I can't reproduce it. Neither with MySQL 5.0 bzr tree nor with MySQL 5.1 bzr tree.

Is it ok for you to set this bug to "Can't repeat" or to "not a bug"?
[8 Nov 2008 0:00] Bugs System
No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".