MySQL Lists are EOL. Please join:

List:Commits« Previous MessageNext Message »
From:Andrei Elkin Date:June 9 2009 6:40pm
Subject:Re: BUG#42851: why two patches in the bug report...
View as plain text  
Luís, hello.

> Hi Zhenxing, Andrei,
>   I have committed a fix for the bug, which is split into two patches.
> What follows are just some comments to help driving you during your
> review.

I agree with the idea of your fixes.

We have discussed today on #rep my proposal to convert 
the new 

   binlog_stmt_suppress_unsafe_warnings of bool


  binlog_stmt_suppress_max_unsafe_warnings of int

       <luis> andrei|around: what would the user benefit from it?
[20:18]<andrei|around> luis, as said, it comes along the existing
		       `max_error_count', plus the user would see something
		       instead of to be complete blind.
[20:20]<luis> andrei|around: yes. on the other hand you have log-warnings
	      which is boolean... but the second argument is a good point.
[20:30]<luis> andrei|around: ok, run the poll. I think I buy your idea, let's
	      hear from jasonh. Thanks

Zhen Xing, you are welcome to express your opinion on this user interface

A second point is testing of --log-warnings=0 to suppress the unsafe
messages in the error log. As I got it, messages get through.
This might be a separate issue, could you please check that 
the pre-patch behavior is similar.
If it is, then your patch, I think, can not intend to fix that.



>   There are several problems reported in the bug report:
>     1. Statements with LIMIT will cause warnings in error log,
>        even though they do not affect replication (this was a
>        consequence of fix for BUG#34768).
>     2. Statements with LIMIT + ORDER BY still cause
>        warning (already reported and handled in BUG#42415).
>     3. Statements with LIMIT + ORDER BY + statement filtered (eg,
>        using binlog-ignore-db) still causes warning (patch 1).
>     4. Given 1., 2., 3., mysqld error log grows up, becoming a
>        maintenance mess (patch 2).
>   Items 1. and 2. are not a problem per se. However, there is
>   collateral damage and feedback for item 4.
>   Item 3. contributes wrongly to item 4. I have proposed a
>   fix in patch 1: binlog filter rules are checked before printing 
>   out the warning. Zhenxing has already approved it - thanks.
>   Item 4. is fixed by 2nd patch in the bug report. In BUG#42415 we
>   had decided to keep the marking of unsafe thus the warning print
>   out when ORDER BY PK was used. However, this can turn it a big 
>   maintenance problem with several Gigabytes sized error log files.
>   Zhenxing, I need you to look at this patch also. Andrei, you still
>   have to take a look at both patches.
>   Item 3 fix:
>   Item 4 fix:
>   Patch fixing for Item 4 is incremental and builds on top of patch 
>   fixing Item 4.
> Regards,
> Luís
Re: BUG#42851: why two patches in the bug report...Andrei Elkin9 Jun
  • Re: BUG#42851: why two patches in the bug report...He Zhenxing10 Jun
    • Re: BUG#42851: why two patches in the bug report...Andrei Elkin10 Jun
      • Re: BUG#42851: why two patches in the bug report...He Zhenxing10 Jun
        • Re: BUG#42851: why two patches in the bug report...Luís Soares11 Jun
        • Re: BUG#42851: why two patches in the bug report...Andrei Elkin11 Jun