List:Bugs« Previous MessageNext Message »
From:sasha Date:May 18 2000 11:22pm
Subject:Re: Repeatable mysql 'update' bug - lockup / cpu drain.
View as plain text  
Sander Pilon wrote:
> 
> >
> > I can verify that this causes an error in 3.22.29, but it works
> > just fine with
> > 3.23.16.
> >
> > So does this warrant a new release of 3.22, or is the solution
> > to upgrade to 3.23 even though it isnt a fully stable system?
> >
> > I havent had any problems with 3.23.16, and as has been said on this list
> > before, it is equally or more stable than 3.22 for anything that
> > 3.22 did, but
> > maybe not production level on all new features (such as BDB transactions,
> > binary replication etc).
> >
> > SELECT 3-22.feature,3-22.stability,3-23.stability
> >  FROM 3-22
> >   LEFT JOIN 3-23
> >   ON 3-22.feature = 3-23.feature;
> >
> > Dana
> >
> 
> IMO, a 3.22 patch should be released. Why? 3.22 is marketed as 'stable'
> release and
> contains reproducable lockups. These things do not belong in 'stable'
> releases :)

First of all this is not an all-out lockup, as you can kill the thread with
mysqladmin. As far as the title of stable is concerned, read my comments below.

> 
> So... you are saying that 3.32.16 is as stable as 3.22.32, as long as I dont
> use any of the new
> tricks?

More stable - there are some inherent 3.22 problems like this one that have been
fixed.

A short explanation of the history to help everyone understand why a more stable
3.23 version is called alpha, and the 3.22 version with a known bug is still
called stable.

First our versioning convention:

alpha - new version that has not met the requirements of beta 
beta - status is obtained if the previous version has been around for a while,
and there has not been any serious bug reports for a month with the exception of
new features that have been documented as alpha

gamma - all the new features are settled in, and there has not been any serious
bug reports for a month

stable - there has been no reproducable bug reports for a very long time

When 3.22 became beta, the installation base of MySQL was significanlty less
that it is now, which, based on our download stats, has crossed 1,000,000 and
keeps growing by leaps and bounds daily. So 3.22 at the time was able to quicly
make it to the stable status because there were not that many users, which
reduced the probability of a somebody hitting a grey area of the code and
finding a bug. As the user base grew, some bugs got found and fixed in the
version that had the status of stable, and for others we have suggested a
workaround rather than make major code modifications. BTW, the workaround for
your bug in 3.22 is UPDATE foo SET id = id + 1 WHERE id + 0 = val -- this will
prevent the query from using the index and will avoid the infinite loop.

With 3.23, the situation is a bit different. With the current user base, and the
increased use of MySQL in high-load environments, where rare and hard-to-find
bugs can be found, we get a lot more bug reports that we used to. This is not
because there are more bugs, only because there are more users and more people
who try to push MySQL to the limit. So 3.23 has been having a hard time getting
out of alpha for a while due to our requirement of no critical bug reports for a
month ( note that if this requirement was enforced in the big bucks software
circles, many products would still be in alpha :-) ). Nevertheless, we really
hope that 3.23.16 will give us no unpleasant surprises and we will be able to
release 3.23.17 as beta.



-- 
Sasha Pachev

+------------------------------------------------------------------+
|     /*/\*\/\*\   /*/ \*\ /*/ \*\ |*|     Sasha Pachev            |
|    /*/ /*/ /*/   \*\_   |*|   |*||*|     mailto:sasha@stripped  |
|   /*/ /*/ /*/\*\/*/  \*\|*|   |*||*|     Provo, Utah, USA        |
|  /*/     /*/  /*/\*\_/*/ \*\_/*/ |*|____                         |
|  ^^^^^^^^^^^^/*/^^^^^^^^^^^\*\^^^^^^^^^^^                        |
|             /*/             \*\ Developers Team                  |
+------------------------------------------------------------------+
Thread
Repeatable mysql 'update' bug - lockup / cpu drain.Sander Pilon18 May
  • Re: Repeatable mysql 'update' bug - lockup / cpu drain.Dana Powers18 May
    • RE: Repeatable mysql 'update' bug - lockup / cpu drain.Sander Pilon18 May
      • Re: Repeatable mysql 'update' bug - lockup / cpu drain.Thimble Smith19 May
      • Re: Repeatable mysql 'update' bug - lockup / cpu drain.sasha19 May
      • RE: Repeatable mysql 'update' bug - lockup / cpu drain.Michael Widenius22 May
  • Re: Repeatable mysql 'update' bug - lockup / cpu drain.sasha19 May
  • Re: Repeatable mysql 'update' bug - lockup / cpu drain.sasha19 May
    • RE: Repeatable mysql 'update' bug - lockup / cpu drain.Sander Pilon19 May
RE: Repeatable mysql 'update' bug - lockup / cpu drain.Sander Pilon18 May