Saturday, January 12, 2013

Fun with Bugs, Issue #1, January 2013

Maybe some of you noted that I write about MySQL bugs on Facebook almost every day. These posts are mostly random and for fun (and addressed usually to my former colleagues in Oracle), but some of bugs I had written about are really serious, coming from Percona and Oracle customers and may affect many MySQL users. Other bugs just show problems in current Oracle's treatment of public bugs database. So I decided to write a series of blog posts that will summarize my recent bugs related comments and highlight problems I consider serious. Being inspired by Sheldon Cooper I called this series "Fun with Bugs". This is issue #1 that will cover period from January 1 till January 10, 2013.

Let me start with bugs that I consider really serious. First of all, all MySQL Cluster users should know about a regression bug introduced in 7.2.9 (that was later removed from downloads without any comments that I've noted) and other recent versions, Bug #67928. I have good news for you. The bug is fixed and version 7.2.10 with the fix is released on January. For some time this bug was not mentioned in the Release Notes, but now we have a very detailed description. Bad news for those Cluster users that may use InnoDB temporary tables while working with cluster binaries - even in 7.2.10 Bug #64663 still happens.

Speaking about InnoDB, you all know about great new features and performance improvements in MySQL 5.6. But does it mean that there are no potentially serious InnoDB bugs in GA versions? Surely, no! Take a look at recently reported ones: Bug #68069 from Domas (no public comments from any Oracle engineer in the bug report at the moment!) or Bug #68022 from Mark Callaghan, Bug #68021 from my colleague Ovais (verified today after my numerous posts about it) or ages old and probably forgotten Bug #64432 from my colleague Laurynas. The worst (as it is a regression bug in 5.5.29) is probably Bug #68051 reported by Davi Arnaut.

Those who are interested in InnoDB internals (and efficiency of disk usage) should check these reports from Jeremy Cole: Bug #67963, Bug #68023 (both are still open at the moment). Good reading, even though hardly any of them will be fixed even in MySQL 5.7.

Users of all kinds of Red Hat Linux should be aware of Bug #67965 when installing recent RPMs. Oracle claims it's all OK and documented to conflict with native RPM, while others may say this is a regression that should be fixed once and forever.

I hope many of you noted that my former colleague Elena Stepanova (now she works for MariaDB) regularly send great bug reports with detailed, ready to use test cases. Recently she had reported some bugs in MySQL 5.6.9. Check Bug #68041 as a good example and be warned that MySQL 5.6 may give not only performance improvements, but also regressions, so you have to test how it works for your applications carefully.

I wanted to write something about more funny reports (that still make a lot of sense) like Bug #68035 (that makes people wonder if they really have recent Core i7 CPU or it is a joke) or Bug #68034 (should MySQL really warn about empty password at the command line or why this matters at all), but probably this blog post is already to long for anybody to still hope to find some fun in it.

So, let me just seriously ask all Oracle customers with valid MySQL support contract who use 5.5.29 and replication in production: please, ask Oracle (in a service request, probably) what Bug #68045 (don't click, it's private) is about and can you be affected. I've heard rumors somewhere that it is a regression bug, but only Oracle customers can get a clear confirmation.

I hope it was fun enough for the beginning. To be continued...

5 comments:

  1. Read http://www.mysqlperformanceblog.com/2013/01/13/cve-2012-4414-in-mysql-5-5-29-and-percona-server-5-5-29/ about the problem that could be behind bug #68045. Still, it's probably better to ask Oracle for comments.

    ReplyDelete
  2. I wrote all the information in the blog post that I posted in the bug report. There's no other information visible to me on that bug report (just that it's been assigned to somebody)

    ReplyDelete
  3. I'm still sad about bug 28432 that was marked as closed but never fixed.

    ReplyDelete
  4. Eric, I've tried to "escalate" this bug on Facebook. Dmitry Lenev claimed that it should be fixed in 5.6 (he had not agreed with the idea that MDL may solve it). Maybe we should just test on 5.6.9?

    ReplyDelete
  5. Re Bug #68069, Marko from InnoDB team posted a patch for internal review on Jan 12, and added a comment to Bug #68069 on Sunday, Jan 13.

    ReplyDelete