This week with a client I saw ext4 used for the first time on a production MySQL system which was running Ubuntu 9.10 (Karmic Koala). I observe today while installing 9.10 Server locally that ext4 is the default option. The ext4 filesystem is described as better performance, reliability and features while there is also information about improvements in journaling.
At OSCON 2009 I attended a presentation on Linux Filesystem Performance for Databases by Selena Deckelmann in which ext4 was included. While providing some improvements in sequential reading and writing, there were issue with random I/O which is the key for RDBMS products.
Is the RAID configuration (e.g. RAID 5, RAID 10), strip size, buffer caches, LVM etc more important then upgrading from ext3 to ext4? I don’t have access to any test equipment in order to determine myself however I’d like to know of any experiences from members of the MySQL community and if anybody has experienced any general problems running ext4.
ext4 References
- Ext 4 How To on kernel.org
- Ext4 on kernelnewbies.org
- ext4ext4 overview via wikipedia.org
- First benchmarks of the ext4 file system
I saw some odd stuff with ext4 and MySQL before:
http://harrison-fisk.blogspot.com/2009/12/ext4-with-mysql-binary-logs-oddity.html
Supposedly recently ext4 has fixed up some of it’s behavior, but I haven’t tried it since then.
Wow, they didn’t even test XFS in that comparison? I thought it was common knowledge that was the best filesystem for high performance MySQL at least, would be at least worth testing in Postgres.
I prefer XFS especially for binlogs.