Jeffrey W. Baker wrote:
Until a few days ago I had a very large postgresql installation on
XFS. We suffered very unusual problems like corrupt tables, missing rows,
corrupt indexes, and so forth. We've never seen those problems on ext2,
to which we have since switched back.
Why would you put a database on a meta data journaling only filesystem?
The original poster mentions ext2, which has no journaling.
In the case of ext3, with metadata only journaling, or other metadata
only journaling filesystems (XFS for sure, JFS?), you would use postgres
because it does all the data journaling (since v7 with WAL).
What's wrong with ext3 full journal? (data=journal)
Hmm. wonder if that would help mysql :)