Dan Harris wrote:
> From my recent experiences, I can say ext3 is probably not a great
> choice for Pg databases. If you check the archives you'll see
> there's a lot of discussion about various journalling filesystems and
> ext3 usually(always?) comes up on the bottom as far as performance
> goes. If you insist on using it, I would at least recommend the
> noatime option in fstab and using data=writeback to get the faster
Based on my knoledge, Ext3 is good with keeping filesystem integrity AND
data integrity while
pressing the reset button.
However, by selecting data=writeback, you gain more speed, but you risk the
data integrity during a crash: Ext3 garantees only filesystem integrity.
This means with database transaction logs: The last transactions are not
guaranteed to be written into the hard drives during a hardware reset,
meaning of a loss of some committed transactions.
Reiserfs is known to do things this false way also.
Is there a way with a Reiserfs filesystem to fulfill both
filesystem AND data integrity requirements nowadays?
See for example "man mount" to see the effects of data=journal,
data=ordered(default) and
data=writeback for Ext3. Only the writeback risks data integrity.
Ext3 is the only journaled filesystem, that I know that fulfills
these fundamental data integrity guarantees. Personally I like about
such filesystems, even though it means less speed.
Marko Ristola
