Re: Possible causes for database corruption and solutions

Bruce Momjian wrote:
> Is changing the OS/X wal_sync_method default something we should
> consider?
>

It's certainly reasonable to consider changing both OS X and Windows so
wal_sync_method defaulted to fsync_writethrough, and provide safer
operation by default on both those platforms. It would also result in a
mass of complaints that PG 9.0 was much slower than 8.4 from people who
were running it in an unsafe way before.

Given that we're already starting to see that bad PR on Linux+ext4:
http://www.phoronix.com/scan.php?page=article&item=ext4_then_now&num=3

I wonder whether it's the right time for the "reliable is the default on
every platform" to just suck up and adopt everywhere, if we're already
going to be fighting this "why is PG so slow on recent Linux versions?"
PR campaign anyway.

--
Greg Smith 2ndQuadrant US Baltimore, MD
PostgreSQL Training, Services and Support
greg(at)2ndQuadrant(dot)com www.2ndQuadrant.us