On Wed, 23 May 2012, Stewart Smith <stewart@flamingspork.com> wrote:
Typically, the log tail is trimmed as things are written back as a checkpoint is performed (writing back things from log to data file). If you have copied 1/2 the data file and checkpoint gets done that is in the first half of the data file and the log wraps... you're pretty hosed no matter what.
Thanks for the explanation.
I've never seen any benchmark that says ZFS should even be considered for databases (nor btrfs for that matter).
Basically, it's still; step 1 for database performance: use XFS.
Does XFS have features equivalent to L2ARC and ZIL? How can a filesystem without such features compare? Also for a database like MySQL that doesn't have anything in the way of data checksums then surely the only thing to do is to use a filesystem like BTRFS or ZFS. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/