On Fri, November 16, 2007 09:31, Alexander Leidinger wrote:
> Quoting Peter Schuller (from Fri, 16
> Nov 2007 01:36:47 +0100):
>>> Can somebody who has done MySQL benchmarks (e.g. Kris Kennaway,
>>> but I don't want to point fingers ) try them when MySQL is
>>> hosted on ZFS? I'm currently having a problem where mysqld gets
>>> stuck in "zfs" state and it becomes completely unusable and
>>> unkillable. The machine generally works, but cannot be rebooted
>>> from the OS in this case (it must be manually power-cycled).

>> I had an unkillable MySQLd a while ago that required reboot
>> (pre-RELENG_7 CURRENT). It was in uninterruptable sleep, but I
>> can't swear to whether it was in zfs state. This was on a quadcore
>> core 2 duo.
>> I can do some sysbenching on a dual core machine with and without
>> ZFS, but it sounds like that won't be more useful than what you
>> already have available.

> If you really do benchmarks and not failure verification, you should
> take care about the recordsize attribute in the corresponding ZFS,
> else you may get suboptimal performance (this is a suggestion of SUN
> for the use of ZFS with DBs, personally I don't have access to a
> machine where the DB is used enough that we could notice a
> performance difference).

Might be a good place to start:


freebsd-fs@freebsd.org mailing list
To unsubscribe, send any mail to "freebsd-fs-unsubscribe@freebsd.org"