Some more insights
The problem seems to happen only when you add a ashift=9 special vdev to
a pool with ashift=12.
Without forcing ashift, the Optane partitions as special vdev uses ashift=9
When I force ashift=12 everything is ok:
The solution should be that a special vdev remove should not proceed
with different vdevs.
This is the behaviour on a regular vdev remove
gea
@napp-it
Am 12.10.2019 um 15:00 schrieb Guenther Alka:
I have made some performance tests with the new Allocation Classes on
newest OmniOS bloody
on ESXi 6.7U3 with Optane partitions (pass-through) as special vdev
and Slog.
As I was told that special vdevs are removable, I tried a vdev remove.
During this, OmniOS crashed and after a few secondes after reboot,
OmniOS crashed again.
I had just enough time to delete /etc/zfs/zpool.cache to hinder a
mount on next reboot.
gea
@napp-it.org
--
------------------------------
Guenther Alka, Dipl.-Ing. (FH)
Leiter des Rechenzentrums
Tel 07171 602 627
Fax 07171 69259
[email protected]
http://rz.hfg-gmuend.de
------------------------------------------
illumos: illumos-discuss
Permalink:
https://illumos.topicbox.com/groups/discuss/T6d8674c40f006633-M435adc886816d08b51d1f386
Delivery options: https://illumos.topicbox.com/groups/discuss/subscription
