Jerry Jelinek wrote:
I wanted to close the loop on the mount interaction with ZFS delegated
datasets under lx.
I closed https://smartos.org/bugview/OS-4157 as 'not a bug' with the
following evaluation:
What is happening here is that the way libzfs works, setting the
mountpoint doesn't cause the mount to happen if the dataset wasn't
mounted to begin with. If the dataset is mounted and I set the
mountpoint to someplace else, the dataset will get mounted at the new
place
So the problem really is that the dataset was not mounted at it's
mountpoint when the zone booted. In a native zone this mount would
happen when the fs-local SMF svc runs the command '/usr/sbin/zfs mount
-va' to mount all of the ZFS file systems. If you run
'/native/usr/sbin/zfs mount -va' under lx it does mount all of the
datasets at their mountpoints. So, depending on which Linux service
manager is being used inside the zone, a new service (for example, a
new rc file) could be created to do the ZFS mounts at boot time.
Hello Jerry,
Your analysis makes sense, but it would be very desirable for the
required service/rc file could be added to the zone image.
The condition this worries me with the current situation is what happens
if there is an unscheduled (or even a scheduled) restart of the zone,
possibly caused by a reboot in the GZ? If the host has a number of LX
brand zones, having to manually go though each one and run a zfs mount
-a isn't an elegant solution...
I have a couple of Centos zones under evaluation by Linux admins who
know very little about ZFS. They would expect filesystems to be
automatically mounted on boot as they are in "normal" Linux system.
--
Ian.
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription:
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com
