On 04/29/2015 01:17 PM, Robert Mustacchi wrote:
On 4/29/15 10:05 , Joe Landman wrote:
1) can I detach (easily without too much pain, ala
https://wiki.smartos.org/display/DOC/Using+vmadm+to+manage+virtual+machines#Usingvmadmtomanagevirtualmachines-AddadisktoaVMthenremoveit
) a storage disk from a zone?
That's strictly for KVM, not for OS virtualized instances. So, no this
model doesn't work. There's no direct equivalent of a disk. There is a
delegated dataset, but that's scoped to the instance itself.
You could instead create your own floating delegated dataset and use the
zoneadm and zfs tools yourself to manage it. You'll also likely need
something inside of the guest to forcibly unount it and mount it on the
other side. But you won't be able to manage it with vmadm.
Got it. Thanks. I'll play with this to see if I can do some of this.
2) can I then attach (preferably live) the same storage disk to the
partner zone?
Again, not a disk, but a zfs file system, presumably. You'd again need
logic in the guest or go into the guest to mount it.
Yes, the storage disk would be a zfs file system that was attached to
the zone.
This gets to another question: if an LX branded zone crashes (I've not
played with them much other than lighting a few up to play), can we turn
them off from the GZ, or do we need to reboot the whole machine? I
don't mind so much forcing the unmounting of the data set (zfs file
system), as most of the caching for the branded zone would be on SmartOS
anyway; e.g. we wouldn't have to worry about most of these:
sysctl -a | grep dirty
vm.dirty_background_bytes = 0
vm.dirty_background_ratio = 0
vm.dirty_bytes = 0
vm.dirty_expire_centisecs = 3000
vm.dirty_ratio = 30
vm.dirty_writeback_centisecs = 500
as the dirty data wouldn't really be cached in the zone, but rather by
the GZ ... ?
And also of great interest (the networking question):
3) could we export (for remote attachment) the storage disk over the
network as a block device? This one would be nice, and quite helpful,
but for the near term I am willing to skip it if needed.
No, because it's not a block device -- it's just a ZFS file system. You
could instead share it with NFS.
Ok. NFS wouldn't work for us.
But in general, I'd note, this model of trying to build HA doesn't
really fit in with the general model of SmartOS. You'll likely find
yourself fighting against it, not that it means it's impossible.
Ok, got it. Thanks!
Robert
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/23047814-09eabc15
Modify Your Subscription: https://www.listbox.com/member/?&;
Powered by Listbox: http://www.listbox.com
--
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
e: [email protected]
w: http://scalableinformatics.com
t: @scalableinfo
p: +1 734 786 8423 x121
c: +1 734 612 4615
-------------------------------------------
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
