[OpenIndiana-discuss] vboxsvc and zvolrights

[Geoff Nordli]

On 13-10-17 09:42 PM, Jim Klimov wrote:
> On 2013-10-17 23:35, Geoff Nordli wrote:
>> I use the built in vbox initiator.
>>
>> Not the highest performing setup, but these are small business services
>> and disk IO isn't the biggest issue we have.
>>
>> I am using it for automated backup system.  I virtualize their servers
>> into vbox, pause the VM(s), zfs snapshot, send to a different pool and
>> send another copy offsite.
>
> In such case note that the recent versions (circa one year now) of the
> script support creation of automatic ZFS snapshots of the backend
> storage after VM stop and before VM start, and at least for file-based
> VMs (including over NFS) and for local zvols it should be able to
> detect the list of involved ZFS datasets, as well as you can provide
> a list of other datasets you know as related (i.e. iSCSI LUN zvols).
>
> I did not have a scenario exactly like yours, but it may be possible
> (either out of the box or after minor scripting) to use the shutdown
> action to pause and un-pause the VMs and have this action wrap the
> creation of the snapshots. Certainly (right now) you can use the
> "vbox.sh zfssnap" method between your own pause/unpause calls with
> just one generic command to call from your automation scripts.
>
> I am not sure if pause/unpause provide for very good backups, since
> they would seem like a reset upon recovery (unlike savestate/restore,
> but at the cost of some more downtime).
>
> Feel free to ask for more features, I might get around to implementing
> them one day... or send patches :)

The vboxmanage pause works great with vbox, because it just pauses all 
activity and it doesn't drop connections to anything that is going on.   
It also flushes the disk buffers and it only takes a few seconds for 
this to happen.

With the script I have, it does backups in batches, so it will pause all 
of the VMs in the batch, then does the snapshot, then resumes all of the 
VMs, which guarantees that everything is at the exactly same point.

Couple feature requests.  If we had volumes linked to a controller/port, 
then if it could automatically create the iscsi target and check to make 
sure that it is attached to the VM, that would be awesome.

Related to that, if you had the option to control flushing.  Any 
production system should not have flushing enabled.

http://www.virtualbox.org/manual/ch12.html#idp56076560

To enable flushing for SATA disks, issue the following command:

VBoxManage setextradata "VM name" "VBoxInternal/Devices/ahci/0/LUN#[x]/Config/IgnoreFlush" 0


I don't know if something like this would go in the vboxsvc, but there 
are three different things that happen with my backups:  1) snapshots 2) 
replication 3) purging/retention.

1)  when I snapshot, it reads a config file, which allows you to have 
multiple "targets" to send the snapshot to. I create a touch file in a 
directory structure that the replication script reads, after successful 
replication it deletes the touch file.
2)  replication looks at the targets that were created when the snapshot 
happened.  I normally run the replication at night.
3)  retention cleans up snapshots (when you run a snapshot you set 
retention policies).

my scripts are a spaghetti mess though :)

have a great day!

Geoff