[OmniOS-discuss] Omios, hvm and AWS

[Peter Tribble]

On Tue, Aug 1, 2017 at 12:18 PM, Tomasz Kłoczko <kloczko.tomasz at gmail.com>
wrote:

> I'll share my method of importing images :)
>
> AWS API provides uploading VHD image (for some reasons using other formats
> is not working).
>

Works for VMDK as well. With VirtualBox, you get the somewhat
misleading "unsupported compression algorithm 0" error, which actually
means that the VMDK is in OVA format whereas the AWS API only
supports the OVF format.


> All what you need is create such image. On top of the Solaris is it quite
> easy. All what you need is VBox.
> Such method is quite handy because using VboxManage command is possible to
> create, setup blank VM and boot it in batch mode.
>
> Below is fragment of my script which could be used as template. It is for
> install Linux inside but it can be adapted to install any OS (Solaris as
> well)
>

Have you tried this with Solaris or illumos?

It doesn't work, because the root ZFS pool embeds the physical paths
to the disk inside the pool, and the emulated disk devices that VBox
provides don't match what comes in an EC2 instance. So yes, you can
create an image, but at some point you need to fix the pool to have the
correct metadata.


> --
> # http://www.perkin.org.uk/posts/create-virtualbox-vm-
> from-the-command-line.html
> VBoxManage setproperty hwvirtexclusive off
>
> VM=$1
> MAC_ADDR="$2"
>
> echo "Unregister $VM"
> VBoxManage unregistervm $VM \
> --delete
> rm -rf $VM.vhd
>
> echo "Create $VM.vhd image"
> VBoxManage createmedium disk \
> --size 5120 \
> --format VHD \
> --filename $VM.vhd
>
> echo "Register $VM.vmdk image as Linux_64 VM"
> VBoxManage createvm \
> --name $VM \
> --ostype "Linux_64" \
> --register
>
> echo "Add SATA controller to $VM"
> VBoxManage storagectl $VM \
> --name "SATA Controller" \
> --add sata \
> --controller IntelAHCI
>
> VBoxManage storageattach $VM \
> --storagectl "SATA Controller" \
> --port 0 \
> --device 0 \
> --type hdd \
> --medium $VM.vhd
>
> echo "Add APIC"
> VBoxManage modifyvm $VM --ioapic on
>
> echo "Add boot sequence order"
> VBoxManage modifyvm $VM --boot1 net --boot2 disk
>
> echo "Add RAM, VRAM and enable RDP access"
> VBoxManage modifyvm $VM \
> --memory 4096 \
> --vram 128 \
> --vrde on
>
> echo "Create $VM0 VNIC with exact MAC: $MAC_ADDR"
> dladm create-vnic -l aggr0 -m $MAC_ADDR $VM0
>
> echo "Network card"
> VBoxManage modifyvm $VM \
> --nic1 bridged \
> --bridgeadapter1 $VM0 \
> --macaddress1 080027f3399d
>
> VBoxHeadless -startvm $VM
>
> echo "Delete VM0 VNIC"
> dladm delete-vnic $VM0
>
> echo "Unregister $VM"
> VBoxManage unregistervm $VM
> mv $VM.vhd ready
>
> ec2-import-instance $VM.vhd –f VHD -t m3.xlarge -a x86_64 -b myawsbucket
> -o AKIAIOSFODNN7EXAMPLE –w wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY -p
> Linux
> --
>
> If kickstart file or AI manifest at the end of installation will be
> powering off VM "VBoxHeadless -startvm $VM" command will be able to finish
> and next steps can be done in batch mode like uploading just generated
> image to VMware, OVM, Azure or AWS.
> BTW ec2-import-instance command: this command is a bit odd because you
> need to pass two keys. First one is account key and second one is AWS
> bucket key (even if AWS account has RW access to bucket).
> http://docs.aws.amazon.com/vm-import/latest/userguide/
> vmimport-image-import.html#upload-image
>
> Real advantage of using this method is sharing the same Linux KS or
> Solaris AI manifest/profile (or even Windows install profiles) with bare
> metal installation descriptions in form of native install descriptions.
> Knowing exact MAC address of the VM used to create exact image is possible
> on install server side associate those MAC addresses with KS or AI files.
>
> So .. you don't need to sit down upside down on you chair or use ISO
> images :)
> All what is needed is possible to do in batch mode and/or using
> regular/native OS install services.
>
> kloczek
>
> Tomasz Kłoczko | Tel: 0774 1209067 | LinkedIn: *http://lnkd.in/FXPWxH
> <http://lnkd.in/FXPWxH>*
>
> On 31 July 2017 at 21:51, Al Slater <al.slater at scluk.com> wrote:
>
>> On 31/07/17 21:30, Eric Sproul wrote:
>> > On Mon, Jul 31, 2017 at 4:05 PM, Al Slater <al.slater at scluk.com> wrote:
>> >> One more question though, is there any way to enable an SMF service for
>> >> the next reboot, but not immediately.  Specifically, I want to enable
>> >> the initial-boot service with a .initialboot file in place, then create
>> >> a new AMI.
>> >>
>> >> I wist to use .initialboot to grab the instance configuration from
>> >> amazon (hostname, root keys etc) and configure appropriately when the
>> >> new instance starts.
>> >
>> > Hi Al,
>> > The initial-boot service isn't really suitable for this sort of thing.
>> > You might want to check out
>> > pkg://omnios/system/management/ec2-credential which specifically
>> > handles setting up the credentials at first boot.  That could be
>> > trivially extended[1] to set the system hostname and probably any
>> > other "standard" thing that operators want.
>> >
>> > Eric
>> >
>> > [1] https://github.com/omniosorg/omnios-build/blob/master/build/
>> ec2-credential/files/install-ec2-credential
>>
>> Thanks for the pointer Eric.
>>
>> --
>> Al Slater
>>
>> _______________________________________________
>> OmniOS-discuss mailing list
>> OmniOS-discuss at lists.omniti.com
>> http://lists.omniti.com/mailman/listinfo/omnios-discuss
>>
>
>
> _______________________________________________
> OmniOS-discuss mailing list
> OmniOS-discuss at lists.omniti.com
> http://lists.omniti.com/mailman/listinfo/omnios-discuss
>
>


-- 
-Peter Tribble
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://omniosce.org/ml-archive/attachments/20170801/cc25e384/attachment-0001.html>