a few doubts about backing up Landscape and MAAS - VM environment

[Carlos Capriotti]

Andreas, hello.

You might want to check this bug report and its very recent activity. It
brings some light to the subject:

https://bugs.launchpad.net/bugs/1520645



On Mon, Feb 22, 2016 at 9:33 PM, Andreas Hasenack <andreas at canonical.com>
wrote:

>
>
> On Thu, Feb 18, 2016 at 3:12 PM, Carlos Capriotti <
> capriotti.carlos at gmail.com> wrote:
>
>> Hello folks.
>>
>> I am currently in an adventure installing MAAS, Landscape and OpenStack.
>>
>> Sparing you excessive details - which I am prone to - let me just quickly
>> describe my environment to you:
>>
>> I am using one VM host (Ubuntu 14.04, KVM+qemu) that runs two VMs: the
>> MAAS server and the Landscape server. Always running Autopilot (or trying
>> to), and my network has two NICs, with public and private networks. All of
>> my nodes (7) are physical.
>>
>> I had ONE successful OpenStack deployment so far, with Autopilot, and ten
>> or twenty "almost there"s, for various reasons.  Trying to save some time
>> "the next time I try", I started cloning the VMs after specific milestones,
>> for instance, an "EMPTY" MAAS, where everything is installed and configured
>> (ssh keys, openstack, etc), but no nodes were enlisted. Other typical
>> milestones were "Landscape added", "All nodes Ready", etc.
>>
>
>
> That shouldn't be needed. Could you elaborate on those failed attempts you
> had? The autopilot collects logs from failed attempts and offers you a
> chance to file a bug with them. If you do that, we can take a look at what
> went wrong.
>
>
>>
>> The next milestone would be after deploying Landscape, and this needs
>> some explaining: instead of enlisting all of my nodes and the future
>> Landscape VM, I only enlisted and commissioned the VM, deployed Landscape,
>> and stopped both VMs (MAAS and the recently installed Landscape), created a
>> clone of both, and then resumed the operation.
>>
>> After my last failure (OpenStack failed installing at 95%...), instead of
>> fetching the logs and opening a new incident, I simply deleted the
>> "tainted" VMs, and started from my last known-good point, restoring the VMs
>> I cloned. I was even careful to use the same MAC addresses, specially on
>> Landscape. Much to my surprise, MAAS did not like the strategy, refused to
>> recognize Landscape, and I saw myself with gigabytes of useless clones,
>> which were promptly deleted and I had to start all from scratch. Literally,
>> from the very installation of MAAS.
>>
>> So, here are the questions:
>>
>> 1) Regardin VMs and the relationship MAAS-Landscape, what would be the
>> best way to make a backup and restore it, making sure both will talk to
>> each other again ?
>>
>
> Deploying Landscape takes just a few minutes, and you shouldn't have to
> tear it down each time. I'd say this snapshotting is not worth it.
>
>
>>
>> 2) Alternatively, can I deploy a new MAAS and "Attach" a pre-existing,
>> previous Landscape server to it ? I see the "Attach" option on the GUI, and
>> even tried it once, but that has already faded from my memory.
>>
>
> Just leave MAAS and Landscape installed, there should really be no need to
> start over really from scratch each time.
>
>
>>
>> In case you are wondering why I would install a new MAAS server and keep
>> the Landscape machine: I have been facing an issue where MAAS is get
>> confused with MAC addresses, during the enlist phase, claiming that a
>> specific MAC already exists, making it impossible to add/commission a node.
>> There are several moving parts on this problem and it is - I believe -
>> object of another email or incident, so, I'll let it be for now.
>>
>
> I haven't seen that before. Specially if you are using real nodes enlisted
> in MAAS.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-openstack-installer/attachments/20160413/b8ca5b10/attachment.html>