Open Stack

Hi Tom,

Unfortunately, Glance does not support rolling db upgrades. Actually,
I'm not aware of any OpenStack project that supports such upgrades. So
far, I believe that deployments that require very high uptime are
advised to carefully manage and even at times rewrite the database
migrations.

I'm not happy with this situation, and supporting rolling upgrades in
Glance is on my personal roadmap. However, there are number of
difficult issues we need to tackle to make that work. First we'll need
to refactor the code away from the use of declarative sqlalchemy
models, towards an approach that allows the domain models and database
schemas to be versioned independently. Second, we'll need to come up
with a testing procedure that can enforce the practices that make
rolling upgrades possible. Last (that I know of) we'll need to figure
out a way to communicate the relative significance of each commit to
deployers, as each can have potentially different strategies for the
best way to distribute the change out into their cluster. If we do
eventually provide support for rolling upgrades, its going to be a
daunting task, made much harder by the variety of possible OpenStack
deployments that need to be considered.

This is a subject I'm very excited about. If there are others with
ideas for how we might make progress towards rolling upgrades, I'm
very happy to brainstorm about ways to incorporate these ideas into
Glance. I anticipate adding support will be easier in Glance, due to
the relative simplicity of the project and its data model.

Thanks,
markwash

On Mon, Mar 11, 2013 at 4:55 AM, Hancock, Tom (HP Cloud Services)
<Tom.Hancock at hp.com> wrote:
> I’m happy to be corrected on this as I’m basing it on reading a diff, not an
> actual test.
>
> It looks like glance migration 19 could introduce a non backwards compatible
> change
>
> to the glance database. In principle, how should such an upgrade be managed?
>
>
>
> A ‘rolling’ upgrade would be an ideal way to minimize disruptive updates.
>
> It looks like there would need to be a bounce upgrade (stop the service on
> all nodes;
>
> upgrade sw; start service) as old and new db clients don’t look like they
> will co-exist.
>
>
>
> thanks,
>
> Tom
>
>
>
> ---
>
> Tomas Hancock, HP Cloud Services, Hewlett Packard, Galway. Ireland
> +353-91-754765
>
> Postal Address   : Hewlett Packard Galway Limited, European Software Centre,
> Ballybrit Business Park, Galway, Ireland
> Registered Office: Hewlett Packard Galway Limited, 63-74 Sir John Rogerson's
> Quay, Dublin 2 Registered Number: 361933
>
> The contents of this message and any attachments to it are confidential and
> may be legally privileged. If you have received this message in error you
> should delete it from your system immediately and advise the sender. To any
> recipient of this message within HP, unless otherwise stated, you should
> consider this message and attachments as "HP CONFIDENTIAL".
>
>
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>