[openstack-dev] [heat] Rolling upgrades vs. duplication of prop data

Zane Bitter zbitter at redhat.com
Thu Jan 5 16:58:26 UTC 2017


On 05/01/17 11:41, Crag Wolfe wrote:
> Hi,
>
> I have a patch[1] to support the de-duplication of resource properties
> data between events and resources. In the ideal rolling-upgrade world,
> we would be writing data to the old and new db locations, only reading
> from the old in the first release (let's assume Ocata). The problem is
> that in this particular case, we would be duplicating a lot of data, so
> [1] for now does not take that approach. I.e., it is not rolling-upgrade
> friendly.
>
> So, we need to decide what to do for Ocata:
>
> A. Support assert:supports-upgrade[2] and resign ourselves to writing
> duplicated resource prop. data through Pike (following the standard
> strategy of write to old/new and read from old, write to old/new and
> read from new, write/read from new over O,P,Q).
>
> B. Push assert:supports-upgrade back until Pike, and avoid writing
> resource prop. data in multiple locations in Ocata.

+1

Rabi mentioned that we don't yet have tests in place to claim the tag in 
Ocata anyway, so I vote for making it easy on ourselves until we have 
to. Anything that involves shifting stuff between tables like this 
inevitably gets pretty gnarly.

> C. DB triggers.

-2! -2!

> I vote for B. I'm pretty sure there is not much support for C (count me
> in that group :), but throwing it out there just in case.
>
> Thanks,
>
> --Crag
>
> [1] https://review.openstack.org/#/c/363415/
>
> [2] https://review.openstack.org/#/c/407989/
>
>
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>




More information about the OpenStack-dev mailing list