Re: Installing Oracle on Machine Images

As I can still say this for one more day- I’m a Microsoft employee in the disclaimer…:)

Our default recommendation is using the OS image from Azure, installing the Oracle home and directory setup, etc. to meet your environment requirements and then create an image in your own image gallery, using this for deployments. You can then perform the final installation during the deployment as part of your automation scripts, (BASH, Terraform/Ansible, Bicep, etc.)

The reason may be different than you may assume though- images can be replaced by the vendor, (Oracle Linux images are provided by Oracle, Redhat by Redhat…) and if you don’t have your own image copy in your gallery, and the image for a recovery has been removed or replaced, you won’t be able to restore from a volume snapshot to a new build.

No matter your final decision, OS images used should have image copies retained in your Azure tenant for recovery. Don’t count on them being retained forever by the cloud vendor.

Hope this helps,
Kellyn

On Wed, May 31, 2023 at 3:14 AM Stefan Koehler <contact_at_soocs.de> wrote:

> Hello Charlotte,
> technically this is possible (and supported) but I would not recommend
> this for the following reasons:
>
> - For every applied RU (or even one-off patch) you need to create a new
> "golden" machine image
> - You have two different procedures to deploy Oracle releases/patches (one
> for the existing database and one with the "golden" machine image)
> - You have to modify the image after cloning and first startup anyway (SSH
> keys, DB and DBID with NID - the latter is especially important as you can
> not have two databases with the same DBID registered in a RMAN catalog,
> etc.)
>
> I understand that you save a few minutes when deploying and spinning up a
> new "golden" machine image for the first time, but IMHO it is not worth to
> have different deployment procedures just to save 10 minutes for the first
> start.
>
> Best Regards
> Stefan Koehler
>
> Independent Oracle performance consultant and researcher
> Website: http://www.soocs.de
> Twitter: _at_OracleSK
>
> > Charlotte Hammond <dmarc-noreply_at_freelists.org> hat am 31.05.2023 10:31
> CEST geschrieben:
> >
> >
> > Hello.
> >
> > We have a requirement to set up some Azure VM Images or AWS AMIs for
> running Oracle DB. My initial plan was to create these images with the
> Oracle software uploaded and then to run (by automation) the actual install
> after the actual VMs were created, followed by DB creation.
> >
> > A colleague has suggested we run the install and DB creation and then
> include those in the images. We tried this and it was absolutely fine in
> casual testing.
> >
> > However, I'm a bit nervous about this as it feels a bit like copying an
> Oracle Home without running clone.pl or OUI to complete the cloning
> process. Of course, booting with an image is not the same as copying a
> directory, so perhaps my concern is unfounded.
> >
> > I just wanted to check the general consensus on this - should a virtual
> machine image contain an installed and start-able database? Whilst it seems
> to work, is this a recommended, or approved approach. Will we hit any
> support issues, or vendor finger pointing, as a consequence?
> >
> > Thank you!
> >
> > Charlotte
> --
> http://www.freelists.org/webpage/oracle-l
>
>
> --

*Kellyn Gorman*
DBAKevlar Blog <http://dbakevlar.com>
about.me/dbakevlar

--
http://www.freelists.org/webpage/oracle-l