[ovirt-users] Remote PostgreSQL 9.5 (was: Answer file key for "nonlocal postgres")

[Yedidyah Bar David]

On Mon, Mar 27, 2017 at 9:07 PM, Jamie Lawrence
<jlawrence at squaretrade.com> wrote:
>> On Mar 25, 2017, at 10:57 PM, Yedidyah Bar David <didi at redhat.com> wrote:
>>
>> On Fri, Mar 24, 2017 at 3:08 AM, Jamie Lawrence
>> <jlawrence at squaretrade.com> wrote:
>
> […]
>
>>> Anyone know what I am missing?
>>
>> Probably OVESETUP_PROVISIONING/postgresProvisioningEnabled
>> and OVESETUP_DWH_PROVISIONING/postgresProvisioningEnabled .
>
> Appreciate the reply - thanks!
>
>> That said, I strongly recommend to not try and write the answer file
>> by hand. Instead, do an interactive setup with the exact conditions […]
>
> I know what I was doing is unsupported. I was wondering down the wrong troubleshooting path for a bit there, but I think ultimately what I need is also unsupported.
>
> It was because I was trying to push this into our extant DB infrastructure, which is PG 9.5. Which I found doesn’t work with a local-install, either. (I was thinking it would work due to past experience with things that demand an old Postgres; IME, PG generally has pretty solid forward-compatibility.)

In "local-install" you mean on the engine machine?

>From RPMs?

Instead of the OS-packaged PG (and not in parallel)?

Were its binaries in /usr/bin (and not some private path)?

If answers for all of the above are 'Yes', then please
share setup logs, perhaps preferably by opening a bugzilla
RFE and attaching them there. It's rather likely that
whatever problems you had are quite easy to solve.

Otherwise, please try that, or see the bug(s) below for
a discussion about this.

>
> So that leads me to my next question: if I install under the supported  version and dump/load/reconfigure to PG9.5.3, is anyone aware of any actual problems (other than lack of official support)?

I personally didn't yet reach that point to be able to tell about,
nor do I know about others that did, but see below.

> In doing answerfile-driven installs repeatedly, the point where it now fails is after the DB load, with ovirt-aaa-jdbc-tool choking and failing the run.
>
> The reason I’m considering that as my fallback, nothing-else-worked option is that the DB needs to live in one of our existing clusters. We are a heavy Postres shop with a lot of hardware, humans and process devoted to maintaining it, and the DBAs would hang my corpse up as a deterrent to others if I started installing ancient instances in random places for them to take care of.

If you do manage to make it work locally, then working with
remote db should be easy, but does require (currently) to have
the local client to be of the same version.

Please see this bug, and the the ones it depends on:

https://bugzilla.redhat.com/show_bug.cgi?id=1324882

Almost all of it is relevant for a "vanilla" 9.5.

I am allowing myself to change the subject of current email,
as I assume the original issue is concluded.

Best,
-- 
Didi