Re: Question: Multiple pg clusters on one server can be reached with the standard port.

On 6/16/23 10:54, Brainmue wrote:
> 16. Juni 2023 17:41, "Ron" <ronljohnsonjr(at)gmail(dot)com> schrieb:
>
>> On 6/16/23 10:18, Laurenz Albe wrote:
>>
>>> On Fri, 2023-06-16 at 14:49 +0000, Brainmue wrote:
>> 16. Juni 2023 14:50, "Laurenz Albe" <laurenz(dot)albe(at)cybertec(dot)at> schrieb:
>>> On Fri, 2023-06-16 at 12:35 +0000, Brainmue wrote:
>>>
>>> We want to minimise dependencies between the application and the associated PostgreSQL DB.
>>> The idea is that the application gets its DB alias and this is then used as a connection string.
>>> This way we can decide in the backend on which server the PostgreSQL DB is running.
>>> There is an existing solution for that: the libpq connection service file:
>>> https://www.postgresql.org/docs/current/libpq-pgservice.html
>>>
>>> If you want to manage the connection strings centrally, you can use LDAP lookup:
>>> https://www.postgresql.org/docs/current/libpq-ldap.html
>> Thank you, I already know this solution, but the LDAP solution is out of the question for us and
>> the file again means an intervention on the client. And that's exactly what we don't want.
>>> Okay.
>>>
>>> Then why don't you go with your original solution, but use a unique TCP port number
>>> for each database? There are enough port numbers available. That way, there is no
>>> collision and no need for a proxy to map port numbers.
>> In practice, that gets very complicated is large organizations: every time you add another
>> database, you must file another request with the CISO RISK office to get yet another non-standard
>> port open from dozens of machines, and the network team implement them.
>>
>> Operationally much simpler to have a listener handle that.
>>
>> -- Born in Arizona, moved to Babylonia.
> Hello Ron,
>
> I have to agree with you there as well. The workflow you have to go through is also often a time
> issue.
> There are many places that have to agree and then application owners still have to provide
> justifications.
> At the same time, we have to be flexible and fast and allocate the resources well at any time and
> provide the application with the maximum possible performance.

There's always The Cloud...  spinning up a new AWS RDS Postgresql is fast
and simple.  (Costly, though.)

--
Born in Arizona, moved to Babylonia.