On 1/5/23 23:43, Laurenz Albe wrote:
> On Tue, 2022-12-27 at 00:48 -0600, Ron wrote:
>> If it really is a critical production database, you will have a CAT/UAT (customer/user acceptance testing)
>> server on which you rigorously run regression tests on a point release for a month before updating the production
server.
>>
>> Otherwise, it's a hope-and-pray database.
> No, that is wrong.
>
> You should not test your application when you install a minor update. The reason is that
> few people are willing to test the application thoroughly every few months, and the outcome
> is that minor releases are *not* applied regularly, as they should be.
>
> You are supposed to trust PostgreSQL development that they don't introduce new bugs.
> Sure, this can happen, even though all possible care is taken with backpatches. I have
> seen it happen once or twice in the 15+ years I have been dealing with PostgreSQL.
> In that case, a new minor release will come out soon afterwards.
It's absolutely standard practice "in the enterprise" to install the latest
patch on the UAT (and possibly Dev and Staging) servers before rolling out
to production.
Have I aver seen a problem in Postgresql? No. But I've seen problems with
other RDBMSs. If a problem did happen, and caused for example, an important
report to suddenly take 3 hours instead of 3 minutes, the client will
scream; there might even be SLA penalties.
Thus, we're cautious with "critical production databases".
--
Born in Arizona, moved to Babylonia.
