Re: patterns for database administration - Mailing list pgsql-general

From Matthew Hixson
Subject Re: patterns for database administration
Date
Msg-id 8DC9E2D3-7D0B-11D8-B6BA-000A95D05926@poindextrose.org
Whole thread Raw
In response to Re: patterns for database administration  (Jonathan Bartlett <johnnyb@eskimo.com>)
List pgsql-general
On Mar 23, 2004, at 11:54 AM, Jonathan Bartlett wrote:

>> One of the reasons this idea was suggested was because my client is
>> concerned that its "crazy" to be modifying business data in a system
>> that is running and processing purchase transactions.  And I'm
>> wondering whether or not this is even a concern when most people build
>> this type of application.  I think its going to be painful to keep
>> track of changes between the two databases (or schemas if you prefer).
>> It sounds like this would be highly prone to errors and cause more
>> problems than it solves.
>>    Thoughts?
>
> It sounds like the problem they have is that they want you to be able
> to
> make changes, but perhaps not make them active until they are all
> finished.  Is that what the problem is?
>
> This can be solved in a number of ways.  You can mark records as
> "testing", and then have an approval step which copies the testing
> records
> over the production records.  You can also have an "active date" on
> your
> records, and then mark your records as being active in the future.

Indeed we're already doing that.

> I think we need more information on the "whys" of this before making
> clearer suggestions.

I agree completely.  Unfortunately I don't have anything more concrete
to go on than my previous post above.    I think the "whys" are
extremely weak.  I'm going to suggest we leave things as they are and
allow the administration application to update the production database.
   Thanks,
    -M@


pgsql-general by date:

Previous
From: Joseph Shraibman
Date:
Subject: Re: partial VACUUM FULL
Next
From: Rob Hoopman
Date:
Subject: self referencing tables/ nested sets etc...