Re: 7.2b3 pg_dump, general 7.2b3 comments

Tom Lane wrote:

> Joe Conway <joseph(dot)conway(at)home(dot)com> writes:
>
>>Is pg_dump supposed to work against prior version databases? I was
>>moving data from an old server (running 7.2devel from several months
>>
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
>>ago) to a new server (running 7.2b3), using pg_dump on the *new* server,
>>and got the following message:
>>
>
>># pg_dump -h 172.16.1.84 -U postgres -t el_names -i lt_lcat > el_names.dmp
>>Password: <types in password>
>>pg_dump: query to obtain list of tables failed: ERROR: Attribute
>>'relhasoids' not found
>>
>
> Oh, never mind: I see it. pg_dump's test is on whether the server calls
> itself 7.2 or not. You've evidently got a copy from back before the
> changes to make OIDs optional. The versioning code is not set up to
> deal with intermediate development versions, so it gets it wrong about
> what query to use. You'll need to use the pg_dump of the same vintage
> as the 7.2devel server.
>
> You're a brave man to be putting production data on CVS-tip servers.
> I wouldn't recommend it ;-)
>
> regards, tom lane
>

Thanks, but I'm not that brave. This has been my proof-of-concept server
for the past 6 or so months. The new server is intended to become a
production, relatively high volume data collection server on our factory
floor. I've finally been successful at selling our upper management on
using PostgreSQL in lieu of commercial databases for this purpose (we
already have multiple instances of brand O and brand M). In fact, given
the current economy, they're enthusiastic about it ;-)

In any case, I've already moved the data, just wanted to report the
possible issue. It sounds like it shouldn't affect any but the highly
adventurous!

BTW, after your first reply, I started to load the new pg_dump into gdb
and discovered it had no debug symbols (recall I installed from RPM). Is
there a way to install the RPM with additional configure options without
rebuilding it? Is there any significant downside (performance or
otherwise) to having --enable-debug on a production server?

-- Joe