> Maybe I'm some crazy, radical DBA, but I've never had a version of
> pgsql get EOLed out from underneath me. I migrated from 7.4 to 8.1
> right around the time 8.2 came out then upgraded to 8.2 around 6
> months later.
>
> Where I work now we are looking at migrating from 8.1 to 8.2 or 8.3
> (depending on whether or not we have the man power to fix a few issues
> with type coercion, our app, and 8.3) These aren't "the DBA got a
> wild hair and just had to upgrade" upgrades. Each time I've migrated
> it's been because there were performance or maintenance issues that
> were solved by upgrading.
Perhaps I'm in a unique situation as well, but as the DBA of a
data-tank style DB, I haven't had a problem at all finding
opportunities to upgrade to later versions of postgresql. My schema
isn't all that complicated; it's just a very large amount of data and
some very complex queries on that data- but the queries have been kept
to extremely standard SQL specifically for migration and
cross-platform reasons. It's definitely been annoying on occasion to
find that I need to do a dump and restore to move to a new version,
but at the same time cheap, large storage is extremely inexpensive
when compared to the sort of storage acceptable for day-to-day use, so
size isn't generally a problem- just dump to a big, cheap disk and
then restore. I'm probably lucky in that I manage a shop that can
tolerate a day's downtime for such a situation, but at the same time,
we also demand the most from database performance for complex queries,
so a day's downtime here could easily save many days' worth of query
time down the line.
8.3, FWIW, was particularly attractive in this regard. I couldn't
quite justify upgrading to the release candidates, but the performance
improvements were pretty tempting.
--
- David T. Wilson
david.t.wilson@gmail.com