Re: Full export and sequence behaviour in 8i

From: Gints Plivna <gints.plivna_at_gmail.com>
Date: Mon, 16 Mar 2009 11:43:41 +0200
Message-ID: <6e49b6d00903160243v3b68e7c6of69f96f8128038f3_at_mail.gmail.com>



2009/3/16 Dan Norris <dannorris_at_dannorris.com>:
> Never is a word I don't see too often. I think a "nocache order" sequence is
> better and more reliable than any other method I've seen. What approach
> would you use instead to have a no-gap list of numbers guaranteed?

Word "guaranteed" isn't appropriate here. Maybe word "less probability" to get gaps could be used to explain this situation. As Neil already explained there are certain scenarios which can grab sequence value and as soon as it is grabbed you'll never get it back. Of course probablity for DB crash and (depending on your app) to rollback a transaction is less than aging sequence out of cache or DB restart, which certainly looses cached sequence values. Also nocached sequence might be sort of bottleneck especially in RAC scenarios. Of course it depends on sequence usage rate. And this requirement to get guaranteed numbers without gaps is asking for bottlenecks per se as I've tried to explain here
http://www.gplivna.eu/papers/gapless_sequences.htm

Gints Plivna
http://www.gplivna.eu

--
http://www.freelists.org/webpage/oracle-l
Received on Mon Mar 16 2009 - 04:43:41 CDT

Original text of this message