Re: v5.05 Logical Log Backups

In article 807560295snz@fourgee.demon.co.uk, Rus Ambler <rizzo@fourgee.demon.co.uk> writes: >Anyone out here who can answer this little conundrum? > >One of our customers has a v5.05 database utilizing 15 10MB logical logs (I >know that they need to be increased). They do not run continious log backups >because they have the entire disk mirrored on a different site, so they >don't care about keeping the LL's on tape. They have the tape device for the >LL's set to /dev/null. > >My understanding of this was that as the logs become full, and once any work >within them is marked as complete (ie, any BEGIN WORK's have been correctly >paired with a ROLLBACK/COMMIT), the log would be cleaned and marked as free. > >What's actually happening is that the logs are sitting there and only a >tbtape -a is actually zapping them (marking them as backed up). > >This works OK on other sites, so any ideas what's gone wrong here? Your understanding is basically correct, and it works fine here under 5.05.UC1. The problem may be this: I have found that, if a tape device has previously been specified in LTAPEDEV and the instance restarted - and later, LTAPEDEV is set to /dev/null, you will have to do a tbtape -a to clear logical logs. If you think this is the case, have the customer restart the On-Line instance in question. The problem should then go away. Bill Williams E-Mail: bwilliam@jaxnet.com Flames: nobody@nowhere.com -or- redirect to /dev/null +-----------------------------------------------------------------+ | Disclaimer: My views, opinions, and judgments are my own. | +-----------------------------------------------------------------+ | "Free scientific inquiry" is a redundancy. | | "Free governmental inquiry" is self-contradictory. | | | | Be sure to read "Atlas Shrugged" by Ayn Rand. It's the world's | | greatest novel, rated #2 of those books readers said had the | | most impact on their lives (NY Times readership poll)! | +-----------------------------------------------------------------+