Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!ames!amdahl!rtech!squid!robf From: robf@squid.rtech.com (Robert Fair) Newsgroups: comp.databases Subject: Re: Referential Integrity in commercial DBMS's? Keywords: INGRES RTI Message-ID: <3632@rtech.rtech.com> Date: 14 Sep 89 12:27:34 GMT References: <5390@tank.uchicago.edu> Sender: news@rtech.rtech.com Reply-To: robf@squid.UUCP (Robert Fair) Organization: Relational Technology, Inc. Lines: 30 >From: supp@tank.uchicago.edu (Steve Upp) writes on a number of topics (most of which it is not appropriate for me to coment on, especially bugs not getting fixed), and also mentions: >If the backend crashes who knows what state your data will be left in! This should normally be no problem in INGRES release 6 - all logging and recovery is done through a independent subsystem (DMFRCP & DMFACP) which is still running if a DBMS server goes down. These will rollback any non-committed transactions, or rollforward any transactions which have been committed but not yet updated to the database. (e.g. with Fast Commit), so your data should remain in a consistent state. All the transaction info is stored in a logfile (often a raw device on UNIX) so that in the event of a machine crash the recovery process will automatically rollback/redo any transactions when it comes back up later. About the only way data can get lost is if the logfile is manually removed - and if you go manually removing files from a DBMS all bets are off anyway... On rules in 7.0, they really do work very nicely, especially with cascading/recursive style updates - I just wish they could have been discussed a few weeks ago in the referential integrity thread, but we try not to discuss new features on the net before they are available. (Rules is just one of many equally significant features in 7.0 - look out for announcements in the near future) Robert L. Fair Technical Support Relational Technology, Inc