home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.databases.ingres
- Path: sparky!uunet!mcsun!sun4nl!rivm!jwd
- From: jwd@rivm.nl (Jan Diesel)
- Subject: Re: Transaction Log File
- Message-ID: <C0oo00.BCA@rivm.nl>
- Sender: news@rivm.nl
- Organization: R.I.V.M.
- X-Newsreader: Tin 1.1 PL5
- References: <1993Jan8.144837.9232@tpghq.com>
- Date: Mon, 11 Jan 1993 09:36:00 GMT
- Lines: 100
-
- Steve Caswell (sfc@tpghq.com) wrote:
- : In article <C0J1q9.FvG@rivm.nl> jwd@rivm.nl (Jan Diesel) writes:
- : >David Kaplan (djk@sei.cmu.edu) wrote:
- : >:
- : >: I'm having some intermittent problems with the Trasaction Log File. Or,
- : >: rather, it's having trouble with me? In either case, the following error
- : >: occurs when running a rather long set of queries:
- : >:
- : >: **** ERROR: E_US1262 Your transaction has been aborted due to the transaction
- : >: log file having reached one of the limits set by the system administrator.
- : >: These limits are log_full, force_abort, and 90 percent of force_abort
- : >: when using the fast_commit option to start DBMS servers.
- : >: (Thu Jan 7 12:21:39 1993)
- : >:
- : >: What's this all about. I mentioned intermittent because some days I get the
- : >: error, and other days I don't?! There are commits after each query (all
- : >: creation and insertion commands), and the queries are broken up just about as
- : >: far as we can get them.
- : >
- : >The transaction logfile is used not only by inserts, updates and deletes but
- : >by selects as well. The obvious solution to your problem seems increasing the
- :
- : Unless something drastic has changed, selects ARE NOT LOGGED to the log file.
-
- I don't know where I got this information but you're right and I'm wrong:
- selects are not logged (unfortunately I can't find this in the manuals).
-
- :
- : >size of the transaction logfile or decreasing the size of your transactions.
- : >The latter can be done by setting autocommit on (thereby causing each statement
- : >to be treated as a singleton transaction). If this won't help and you don't
- : >want to increase the transaction logfile you'll have to split up your statements.
- : >
- : Setting autocommit on is not a good thing to do in normal application
- : situations. This command is in INGRES primarily as a conversion tool to allow
- : converted release 5 applications to behave under 6 as they did in 5. Using
- : set autocommit on in a production 6 application is not a good idea for three
- : reasons:
- :
- : 1. It is NOT ANSI standard. It in fact makes transaction semantics behave
- : exactly opposite of ANSI standard. If someone comes in later to
- : maintain your code, and is not aware that set autocommit on has been
- : executed, and is expecting ANSI transaction semantic behavior, they will
- : be in for a rude surprise.
- :
- : 2. Because set autocommit on is a conversion feature, there is no guarantee
- : that is will be supported in the future when Ingres decides that everyone
- : has converted and they remove the command.
- :
- : 3. You lose rollback capability. Unless EVERY statement in an application
- : is a single statement transaction, you are giving up a lot. Even if
- : every statement is a single statement, the incremental cost of putting
- : in a commit statement is minor.
-
- I agree with you on most points, but I do think that 'autocommit on' is a usefull
- ingres-feature. By the enforcement of singleton transactions endusers aren't
- bothered with the transaction mechanism anymore. Unless autocommit is set off
- most interference with more complex applications (not using singleton transactions)
- is no longer a problem.
- The only price that has to be paid is setting autocommit off in your applications.
- (this also makes sure you won't suffer from removing the autocommit command,
- which I hope won't happen).
-
- : >:
- : >: Could this error be an indicator that something else is wrong? The fact that
- : >: it is intermittant EVEN when the data being used has not change is what really
- : >: has me puzzled.
- : >:
- : >: Any clues appreciated ;-)
- : >:
- : >: Thanks,
- : >:
- : >: -Dave.
- : >:
- : >:
- : >: ---------------------------------------------------------------------------
- : >: David J. Kaplan djk@sei.cmu.edu
- : >: Software Engineering Institute dk3g+@andrew.cmu.edu
- : >: Carnegie Mellon University (412) 268-7736
- : >
- : >Regards,
- : >-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
- : >Jan Diesel | jwd@rivm.nl
- : >Informatics Service Centre | tel. ++31 30 742067
- : >P.O.Box 1, 3720 BA BILTHOVEN, the Netherlands | fax. ++31 30 282316
- : >RIVM - National Institute of Public Health and Environmental Protection
- :
- :
- : --
- :
- : Steve Caswell | (404) 448-7727 | "The opinions expressed are my
- : Principal Consultant | sfc@tpghq.com | own. They may not be perfect,
- : The Palmer Group | uunet!tpghq!sfc | but they're all I've got."
-
- --
- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
- Jan Diesel | jwd@rivm.nl
- Informatics Service Centre | tel. ++31 30 742067
- P.O.Box 1, 3720 BA BILTHOVEN, the Netherlands | fax. ++31 30 282316
- RIVM - National Institute of Public Health and Environmental Protection
-