Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: NOLOGGING - too much logging!
The is no way to turn off redo / rollback geneation for dictionary
operations, e.g. updates to FET$ / UET$ must be written (new extent
creation). Setting a Table / Index to NOLOGGING ensures that Create Table As
Select, Insert direct path and sqlload direct path operations do not
generate rollback/redo, but standard DML operations still generate
rollback/redo. I have a number of customers who would like to have the
database in NOLOGGING mode, maybe Oracle9 will offer this feature.
To find out how much redo is being generated check v$sysstat for "redo blocks written" (check the statistic# from v$statname).
Checkpoint not complete means that logwritter is waiting for a checkpoint to complete on the first log group before switching into the log, you need to resize your redo logs (this WILL be affecting performance).
Regards
David Russell
Steve Haynes wrote in message <2si5uKA+6We1Ew5u_at_rwx777.demon.co.uk>...
>Hi All,
>RE: 8.0.4 on NT.
>I'm designing a small (2GB) datamart and since it
>is easily recreated, there is no need for backups,
>so all redo is a waste of time.
>I made all tables and indexes no logging but was
>amazed to see lots of redo still happening.
>When I checked all_tables I found all the oracle
>internal stuff set to logging=YES, can I make all that
>NO as well?
>The main question is exactly how I can tell the
>amount of logging going on, the alert log tells
>me "swtiching to thread 1" etc about every 20 secs
>(I have 5X5Mb logfiles), this seemed too much
>so I check trc files from CKPT and something else
>(sorry can't recall) and these indicated logfile
>switches every hour or so. What do I believe?!
>I sometimes see "checkpoint not complete" in
>the alert log too. Any info around on this, all
>my million books don't go deep enough.
>Thanks
>Steve
>--
>"The floggings will continue until morale improves."
>
Received on Sat Jun 06 1998 - 16:41:48 CDT