Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: reduce archive logs

Re: reduce archive logs

From: Howard J. Rogers <howardjr2000_at_yahoo.com.au>
Date: Sat, 17 May 2003 06:32:10 +1000
Message-ID: <3Kbxa.35666$1s1.518547@newsfeeds.bigpond.com>

"bcjm" <oldcar_at_hotmail.com> wrote in message

> First of all, thank you for all the replies. Setting the table to
> nologging does not work since nologging only works in batch or direct
> load. I assume moving them to a nologging tablespace is the same as
> setting the table to nologging.

Correct: it just means that any table/other sort of segment that gets created within that tablespace gets the NOLOGGING attribute set by default... which is then promptly ignored in all cases except the rare bacth or direct loads you mentioned!

>If this is what I have to live with
> that is fine. I just want to make sure there is nothing else I can
> do.
>
> Having said that, do you all adjust the redo log size periodically?

Well, it isn't easy to adjust the size of redo logs. Instead you have to add new ones, and drop old ones. And do log switches etc when Oracle complains that you can't drop this group because it's the Current one. So no... it's not usual to 'periodically' adjust the redo. Certainly when first setting the thing up. Yes, just after it hits production for the first time (or should that rather be, when poduction hits it?!). And maybe if something radical happens to the nature of the work being done on the database. Under all those circumstances, I'd probably be having a look to see if my previous log size was still up to the task.

>I
> was told to keep it about one log file/hour is a good rule of thumb.

It's not a bad rule of thumb, but it is just a rule of thumb. Some people prefer once every half hour. Some prefer no switches at all. It all comes down to two things: how much data can you afford to lose if, perish the thought, you lose the current redo log (which you won't, of course, because you've multiplexed all your groups, as well as hardware-mirrored them, haven't you?). And second: in the event of an instance crash, how long do you want your instance recoveries to take (although in 8i and 9i, the two issues can be more or less separated by switching on incremental checkpointing, thus meaning that the log switch-induced checkpoint is not the sole determinant of when checkpoints occurr).

> Too many log files (switches) reduce the performance and hard to
> manage.

I'd go along with that.

>keeping giant log file is not good either (why?).

That one I won't buy, but as for the why, see above: enormous logs which never switch, and then which disappear in a puff of blue smoke, mean lots of data is lost. (My argument is that we can configure resilience and redundancy into the system, so this shouldn't be a concern. But I guess if you stick lots of eggs into large baskets, you're just waiting for the lot of them to be smashed).

Regards
HJR Received on Fri May 16 2003 - 15:32:10 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US