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: Optimal size for Redo Logs?

Re: Optimal size for Redo Logs?

From: Joel Garry <joel-garry_at_home.com>
Date: 21 Oct 2003 14:15:42 -0700
Message-ID: <91884734.0310211315.492091b0@posting.google.com>


hari_om_at_hotmail.com (Hari Om) wrote in message news:<d1d5ebe4.0310171316.5c2988f9_at_posting.google.com>...
> I am running DB in ARCHIVEMODE. Using Oracle 9.2.0.1
>
> What is the optimal size for the Redo Logs....I have heard if Redo
> Logs are larger then LGWR would call ARCH few times then if it would
> have been smaller...is that right....

It may or may not be, depends on your transaction volume and type and redo log buffer size and SGA size and some init parameters and disk drive speeds and whether people are modifying contentious data blocks and what kind of recovery time you want. Optimal size can only be determined empirically using a recursive tuning methodology with this many variables. I usually size the logfiles to the largest volume generally expected and force log switches to a particular time granularity with a cron job. But's that's because the systems I work on usually have a small proportion of high-volume loads or updates and a large proportion of online transaction processing. In the classical tuning scenario, you would start writing the buffers to the logs in time to avoid a wait, with the buffers big enough to not keep people waiting to write to them.

>
> Say I have 3 log files... Log1, Log2, Log3 each 100M and say LGWR is
> currently using Log2. if Log2 becomes Full...LGWR would call ARCH
> which would then write Log2 to the archive destination. At this stage
> is LGWR idle (waiting for ARCH to finish) or is it busy making logs to
> Log3....? I mean would there be any contention between these 2
> processes? Can anyone tell me the BIG picture?

It will temporarily hang your db if it is not tuned right, that is, it will be busy writing Log3, and if it finishes before ARCH finishes with Log2, users scream. If they are writing to the same disk device, there may be severe contention.

Like everyone said, read the concepts manuals and learn the basics. google for "checkpoint not complete" and "redo log buffers."

>
> THANKS!
jg

--
@home.com is bogus.
http://www.signonsandiego.com/news/uniontrib/mon/currents/news_1c20air.html
Received on Tue Oct 21 2003 - 16:15:42 CDT

Original text of this message

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