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: Increase frequency of log switches

Re: Increase frequency of log switches

From: Ed Stevens <nospam_at_noway.nohow>
Date: Fri, 09 Jul 2004 07:53:23 -0500
Message-ID: <rr4te055vt5bbg1ie79tt8gvnje5ppainp@4ax.com>


On 8 Jul 2004 15:06:19 -0700, joel-garry_at_home.com (Joel Garry) wrote:

>"Tom" <tomNOSPAM_at_teameazyriders.com> wrote in message news:<1089302964.20354.0_at_spandrell.news.uk.clara.net>...
>> > First, if you are properly set up in archive log mode, have duplexed
>> > redo log groups, and are taking regular backups, the amount of data
>> > lost in the event of "a major machine failure" is already near zero,
>> > and has nothing to do with the frequency of log switches.
>> >
>> > Second, the reason you don't find a parameter to control the maxsize
>> > of a redo log before a switch is because there is none. You're redo
>> > log files are a fixed size. A switch occurs when one fills up or you
>> > manually issue ALTER SYSTEM SWITCH LOGFILE.
>>
>> Thanks - Yes i already have everything running as you state and can recover the whole instance onto a test machine.
>>
>> My query though is that if online redo has not been archived then this is not included in the backup. Therefore the time from last
>> RMAN backup to failure could, in my case, be around a day. If i could have the online redo archived more often i would reduce this
>> potential data loss.
>>
>> Seems i need to run ALTER SYSTEM SWITCH LOGFILE every couple of hours to make sure i minimise this loss. This is as i don't backup
>> online redo as per the information i have found.
>>
>> FYI - when i say 'major machine failure' i mean the building burns down or all disks eat themselves at the same time. As i have
>> archived logs, backup control files and RMAN backups being duplicated offsite then i can maintain minimal data loss providing i can
>> reduce the time between log switches. Unless i have completely got the wrong end of the stick.
>>
>> thanks for all your help
>
>You seem to understand it a bit better than Ed here (sorry Ed!).
>Pre-9, it is common on unix systems to run periodic logfile switches
>in cron to achieve a given time granularity. 'Bout time they added
>that lag parameter.
>
>jg

No problem. I learn by getting corrected and I only get corrected when I stick my neck out.

And we're still pretty much Pre-9. I have one set of 9.2 running on Solaris, but that was all set up by the app vendor. This weekend we're converting a set of 8.1.7 on Win2k to 9.2. The real problem in trying to pick up on new features (besides a shortage of study time) is that my partner NEVER wants to use new features until he is forced to. We're still living with a lot of stuff we did in our own ignorance when we got started with Oracle at 7.3 (things like bad naming conventions, very-non-OFA file placements, user permissions, etc.) and I can't get him to agree to go back and change any of it. Puts a real drag on my own learning opportunities. Received on Fri Jul 09 2004 - 07:53:23 CDT

Original text of this message

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