| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Usenet -> c.d.o.server -> Re: LOG_CEHCKPOINT_INTERVAL AND RECOVERY
A copy of this was sent to boulke_at_globalnet.co.uk (keith boulton)
(if that email address didn't require changing)
On Sun, 28 Mar 1999 13:35:12 GMT, you wrote:
>On Sat, 27 Mar 1999 17:38:45 -0000, "Jonathan Lewis"
><jonathan_at_jlcomp.demon.co.uk> wrote:
>
>>It's probably the usual fuzzy-history syndrome,
>>in this case failure to discriminate between log
>>file switches and checkpoints.
>>
>>Log switches every N minutes makes sense if
>>you want to cater for no more than N minutes of
>>data loss when the machine is destroyed.
>
>Sorry for the tone of my previous response, I wasn't in a good mood at
>the time.
>
>I had thought about this, but surely this only applies if you are
>copying archived log files to a different machine.
>
No, it applies in all cases especially if you are getting your achived logs off of the machine. If you have your logs sized to switch every hour and you are archiving to tape on log switches and your disks blow up -- you can lose upto 1 hour of work since all copies of the online redo log are gone. If, on the other hand, the logs are sized to swith every N minutes, you stand a small change of losing N minutes of work....
Also, if you are archiving to the same machine it still applies. If you have your logs mirrored to 2 disks and have your archive going to a 3'rd (and/or 4'th with o8 and/or a 5'th, 6'th, 7'th with Oracle8i) you can have a catastrophic failure but not total failure of the machine and still limit your loss by limiting the amount of data in a redo log. If you lose all of the copies of the log (online and redo) -- you are going to lose a lot (thats why you want to get archive off of the machine in a timely basis). On the other hand, if you lose all of the online redo and some of the archive copies -- you'll only lose what has not been archived (and since you can archive to 2 places in O8 and 5 places in O8i, you can protect yourself pretty well). In O7.x, i sent up OS jobs to compress and multiplex the archive.
Its for disasters, not day to day stuff .
>I tend to work with small to medium sized databases, (because I like
>the dba/database designer/devloper combined role) where hardware cost
>is not an issue, so if the consequences of loss of a few hours data
>were significant, I would try to build my application to write to two
>databases on different machines.
Thomas Kyte
tkyte_at_us.oracle.com
Oracle Service Industries
Reston, VA USA
--
http://govt.us.oracle.com/ -- downloadable utilities
 |
 |