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: log switching question

Re: log switching question

From: Steve Adams <steve.adams_at_ixora.com.au>
Date: Tue, 31 Oct 2000 11:58:14 GMT
Message-ID: <39feaecb.466051086@nsw.nnrp.telstra.net>

Hi Scott,

Yes, what the analyst told you is rubbish. The only reasons I know of for a log file switch are if you do it manually using either the SWITCH LOGFILE or the ARCHIVE LOG clauses of the ALTER SYSTEM command, or if it happens automatically because the current log file is full, or in parallel server to secure recoverability in a relatively idle instance.

I'd firstly check that the logfiles are all the right size as follows:

        select lesiz, count(*) from x$kccle group by lesiz; Then if you are certain that nothing is doing ALTER SYSTEM commands, I would reopen the TAR and press for a real explanation.

@ Regards,
@ Steve Adams
@ http://www.ixora.com.au/
@ http://www.christianity.net.au/

-----Original Message-----
From: gambleNOSPAM_at_iwc.net (Scott Gamble)

Hello everyone,

   Looking for some information on log_checkpoint_interval with respect to log switches.

This may be purely a complete misunderstanding on my part, but I think I am missing something here or I cannot find the appropriate documentation.

We have a database that is using 1 gig log files. and we are doing log switches/checkpoints _very_ frequently, in some cases in less than 10 minutes and in _most_ cases those log files are less than the full 1 gig, in fact they are considerablly less, in the area of 100-200m. We think this started after the recent upgrade to 8.0.6 but are not 100% sure.

To start with here are the parameters as we have them defined.

log_checkpoint_interval             integer 60000000                      
log_checkpoint_timeout              integer 0                             

From what I have read log_checkpoint_timout, and log_checkpoint_interval control checkpoints. Neither of which should force a log switch based on the documention.  

Oracle seems to present conflicting information on this.

The analyst that handled the TAR i opened said that a checkpoint will force a log switch and the solution to slow the check points down was to raise log_checkpoint_interval to a number that could not be reached prior to the log filling up.

From what I understand log_checkpoint_interval is the number of OS blocks, 60M OS blocks is bigger than our log size. Obviously I am missing something here.   

Looking at a post in Metalink, that oracle people replied to, what the analyst told me above is incorrect.

I guess my question comes down to this

  1. in the abscence of anyone doing an 'alter system switch logfile' why are the log files switching when they aren't full.

Portion of v$archived_log

    BLOCKS BLOCK_SIZE TO_CHAR(COMPLE
---------- ---------- --------------

    460952       1024 10/30/00 10:27
   1024000       1024 10/30/00 11:54
    226364       1024 10/30/00 11:02
    405973       1024 10/30/00 11:25



Portion of the alert log....

Mon Oct 30 10:55:58 2000
Beginning log switch checkpoint up to RBA [0x6a1.2.10], SCN: 0x0000.64cf083a
Thread 1 advanced to log sequence 1697
  Current log# 4 seq# 1697 mem# 0: /u16/oradata/P013/redo4aP013.log   Current log# 4 seq# 1697 mem# 1: /u17/oradata/P013/redo4bP013.log Mon Oct 30 11:02:11 2000
Beginning log switch checkpoint up to RBA [0x6a2.2.10], SCN: 0x0000.64cf15d5
Thread 1 advanced to log sequence 1698
  Current log# 5 seq# 1698 mem# 0: /u17/oradata/P013/redo5aP013.log   Current log# 5 seq# 1698 mem# 1: /u18/oradata/P013/redo5bP013.log Received on Tue Oct 31 2000 - 05:58:14 CST

Original text of this message

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