Printable View
Hi
I have Log file size of 51M, and obviously, when ever this gets filled, log switch occurs and archive log file of 51M will be created. But very few times I am finding that, log file is switching without getting filled completly, that at 48M or less than 51M.
I am sure no other program is swithching log automatically. Can any one give any idea on this. And this his happening very few times. May be once or twice in a day
Thanks
Kishore
Hi,
Is this happening at specific times like during hot backup? Generally at the end of the hot backup logfile switch is done to get an archived logfile.
OR any other process is issuing "ALTER SYSTEM SWITCH LOGFILE" command?
Baliga
I noticed the same thing on my instances. However, a couple of my archived logs are 1 OS block smaller than the size I expect, not 3M different. hmmm....
I have two files at present, one is 49.99 M and another is 50M and all others are 51M only. And it is not during HOT BACKUP.
What about the possibility that Oracle won't span a transaction across redo logs? I'm pretty sure this is true....it oughta be, but I couldn't find a confirming reference.
You doing any huge transactions?
I also feel what JDoyle just mentioned; May be you have long running transactions that can't finish in the remaining portion of the current log and so Oracle needs to switch log becasue it won't allow log spanning for one transaction (something of-course to confirm)
Are you getting errors like followoing in the Alert log --
Thread 1 cannot allocate new log, sequence ....
Checkpoint not complete
- Rajeev
That's definitely not true. If you had 1M redo logs you could only have 1M worth of transactions. Transactions definitely do span redo logs.Quote:
Originally posted by JDoyle
What about the possibility that Oracle won't span a transaction across redo logs?
The test:
[email protected]> select max( SEQUENCE#) from v$loghist; -- show me what the current log seq# is
MAX(SEQUENCE#)
--------------
519
[email protected]> alter system switch logfile; -- change the seq#
System altered.
[email protected]> select max( SEQUENCE#) from v$loghist;
MAX(SEQUENCE#)
--------------
520
[email protected]> commit; -- start the TX
Commit complete.
[email protected]> declare -- big TX
2 i number;
3 begin
4 for i in 1..100000 loop
5 insert into jeff.jh_test1 values (i,lpad(to_char(i),20,'_'));
6 end loop;
7 end;
8 /
PL/SQL procedure successfully completed.
[email protected]> select max( SEQUENCE#) from v$loghist; -- get seq#; still in TX
MAX(SEQUENCE#)
--------------
565
[email protected]> commit; -- TX over
Commit complete.
[email protected]> select max( SEQUENCE#) from v$loghist; -- 40 logs were archived during this time, all the correct size
MAX(SEQUENCE#)
--------------
565
[email protected]>
[Edited by marist89 on 01-29-2001 at 03:45 PM]
I don't find any errors in alert logs. I am only qurious about what's heppening inside. Even not 'CANNOT ALLOCATE THREAD 1'. And I feel that Transaction can span Redo Logs.
what Oracle Database Release you are in????
8.0.6