Are you identifying the fact that ARCH is archiving Group3 but the LGWR is
attempting to write to Group1
from some messages in the alert.log ?
Do those messages say "checkpoint not complete" ?
When LGWR finishes writing to a Logfile (eg Group3), it does three things :
a) signals ARCH to start archiving it
b) signals DBWR to execute a database checkpoint
c) checks if the next Logfile (Group1) is available for writing and, if
so, overwrrites it,
if not, starts waiting
It could be so happening that either even if a) has completed, b) has not
yet completed,
so LGWR will not overwrite the Logfile.
Solutions :
a) I/O performance tuning / load balancing to increase the speed of
the checkpoints
(eg even actually reducing the db_block_buffers ! -- to reduce the number
of dirty buffers
to be written)
b) add more log files so that LGWR does not have to wrap from Group3 to
Group1 but
can go to Group4, Group5, Group6 before it comes back to Group1, by which time
DBWR should have finished the checkpoint
c) create larger log files so that LGWR spends more time inside each
Logfile before
attempting to switch to the next one
each has it's own pros and cons ! (eg if you have a Standby Database, you
would *want*
more frequent Log Switches (eg smaller log files or a job issuing "alter
system archive log current"))
Hemant
I'm having a big problem my database is stopping for like 20 seconds because it can not write to it redo log
As it looks it seems to be full, but this should not be. The acriver is archiving group 3 and the lgwr wants to write to group 1. Is there anyone how has had this problem before.
Or who might have some tips on where or how i can solve this.
I have looked for I/o problme but this does not seen to be the problem
-- Vincent Verpoort
,.-~`"'~-.,_,.-~`"'~-.,_,.-~`"'~-.,_,.-~`"'~-.,_,.-~`"'~-.,_ Communiceren is begrepen worden ^*<-._,.->*^*<-._,.->*^*<-._,.->*^*<-._,.->*^*<-._,.->*^*<-.
Hemant K Chitale http://web.singnet.com.sg/~hkchital
-- //www.freelists.org/webpage/oracle-l
