Re: very (I repeat) slow DBWR

From: Piotr Kolodziej <pkol_at_otago.gda.pl>
Date: Wed, 5 May 1999 12:32:03 +0200
Message-ID: <7gp743$ehg$1@korweta.task.gda.pl>

[Gerard H. Pille <ghp_at_skynet.be>]
>redo allocation 298484 120 1 1
>008
>redo copy 1172 1157 .013 1530
>1.322

What's the value of log_simultaneous_copies ? Specify it 16 as 2*cpu_count (default=cpu_count) in init<sid>.ora. But it is not directly related to dbwr.

>TABLE_SPACE FILE_NAME
>READS BLKS_READ READ_TIME WRITES BLKS_WRT WRITE_TIME MEGABYTES
>------------------------------
>------------------------------------------------
>---------- ---------- ---------- ---------- ---------- ----------
>----------
>ORADATA /misb/oradata02MIS.dbf
> 0 0 0 11368 11368 122441112
>1074
>ORADATA /misb/oradata03MIS.dbf
> 15 15 21 4003 4003 38897899
>944

Oupsss.... However Oracle docs say that write time stats may be misleading, but I don't think that in such large scale.... I'm affraid dbwr may wait for sth (lgwr? have you checkpoint process started ?).
Do they contain temp segments ? It looks a bit like that ... If so configure using sort_direct_writes. What about waiting event sections (background and non-background processes) in report? It is very interesting. What's the value of db_block_lru_latches ? If it is default value (=cpu_count/2) set it to 8 (=cpu_count) when possible.

--
Piotr Kolodziej pkol_at_otago.gda.pl
Just my private opinion. Received on Wed May 05 1999 - 05:32:03 CDT