Noons wrote:
> Kewl, no apparent problem there.
>
>> filesys EXT3
>
>
> Content of /etc/fstab?
/dev/hda1 / ext3 defaults 1 1
none /dev/pts devpts mode=0620 0 0
/dev/hda8 /home ext3 defaults 1 2
none /mnt/cdrom supermount dev=/dev/scd0,fs=udf:iso9660,ro,--,iocharset=iso8859-15 0 0
none /proc proc defaults 0 0
/dev/hda9 /scratch ext3 defaults 1 2
/dev/hda6 /usr ext3 defaults 1 2
/dev/hda7 /var ext3 defaults 1 2
/dev/hda5 swap swap defaults 0 0
>
>> There are lots of "*init*ora", but I think
>> I've found the right ones
>
>
> Sorry, I should have provided more info.
> 9.2/dbs/initscan.ora sounds like the right one:
> it usually is in either $ORACLE_HOME/dbs or a
> symbolic link to it is there. Make sure the
> database is not using a SPFILE (binary) in which
> case this file is irrelevant.
> To check, log on to sqlplus as sysdba and do:
>
> select * from v$spparameter
> where value is not null;
>
> If that returns no rows you are using a pfile
> instead of a spfile and the following is valid.
> Otherwise, you have to do all the work using
> Enterprise Manager (and that is another pain...)
So
"32 rows selected." indicates pain?
>
>> (9.2/dbs/initscan.ora)
>> (decommented for space)
>> db_block_size=8192
>
> ***bzzt> db_cache_size=33554432
> make it =50M at least
>
>> db_file_multiblock_read_count=16
>> open_cursors=300
>> db_domain=################### (not shown on USENET)
>> db_name=scan
>> background_dump_dest=/home/oracle/admin/scan/bdump
>> core_dump_dest=/home/oracle/admin/scan/cdump
>> timed_statistics=TRUE
>> user_dump_dest=/home/oracle/admin/scan/udump
>
> ***bzzt> control_files=("/home/oracle/oradata/scan/control01.ctl",
>
>> "/home/oracle/oradata/scan/control02.ctl",
>> "/home/oracle/oradata/scan/control03.ctl")
>
> if you are putting all your control files in the same directory,
> there is no need to triplicate them, one only is plenty! Multiple
> control files is for multiplexing to avoid data loss. Right now
> you are doing a "tobesure" thrice...
>
>> instance_name=scan
>> job_queue_processes=10
>> dispatchers="(PROTOCOL=TCP) (SERVICE=scanXDB)"
>> aq_tm_processes=1
>> compatible=9.2.0.0.0
>> hash_join_enabled=TRUE
>> query_rewrite_enabled=FALSE
>> star_transformation_enabled=FALSE
>
> ***bzzt> java_pool_size=83886080
> if your java client is not inside the database, drop this to 0
>
>> large_pool_size=16777216
>
> ***bzzt> shared_pool_size=83886080
> get in the habit of specifying these sizes as M, in this case = 80M
> ***bzzt> processes=150
> a bit too high for a small test? Drop down to 50?
>
>> fast_start_mttr_target=300
>> remote_login_passwordfile=EXCLUSIVE
>> pga_aggregate_target=25165824
>
> ***bzzt> sort_area_size=524288
> when you use pga_aggregate_target, this one is redundant: drop it.
>
>> undo_management=AUTO
>> undo_retention=10800
>> undo_tablespace=UNDOTBS1
>
> and now add these:
> log_buffer=1048576
> log_checkpoint_interval=10000
> db_writer_processes=2
>
>
> and make triple sure you have "noatime" as one of the
> mount parameters for file system where all dbfiles are.
>
> Check the size of your redo log files. If they are anything
> less than 100Mb, make them that size (you need a min of 3).
>
> You can do that by allocating additional ones of the right size
> then dropping the old ones, check the SQL manual for that (ALTER DATABASE):
> http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96540/toc.htm
> (you will need an email to register, it's ok: they don't spam)
> If it doesn't let you drop a logfile in use, ALTER SYSTEM SWITCH
> LOGFILE toggles to the next one. Of course: you need dba powers
> for all this.
>
> While you are at it go to tahiti.oracle.com, it's a front-end for
> all manuals. Look for 9i ones, Concepts manual: it's a leisure read
> for the weekend and it will 'splain a lot. Same for Application
> Developer's Guide and the JDBC User's Guide. All the params
> above are on Reference manual if you feel curious.
>
> Anything else I can help with ping me on the email, I'm
> nursing databases this weekend and cannot be on Usenet.
No problem; you've already greatly exceed the call
of USENET duty.
> Yup, 1 f/s is a killer for 9i. You definitely should put redo logs
> somewhere else. 10g does all sorts of default magic to get around
> this problem, hence why you are likely to not feel it. 9i just
> needs a bit more TLC.
I'm getting that impression.
>
>> But we would like our test to finish!!
>
>
> Fair enough! :)
>
Received on Thu Jun 30 2005 - 10:59:02 CDT
