Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> RE: redo log space requests growing and log stats : beyond the basics

RE: redo log space requests growing and log stats : beyond the basics

From: Steve Adams <steve.adams_at_ixora.com.au>
Date: Fri, 11 Aug 2000 16:46:19 +1000
Message-Id: <10585.114332@fatcity.com>


Hi Tom,

Yes, you have clearly gained something here with a larger log buffer, and in a batch environment that is more important than the commit response time. Anyway, a 560K log buffer is not exactly huge. It is just towards the upper end of the "normal" range, and the negative impact on commit response time from needing to do bigger syncs would be no more than 20ms. If your redo generation is intensive, then you may also be doing better in terms of reduced contention for the 'redo allocation' latch. So what you've done seems right.

However, for future reference, the 'redo buffer allocation retries' statistic is not as helpful here as 'log buffer space' waits, because it includes retries after log switches as well as those related to buffer space.

Also, your suggestion that you have 8192 byte log buffer blocks surprises me. I am not aware of any platform on which Oracle uses an 8K log block size. Could you please check using my log_block_size.sql script.

Thanks,
Steve Adams

http://www.ixora.com.au/
http://www.oreilly.com/catalog/orinternals/
http://www.christianity.net.au/


-----Original Message-----
From: Tom Pall [mailto:tpall_at_cdproc.com] Sent: Tuesday, 8 August 2000 2:27
To: Steve Adams; J. Laurindo Chiappa; DBAs - TeleLists; LazyDBA/Quickdocs -
Subject: Re: redo log space requests growing and log stats : beyond the basics

Steve,

How is the too large lob_buffer damaging commit response times?

I ask this because I've been upping log_buffers on an at times very batch intensive database from
20 8192 byte buffers and 8,700 redo buffer allocation retries per day to current 70 8192 byte buffers,
and 261 redo buffer allocation retries per day.

> Hi Laurindo,
>
> Your 'redo log space requests' are a symptom of log switches, not an
> indication that the log buffer is too small. Indeed your log buffer is
> too large and this is damaging your commit response times. There is a
> detailed explanation of these issues available on my web site, as well
> as an explanation of 'redo wastage' and some of other issues you have
> raised.
>
> Regards,
> Steve Adams
> http://www.ixora.com.au/
> http://www.oreilly.com/catalog/orinternals/
> http://www.christianity.net.au/
>
>
> -----Original Message-----
> From: J. Laurindo Chiappa [mailto:jlchiappa_at_uol.com.br]
> Sent: Wednesday, 2 August 2000 10:21
> To: DBAs - TeleLists; LazyDBA/Quickdocs -; ODTUG Oracle DBA List;
VOUG -
> DBAs Virginia
> Subject: redo log space requests growing and log stats : beyound the
> basics
>
>
> Hi ! I'm receiving many redo log spacerequests in my prod db; I
followed
> the common steps in this case (add
> more log_buffer space; more log member, etc , but no avail. the
hardware
> is : IBM F50 with AIX 4.3.3, 2
> processors, 1Gb RAM, 1 local (non-raid) disk (with the logs and the
> control files), and 1 array of disks (due to
> hardware reasons, unfortunately, mounted as RAID-5), db version is
> 7.3.4.5, and Oracle is the only app in
> the box.
> Parameters of the db :
>
> pre_page_sga FALSE
> log_archive_start FALSE
> log_archive_buffers 4
> log_archive_buffer_size 127
> log_archive_dest ?/dbs/arch
> log_archive_format %t_%s.dbf
> log_buffer 20971520
> log_checkpoint_interval 0
> log_checkpoint_timeout 0
> log_block_checksum FALSE
> log_small_entry_max_size 80
> log_simultaneous_copies 4
> log_files 255
> log_checkpoints_to_alert TRUE
> delayed_logging_block_cleanouts TRUE
> dblink_encrypt_login FALSE
> db_block_buffers 22000
> db_block_checksum FALSE
> db_block_size 4096
> db_block_checkpoint_batch 8
> db_block_lru_statistics FALSE
> db_block_lru_extended_statistics 0
> db_block_lru_latches 2
> db_files 20
> db_file_simultaneous_writes 4
> db_file_multiblock_read_count 16
> gc_db_locks 100
> dblink_encrypt_login FALSE
> db_domain WORLD
> db_name prod
> shared_pool_size 130000000
> shared_pool_reserved_size 0
> shared_pool_reserved_min_alloc 5000
>
> (OBS : I have doubts about the shared_pool_reserved_size=0 and the
> shared_pool_reserved_min_alloc=5000 ).
>
> Note the obvious points : big log buffer (20 Mb) , 2 db writers , 2
> db_block_lru_latches and 4
> db_file_simultaneous_writes (because of the 2 processors), big
> db_blocks (22000, meanining more than 90
> Mbs), but no luck . See the stats :
>
> elect name, value from v$sysstat where name like '%redo%';
> NAME VALUE
> ------------------------------ -----------------
> redo synch writes 52505.0000
> redo synch time 103900.0000
> redo entries 1033556.0000
> redo size 234973983.0000
> redo entries linearized .0000
> redo buffer allocation retries 4.0000
> redo small copies 289873.0000
> redo wastage 17960239.0000
> redo writer latching time 80.0000
> redo writes 57997.0000
> redo blocks written 511296.0000
> redo write time 109835.0000
> redo log space requests 4.0000
> redo log space wait time 172.0000
> redo log switch interrupts .0000
> redo ordering marks .0000
>
> Log switch is reasonably :
>
> sys_at_prod::SQL>select * from v$log
>
> GROUP# THREAD# SEQUENCE# BYTES MEMBERS ARC STATUS
> FIRS
> T_CHANGE# FIRST_TIME
> ---------- ---------- ---------- ---------- ---------- --- -----------

--

> --- ----
> --------- --------------------
> 1 1 2545 83886080 1 NO CURRENT
> 5363474 08/01/00 16:50:41
> 2 1 2543 83886080 1 NO INACTIVE
> 5325532 08/01/00 15:13:58
> 3 1 2544 83886080 1 NO INACTIVE
> 5348822 08/01/00 15:45:33
>
> OBS : the db was bounced at 07:00 and the stats were queried at 17:00
> (only a day of work, and 4 redo log
> space requests! )
> I'm concerning about the high redo wastage, too, and the wait time.
> Some ideas ?
>
> []s
> Chiappa
>
>
> --------
> If you're bored, then visit the list's website: http://www.lazydba.com
> (updated daily)
> to unsubscribe, send a blank email to
> oracledba-unsubscribe_at_quickdoc.co.uk
> to subscribe send a blank email to oracledba-subscribe_at_quickdoc.co.uk
>
>
> --------
> If you're bored, then visit the list's website: http://www.lazydba.com
Received on Fri Aug 11 2000 - 01:46:19 CDT

Original text of this message

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