Xref: alice comp.databases.oracle.server:48243
Path: alice!news-feed.fnsi.net!news.maxwell.syr.edu!news.belnet.be!krypton.inbe.net!INbe.net!not-for-mail
From: "Gerard H. Pille" <ghp@skynet.be>
Newsgroups: comp.databases.oracle.server
Subject: AIX: very (I repeat) slow DBWR
Date: 5 May 1999 08:06:49 GMT
Organization: Creyf's
Lines: 52
Message-ID: <01be96cf$41038920$390120c0@GHP.creyfs.be>
X-Newsreader: Microsoft Internet News 4.70.1161

Hai Folks,

I have a huge performance problem with the DBWR-proces(ses) on an AIX
4.2.1.  While the redo-logs (4x50Mb) get filled in a couple of minutes
(during which there is high IO on the disk containing them), the
checkpointing (which starts when a redo-log is half-full) takes more than
an hour, resulting, of course, in locking up when all of the redo-logs are
full.

The data-files of the tablespace being loaded via import, are on one
SSA-disk.  This disk also contains the log of the Journaled File System. 
The DBWR(s) show(s) activity of around 70-80.  But iostat (5 sec sample)
shows:
Disks:        % tm_act     Kbps      tps    Kb_read   Kb_wrtn
hdisk3_b         4.0      16.8       4.2          0        84
hdisk3_d         0.0       0.0       0.0          0         0
hdisk3_e         0.0       0.0       0.0          0         0
hdisk3_g         0.0       0.0       0.0          0         0
hdisk4_b         9.4      31.9       8.0          0       160
hdisk4_d         0.0       0.0       0.0          0         0
hdisk4_e         0.0       0.0       0.0          0         0
hdisk4_g         0.0       0.0       0.0          0         0

hdisk3_b and hdisk4_b each contain a copy of the JFS-log, so less than 80
Kb were written to the data-files in 5 seconds.

I started with 4 DBWR-processes with use_async_io = false, then switched to
async_io which resulted in the DBWR doing all the work (sic) and the 4
child-processes idling.  Same lack of performance.

A very funny symptom: I just shut down abort the database, but the dbwr
kept running for a while, never saw that before.  See for yourself:
SVRMGR> shutdown
^CORA-01013: user requested cancel of current operation
Database dismounted.
^CORA-01013: user requested cancel of current operation
SVRMGR> shutdown abort
ORACLE instance shut down.
SVRMGR> host ps -fu orapre | grep MIS
  orapre 22952 25970   5 09:39:38  pts/0  0:00 grep MIS
  orapre 26228 25970   0 09:25:29      -  0:00 oracleMIS
(DESCRIPTION=(LOCAL=YES
)(ADDRESS=(PROTOCOL=beq)))
  orapre 34194     1  69 23:17:19      - 287:26 ora_dbwr_MIS
SVRMGR> exit
Server Manager complete.

Any brightness is, as always, welcome.
-- 
Kind Regards,

Gerard