SAP/Oracle/AIX problems with async i/o and no mwc

From: Dennis Jarvis <jarvis_at_apple.com>
Date: Wed, 12 May 1999 14:19:32 -0700
Message-ID: <3739EFF2.87E5A060@apple.com>

Hi,

In the continuing saga of hosting SAP and Oracle on AIX:

We have experienced two corruptions in the last month which we and/or IBM are as of yet unable to attribute any cause to.

We never had a corruption prior to enabling async i/o and disabling mirror write consistency for performance sake, so there is a tendency among some to say that they are the cause. No one (neither customer nor vendors) however, has been able to support such a statement with hard evidence. Following a system crash we updated AIX to 4.3.2.4 with the latest async i/o patch from IBM, and the latest SSA adapter microcode and device drivers. We are at Oracle 7.3.4.2 with no plans to go to 8.x anytime soon. Our SAP version is 3.1H, with the 3.1I kernel, patch 244.

Does anyone have any experience with corruptions on the latest AIX as a result of one or more of these parameters?

Here is an extract of the DBV we ran for the latest corruption. Consists of +50,000 lines.

dbv file=[filename and path removed for security purposes] blocksize=8192

DBVERIFY: Release 7.3.4.0.0 - Production on Sat May 8 00:07:57 1999

DBVERIFY - Verification starting : FILE = [filename and path removed for security purposes]
Page 96426 is marked software corrupt
Page 96433 is marked software corrupt
Page 96434 is marked software corrupt

and the user trace...

Dump file [filename and path removed for security purposes].trc Oracle7 Server Release 7.3.4.3.0 - Production With the distributed, replication and parallel query options PL/SQL Release 2.3.4.3.0 - Production
ORACLE_HOME = [path removed for security purposes] System name: AIX

Node name:      [hostname removed for security purposes]
Release:        3
Version:        4
Machine:        000912184C00

Instance name: [SID removed for security purposes] Redo thread mounted by this instance: 1 Oracle process number: 67
Unix process pid: 108152, image:

1999.05.07.23.00.01.000
SESSION ID:(66.15) 1999.05.07.23.00.01.000
***
Corrupt block dba: 0x4d0178aa file=77. blocknum=96426. found during buffer read on disk type:6. ver:1. dba: 0x4d0178aa inc:0x00000020 seq:0x00002cbd incseq:0x00202cd0

Reread of block=4d0178aa file=77. blocknum=96426. found same corrupted data
*** 1999.05.07.23.23.42.000
***

Corrupt block dba: 0x4d0178aa file=77. blocknum=96426. found during buffer read on disk type:6. ver:1. dba: 0x4d0178aa inc:0x00000020 seq:0x00002cbd incseq:0x00202cd0

Reread of block=4d0178aa file=77. blocknum=96426. found same corrupted data

Dennis

Received on Wed May 12 1999 - 16:19:32 CDT