Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> dbverify, datafile corruption, Oracle 9i R2
Oracle 9.2, Suse 8.0, physical RAM 1GB, 700M for Oracle, 1GB swap space
and no other RAM consuming processes on this machine.
Dear all,
in my current project I have to set up Oracle 9.2 EE on Suse 8.0 on an
ext3 system. Please bear with me, as none of the three was my choice.
Installation of the software on a brand new installed suse 8.0 was
sucessful, install log reported no errors. I created the database
manually via sql statements. All went fine and yesterday I trashed the
database with inserts to verify my initial space requirements. Oracle
reported no errors, but tomorrow morning alert.log stated corrupt
datablocks in the data TS (inserts were finished 10 hours ago).
After that alert I run dbverify against all the datafiles and the
utility found in 3 of them corrupt blocks.
The datafile size ranges from 200M - 3 GB (for BLOBS), total size of db
around 10GB.
After that I run e2fsck with badblock option, e2fsck found no errors on
the partitions.
I decided to start from scratch, deleted the database and created it for
the second time. After each adding of a tablespace (or datafile) I run
dbverify against the new created files. Dbverify again reported errors
(this time only 3 blocks corrupt, the first time about 15).
A google search revealed that other some users with Suse 8.0 and Oracle
9.0.1 had problems with dbverify, but the underlying reason was not
pointed out. There was only a suggestion that dbverirfy might be broken.
Has anybody encountered similar problems?
The utilites manual
http://otn.oracle.com/docs/products/oracle9i/doc_library/release2/server.920/a96652/ch13.htm#SUTIL013
states that
<snip>
Total Pages Marked Corrupt = number of blocks for which the cache header
is invalid, thereby making it impossible for DBVERIFY to identify the
block type
</snip>
which I interpret as 'dbverify notices that there is something wromg in
the datablock, but does not mark the datablock (mark it not as unusable
for Oracle, so Oracle never writes to this block)'. Am I correct with
this interpretation?
I'm aware of dbms_repair package, but the frequence and the occurence of
data block corruption shortly after the installation makes me uneasy. I
have never encountered data block corruption with that frequency on
various 8i installations I performed.
Any suggestions are welcome
Thanks and have a nice day
Manuela Mueller
Received on Wed Jun 12 2002 - 13:25:16 CDT