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: Migration from HP-UX 10.20 to HP-UX 11 and Oracle 7.3.4 t

RE: Migration from HP-UX 10.20 to HP-UX 11 and Oracle 7.3.4 t

From: Carle, William T (Bill), NBSO <wcarle_at_att.com>
Date: Fri, 18 Aug 2000 13:37:35 -0500
Message-Id: <10593.114999@fatcity.com>


Ron,

    I believe this is the problem you are referring to and it doesn't address the 32 bit/64 bit issue.

Bill

Alert: Database may crash after upgrade to 8.1.6 (Oracle8i Release2)


  This alert describes some essential checks to perform if upgrading   to Oracle 8.1.6 from an earlier release.

  Oracle 8.1.6 is the first server release to perform block checking   for dictionary objects by default. This block checking can detect   problems which went un-noticed in earlier releases of Oracle.

  In particular <Bug:839342> , which affects Oracle releases 7.3 to   8.1.5 inclusive, may have silently introduced a minor corruption in   the data dictionary. If a database with this corruption is upgraded   to 8.1.6 or higher then the new block checking code can notice the   corruption causing the instance to be aborted or the dictionary   block to be marked as permanently corrupt.

Versions of Oracle Affected



  This problem potentially affects all databases upgraded as follows:

    Upgrade path Potentially affected?

  Note, if a database is upgraded from 7.3.4 to 8.1.6 via 8.0.X
(i.e. it has been run under 8.0.X) it is potentially affected by this
  problem.

Platforms Affected



  GENERIC - affects ALL platforms.

Description



  Due to <Bug:839342> the SYS.C_TS# data dictionary cluster can become   silently corrupted in Oracle release 7.3.4 to 8.1.5 inclusive. This   corruption manifests itself as a single byte space 'leak' from a cluster   block. Notably, but not exclusively, this behaviour has been observed   on some databases after the CATREP.SQL script has been run.

  The space leak is benign per se, and on Oracle versions up to (and including) 8.1.5 is likely to go unnoticed as long as the database instance is running WITHOUT any special block checking events.

  Under 8.1.6, block checking is automatically enabled for data objects in the SYSTEM tablespace. If a block in the SYS.C_TS# cluster was previously corrupted then the database may crash after the block is modified in 8.1.6 .

Likelihood of Occurrence



  The problem described here does NOT affect databases created using Oracle8i Release 2 (8.1.6) or higher.

  It can affect databases upgraded from earlier 8.0 or 8.1 releases.   The likelihood of the database crashing as described here is high if the dictionary contains any logically corrupt data blocks. This can be checked prior to upgrade as described in the section "Checking for Corruptions" below.

  Databases migrated DIRECTLY from 7.3.4 to 8.1.6 will not be affected by this problem as the data dictionary objects (including SYS.C_TS#) are rebuilt during migration.

Possible Symptoms



  Possible symptoms of this problem are:

(1) Recursive DML that modifies the datablock raises:

      ORA-00604: error occurred at recursive SQL level 1
      ORA-00607: Internal error occurred while making a change to a 
                  data block
      ORA-00600: internal error code, arguments: 
                  [kcoapl_blkchk] [afn] [block]


(2) Database writer process may terminate with:
ORA-00600: internal error code, arguments: [kcbzpb_2] [afn] [block]

  Where 'afn' is the absolute datafile number, and 'block' is the datablock in that datafile.

  If you see either of these symptoms contact your local support centre for advice on what steps to take.

Checking for corruptions


  BEFORE upgrading to 8.1.6, Oracle Support strongly recommends that you   check the data dictionary objects for possible corruptions. A SQL script to do this can be generated as follows from SQL*PLUS (connected as SYS):
(Spool to an appropriately named file for your platform. This example
uses Unix file naming conventions)

    set head off feedback off pagesize 500 echo off     spool /tmp/analyze_objects.sql
    select 'ANALYZE TABLE "'||table_name||'" VALIDATE STRUCTURE CASCADE;'

      from dba_tables
      where tablespace_name='SYSTEM'
      and owner='SYS';

    select 'ANALYZE CLUSTER "'||cluster_name||'" VALIDATE STRUCTURE CASCADE;'
      from dba_clusters
      where tablespace_name='SYSTEM'
      and owner='SYS';

    spool off

  Because 8.1.6 enables block checking for all SYSTEM tablespace objects, it would be worthwhile running a script that analyzes ALL (SYS and non-SYS owned) objects in this tablespace. Although it is bad practice to use the SYSTEM tablespace for non-SYS objects, it is not an uncommon one. To generate a script that analyzes all SYSTEM tablespace objects, execute the following SQL:

    set head off feedback off pagesize 500 echo off     spool /tmp/analyze_objects.sql
    select 'ANALYZE TABLE "'||owner||'"."'||table_name||

           '" VALIDATE STRUCTURE CASCADE;'
      from dba_tables
      where tablespace_name='SYSTEM';
    select 'ANALYZE CLUSTER "'||owner||'"."'||cluster_name||
           '" VALIDATE STRUCTURE CASCADE;'
      from dba_clusters
      where tablespace_name='SYSTEM';

    spool off

  Edit /tmp/analyze_objects.sql to remove everything but the ANALYZE commands,
  and run the script, connected as a privileged user. If ANALYZE fails for any of the tables or clusters, RAISE A SUPPORT CALL WITH ORACLE SUPPORT BEFORE PROCEEDING WITH THE UPGRADE. If ANALYZE detects a leaked byte in a cluster, it will create a trace file (which should be saved) and fail with the following error:

    ORA-1498: block check failure - see trace file

  AT THE VERY LEAST, THE SYS.C_TS# CLUSTER MUST BE ANALYZED PRIOR TO   UPGRADING TO 8.1.6 using the following command:

    ANALYZE CLUSTER SYS.C_TS# VALIDATE STRUCTURE CASCADE; What should I do I an "ANALYZE .. VALIDATE.." reports an error ?



  If any of the ANALYZE commands reports an error then you should do the

  following:

  1. Check the meaning of the error reported. Eg: "ORA-942 table or view does not exist" may mean a table was dropped between generating the script and running the ANALYZE.
  2. Re-run the ANALYZE for the affected table / cluster to see if the error is repeatable.
  3. For ORA-1499 or ORA-1498 errors there should be a trace file produced in your USER_DUMP_DEST directory. Collect all such trace files together, along with details of the ANALYZE command which fails and the exact error reported.
  4. Contact your local support center with all the details and trace iles collected.

Workaround / Patches



  <Bug:839342> is fixed in 8.1.6.
  Any database with the corruption introduced by this bug should be looked
  at by Oracle Support prior to the database being upgraded to 8.1.6.

Articles


  Corruption detection features in Oracle               <Note:32969.1>
  Parameter reference: DB_BLOCK_CHECKING                <Note:68483.1>


		-----Original Message-----
		From:	Ron Rogers [mailto:RROGERS_at_galottery.org]
		Sent:	Friday, August 18, 2000 2:15 PM
		To:	Multiple recipients of list ORACLE-L
		Subject:	RE: Migration from HP-UX 10.20 to HP-UX 11
and Oracle 7.3.4 t

                If I understand correctly, 8.0.x can be 32/64 bit but migrating to 8.1.x causes data dict corruption id the 8.0.x is 64 bit. The 8.1.x can only be 32 bit.

                 Is this correct? I thought that it is what I was told ay an application users group meeting.

		2bits:
		Ron ROgers
		DBA OCP
		Atl.GA
		>>> lkoivu_at_qode.com 08/18/00 11:44AM >>>
		Dick, you say that 64-bit is faster than 32-bit.  We
recently had to
		investigate whether or not to go 64-bit (we are also HP/UX
11.0) with the
		O/S and the RDBMS.  I searched the net exhaustively trying
to find something
		other than 'you can have an SGA larger than 2GB' for a
reason to choose
		64bit.  I read that 64-bit was only really beneficial to
very large OLTP
		systems (which we aren't, really).  

		Can you substantiate your statement below?  I would be
interested in any
		metrics, white papers, ANYTHING that would explain this in
more detail.  I
		understand the difference between 32-bit and 64-bit, I just
am at a loss to
		understand what exactly this buys me.  

		I was also wary of choosing 64-bit for the RDBMS because we
are HP/UX, the
		second tier for patches, upgrades, etc.  Undesirable
situation to be in, to
		begin with (in my opinion).  I thought if we then went
64-bit RDBMS we would
		have to wait even longer for patches and upgrades.  

		Thanks in advance for your comments. 
		Happy Friday everyone, I'm off to Vegas!!
		Lisa
		Ft. Lauderdale, FL, USA

		-----Original Message-----
		Sent: Friday, August 18, 2000 10:17 AM
		To: Multiple recipients of list ORACLE-L


		Rafiq,

		    It depends on whether your going to be on HP-UX 11 32 or
64 bit.  HP-UX
		10.2
		was strictly 32 bit, but HP-UX 11 gives you a choice,
somewhat dependent on
		the
		hardware.  If your remaining on older HP9000 E,G,K class
machines then your
		stuck with 32 bit.  If your headed for the newer L & N class
then you have a
		choice (BTW: take 64 bit if it's available, it is faster).

		Now, from an Oracle perspective, If your staying with 32 bit
then the
		datafiles
		are compatible & it's just a version upgrade.  If it's 64
bit, then an
		export,
		database rebuild & import will be required.  BUT, since you
state that your
		headed for new hardware (of whatever flavor) in the first
place, this is an
		optimum time for a little house cleaning so take it.  Golden
opportunities
		like
		this are few & far between.

		Dick Goulet

		____________________Reply Separator____________________
		Author: "Mohammad Rafiq" <rafiq9857_at_hotmail.com>
		Date:       8/17/00 12:17 PM

		Hi Gurus

		I need expert guidance for our future move to HP-UX 11 from
HP-UX 10.20.
		Existing setup:
		Platform HP
		HP-UX 10.20
		Oracle 7.3.4.5
		Oracle Financials 10.7(Char) with Oracle Manufacturing/Bill
Of Materials
		Customized forms(2.3) and reports(2.5)


		Future Move:
		Platform HP
		New Machine
		HP-UX 11
		Oracle 8.1.6
		Oracle Financials Release 11i
		Conversion of Customized Forms & Reports to Forms/Reports 6i

		Please advise your experience or opinion to carry out above
migration
		smoothly as Oracle Support guidance is so far contradictory.

		Regards
		Rafiq

	
________________________________________________________________________
		Get Your Private, Free E-mail from MSN Hotmail at
http://www.hotmail.com
		-- 
		Author: Mohammad Rafiq
		  INET: rafiq9857_at_hotmail.com 

		Fat City Network Services    -- (858) 538-5051  FAX: (858)
538-5051
		San Diego, California        -- Public Internet access /
Mailing Lists         

                To REMOVE yourself from this mailing list, send an E-Mail message

                to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in

		the message BODY, include a line containing: UNSUB ORACLE-L
		(or the name of mailing list you want to be removed from).
You may
		also send the HELP command for other information (like
subscribing).
		-- 
		Author: 
		  INET: dgoulet_at_vicr.com 

		Fat City Network Services    -- (858) 538-5051  FAX: (858)
538-5051
		San Diego, California        -- Public Internet access /
Mailing Lists         

                To REMOVE yourself from this mailing list, send an E-Mail message

                to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in

		the message BODY, include a line containing: UNSUB ORACLE-L
		(or the name of mailing list you want to be removed from).
You may
		also send the HELP command for other information (like
subscribing).
		-- 
		Author: Ron Rogers
		  INET: RROGERS_at_galottery.org

		Fat City Network Services    -- (858) 538-5051  FAX: (858)
538-5051
		San Diego, California        -- Public Internet access /
Mailing Lists         
Received on Fri Aug 18 2000 - 13:37:35 CDT

Original text of this message

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