Re: Oracle data block corruption....
Date: 29 Jun 1994 23:14:14 GMT
Message-ID: <2usv86$rob_at_pandora.sdsu.edu>
I have also seen this problem in both oracle 6 and 7 and Sun and Hp 7000. I am posting the responses I collected earlier. Perhaps one of the tips will be of use to you.
Greg Tupper
SAIC Comsystems
gtupper_at_nosc.mil
(808) 471 0587
ORA-01578: ORACLE data block corrupted (file # 2, block # 24284) ORA-01110: data file 2: '/u01/ORACLE/tmsdb/master01.dbf' ORA-00600: internal error code, arguments: [3339], [134242020], [134242012],[], [], [], [], []
The above error prompted me to post the the oracle news group asking others if they had similar problems with their database. The results of the mail I received and some of the threads followed. Note the message from Rick Wessman of Oracle saying this should be reported to oracle as a bug. If you have had this problem please send Oracle a note.
Date: Fri, 18 Mar 94 11:39:26 GMT
From: dcrowson_at_amoco.com (David Crowson) To: gtupper_at_monkfish.nosc.mil
Subject: Re: ORACLE data block corrupted survay Status: O
In article crv_at_pandora.sdsu.edu, gtupper_at_monkfish.nosc.mil (Greg Tupper)
writes: -->I am running oracle 7.0.13 on an hp 700 with hpux 9.01 I have --
>had five occations now when I have seen an ORACLE error in the 600 range --
>reporting a corrupt oracle data block while selecting from a table. -->In all
cases I have exported all the data I could, dropped the table, -->and imported
the export file.
-->
-->I have seen this same problem on a 7.0.12 database on an hp700 with 9.00 --
>
-->Is there a bug with the RDBMS code? --> -->Is there a way to recover in a better way (I know... Use the backup) --> -->Does anyone else see this problem?
I've had this problem.. it occurs when there isn't enough SGA to enable oracle to buffer *a lot* of records quickly enough. Increasing the SGA (or fitting quicker disks) fixes this. (We we using a 50Gbytes RAID array which appears to be too slow for Oracle on tables > 15 Million rows). We've had no reoccurence since the SGA was increased. Apparently there is no corruption , it just happens when the transfer of data from disk to memory takes too long and Oracle then gets confused about it and calls it a bad data segment.
A friend also had this problem on tables > 10 Million rows on a mirrored system and it went away when they turned off the mirroring and upped the SGA size
David Crowson =%^) |"For my part, I travel not to go anywhere, but to go. Oracle DBA(Ver.4,5,6,7)| I travel for travel's sake. The great affair is to Amoco Exploration | move, to get down off this featherbed of civilisation Ealing, London, UK. | and to find the globe granite underneath and strewn "My views not Amoex's" | with cutting flints" : Robert Louis Stevenson
Date: Fri, 18 Mar 94 12:38:19 GMT
From: amrattray_at_amoco.com (Sandy Rattray) To: gtupper_at_monkfish.nosc.mil
Subject: Re: ORACLE data block corrupted survay Status: O
Greg -
We've had corrupt data blocks too - Sunos4.1.3& Oracle 6.0.36 - but
we were hook-up to a raid box ; increasing the size of sga seemed
to make the problem go away. (Only seemed to occur on *large* tables
Sandy
- Sandy Rattray (Programmer/Analyst) phone 081-8497554 Amoco (U.K.) Exploration fax 081-8497125 Internet sandy_at_amoco.com X400 C=US A=ATTMAIL P=AMOCO S=RATTRAY G=ALEXANDER I=M
Status: O
Greg:
Unless something really extraordinary is going on like a power hit which left the file system in a bad state, this is a bug which should be reported.
There is no way safe that I know of to recover a file other than reimporting it.
Rick
Rick Wessman
Server Techologies - Network Products
Oracle Corporation
rwessman_at_us.oracle.com
>Path:
oracle!gatekeeper.us.oracle.com!barrnet.net!sgiblab!swrinde!ihnp4.ucsd.edu!new
shub.sdsu.edu!monkfish!gtupper From: gtupper_at_monkfish.nosc.mil (Greg Tupper)
Newsgroups: comp.databases.oracle
>Subject: ORACLE data block corrupted survay Date: 18 Mar 1994 08:04:03 GMT
>Organization: San Diego State University, College of Sciences Lines: 19
>Message-ID: <2mbn9j$crv_at_pandora.sdsu.edu> NNTP-Posting-Host:
monkfish.nosc.mil
>X-Newsreader: TIN [version 1.2 PL2]
>I am running oracle 7.0.13 on an hp 700 with hpux 9.01 I have had five
occations now when I have seen an ORACLE error in the 600 range reporting a
corrupt oracle data block while selecting from a table. In all cases I have
exported all the data I could, dropped the table, and imported the export
file.
>I have seen this same problem on a 7.0.12 database on an hp700 with 9.00
>Is there a bug with the RDBMS code?
>Is there a way to recover in a better way (I know... Use the backup)
>Does anyone else see this problem?
>Thanks
>Greg Tupper
>SAIC Comsystems
>gtupper_at_nosc.mil
Date: Mon, 21 Mar 94 09:39:29 -0500
From: kynug.org!neil_at_ms.uky.edu
To: Greg Tupper <gtupper_at_monkfish.nosc.mil> Subject: Re: ORACLE data block
corrupted survay Sender: kynug.org!neil_at_ms.uky.edu
In comp.databases.oracle article <2mbn9j$crv_at_pandora.sdsu.edu> you wrote: II am running oracle 7.0.13 on an hp 700 with hpux 9.01 I have
>had five occations now when I have seen an ORACLE error in the 600 range
reporting a corrupt oracle data block while selecting from a table. In all
cases I have exported all the data I could, dropped the table, and imported
the export file.
>I have seen this same problem on a 7.0.12 database on an hp700 with 9.00
>Is there a bug with the RDBMS code?
>Is there a way to recover in a better way (I know... Use the backup)
>Does anyone else see this problem?
I am currently having this problem with my 'rulings_data' tablespace in my 'rulings' table. The table complains of a corrupt data block. I have tried exporting the table, but the export complains as well. How did you export the table and its data through export to override this error. Is this a new feature of Oracle7 that allows this? I am running an older version right now, V6.0.30.3.1 under NEXTSTEP. I won't migrate this system over to Oracle7 on our Sparcs until I can correct this error.
tia
-- Neil Greene benchMark Developments, Inc. [NeXT VAR] 2040 Regency Road, Suite C Lexington, KY 40503 Phone: 606-231-6599 / Fax: 606- 254-4864 Email: Neil_at_bMD.com [NeXTmail] Date: Tue, 22 Mar 1994 19:43:05 +0300 (EET) From: "Esa Nurmi, NCS, Tampere, Finland. Tel. +358 31 2407706 Fax. ...666" <ENURMI_at_tnclus.tele.nokia.fi> To: gtupper_at_monkfish.nosc.mil Subject: Re: ORACLE data block corrupted survay X-News: tnclus comp.databases.oracle:11303Received on Thu Jun 30 1994 - 01:14:14 CEST
>From: gtupper_at_monkfish.nosc.mil (Greg Tupper) Subject:ORACLE data block
corrupted survay Date: 18 Mar 1994 08:04:03 GMT
>Message-ID:<2mbn9j$crv_at_pandora.sdsu.edu>
>I am running oracle 7.0.13 on an hp 700 with hpux 9.01 I have had five
occations now when I have seen an ORACLE error in the 600 range reporting a corrupt oracle data block while selecting from a table. In all cases I have exported all the data I could, dropped the table, and imported the export file.
>I have seen this same problem on a 7.0.12 database on an hp700 with 9.00
>Is there a bug with the RDBMS code?
>Is there a way to recover in a better way (I know... Use the backup)
>Does anyone else see this problem?
>Thanks
>Greg Tupper
>SAIC Comsystems
>gtupper_at_nosc.mil
Hello, We are presently having a very similar problem, in a very similar environment - thus we would be VERY grateful, if you could mail us any replies received. Are you using the Oracle call interface, BTW ? We have discovered a call interface related problem, that results in an "600 range" error and MIGHT be the reason for the corruption. We are not yet 100% sure, whether this is the actual reason, because the crash & corruption occurs later ... regards, Esa -- /***************************************************************************** / /* Esa Nurmi * #include <std/disclaimer.h> */ /* enurmi_at_tnclus.tele.nokia.fi * */ /* tel. +358 31 2407 706 * */ /* fax. +358 31 2407 666 * */ /***************************************************************************** / From newshub.sdsu.edu!nic- nac.CSU.net!charnel.net.csuchico.edu!charnel!yeshua.marcam.com!MathWorks.Com!e uropa.eng.gtefsd.com!howland.reston.ans.net!cs.utexas.edu!swrinde!sgiblab!sgig ate.sgi.com!olivea!uunet!frame.frame.com!dlogics.dlogics.com!dlogics.dlogics.c om!nobody Thu Mar 31 16:50:33 1994 Path: newshub.sdsu.edu!nic- nac.CSU.net!charnel.net.csuchico.edu!charnel!yeshua.marcam.com!MathWorks.Com!e uropa.eng.gtefsd.com!howland.reston.ans.net!cs.utexas.edu!swrinde!sgiblab!sgig ate.sgi.com!olivea!uunet!frame.frame.com!dlogics.dlogics.com!dlogics.dlogics.c om!nobody From: dlm_at_dlogics.dlogics.com (Dave Mausner) Newsgroups: comp.databases.oracle Subject: Re: ORACLE data block corrupted survay Followup-To: comp.databases.oracle Date: 24 Mar 1994 11:34:33 -0600 Organization: Frame Technology/Datalogics, Chicago, IL Lines: 22 Sender: dlm_at_dlogics.com Message-ID: <2msiv9$km1_at_dlogics.dlogics.com> References: <2mbn9j$crv_at_pandora.sdsu.edu> NNTP-Posting-Host: dlogics.dlogics.com Summary: corrupted blocks In article <2mbn9j$crv_at_pandora.sdsu.edu>, Greg Tupper <gtupper_at_monkfish.nosc.mil> wrote:
>had five occations now when I have seen an ORACLE error in the 600 range
>reporting a corrupt oracle data block while selecting from a table.
>Is there a bug with the RDBMS code?
>Does anyone else see this problem?
In four months i saw three bad-block hits in databases running 7.0.15.4 under vms 5.5. oracle support is surprisingly unconcerned, taking the attitude that these things happen sometimes. they generally declined to analyse the traces in any great detail mainly because we reloaded tables from backup in order to continue operations. we were able to prove that the bad blocks were not due to faulty media or electronics. i therefore deduce that there is a bug in the rdbms as recently as 7.0.15.4. i have not seen the 7.0.16 release notes to judge whether anything has been corrected in this regard; but i am reluctant to use 7.0.16 because of a number of new pl/sql bugs. -- -- Dave Mausner, Senior Consultant / Datalogics Inc / Chicago IL / 312-266-4450 dlm_at_dlogics.com Motto: Just show me the code From newshub.sdsu.edu!ihnp4.ucsd.edu!usc!yeshua.marcam.com!zip.eecs.umich.edu!newsx fer.itd.umich.edu!nntp.cs.ubc.ca!utcsri!newsflash.concordia.ca!canopus.cc.uman itoba.ca!tribune.usask.ca!kakwa.ucs.ualberta.ca!quartz.ucs.ualberta.ca!alberta !hannigan Thu Mar 31 16:50:33 1994 Path: newshub.sdsu.edu!ihnp4.ucsd.edu!usc!yeshua.marcam.com!zip.eecs.umich.edu!newsx fer.itd.umich.edu!nntp.cs.ubc.ca!utcsri!newsflash.concordia.ca!canopus.cc.uman itoba.ca!tribune.usask.ca!kakwa.ucs.ualberta.ca!quartz.ucs.ualberta.ca!alberta !hannigan From: hannigan_at_cs.ualberta.ca (Barry Hannigan; EdTel) Newsgroups: comp.databases.oracle Subject: Re: ORACLE data block corrupted survay Date: 25 Mar 1994 02:27:56 GMT Organization: Computing Science, U of Alberta, Edmonton, Canada Lines: 29 Message-ID: <2mti7c$1mi_at_scapa.cs.ualberta.ca> References: <2mbn9j$crv_at_pandora.sdsu.edu> <2msiv9$km1_at_dlogics.dlogics.com> NNTP-Posting-Host: scapa.cs.ualberta.ca dlm_at_dlogics.dlogics.com (Dave Mausner) writes:
>In article <2mbn9j$crv_at_pandora.sdsu.edu>,
>Greg Tupper <gtupper_at_monkfish.nosc.mil> wrote:
>>had five occations now when I have seen an ORACLE error in the 600 range
>>reporting a corrupt oracle data block while selecting from a table.
>>Is there a bug with the RDBMS code?
>>Does anyone else see this problem?
>In four months i saw three bad-block hits in databases running 7.0.15.4
three for me too! we are condomizing the system by executing an ANALYZE TABLE .. VALIDATE STRUCTURE on every table in the production system. unfortunately for us, the hit we took was on the SDD_STRUCTURE_ELEMENTS table, whacking the case model. recovery was complicated by a "organizational artifact". ended up synthesizing all possible rowid's for the tablespace and doing a insert into foo_like select * from foo where rowid = MyRowid; for each rowid. admittedly, a bfi approach but this gave up everything but half a block. you get the feeling though that some of the tech support staff wonder about the debugging abilities of those on the left coast. barry -- 4 players | barry hannigan | inferno | 1960-10303 jasper ave | death to videodrome nightmare | edmonton, alberta | long live the new flesh deathmatch on net 666| t5a 4k5 |