Re: OS 33 & 221 with REDO LOGS

From: Kyle <kyle_at_gadgets.co.nz.nospam>
Date: Tue, 11 Jun 2002 09:52:39 +1200
Message-ID: <4b9N8.7547$7N.1256983@news02.tsnz.net>

Thanks for that. No, oodles of space.
I removed the duplicate logs and the database still crashed. I have just heard back from Oracle and received the note I have attached. I have rebuilt the three control files and am keeping an eye on it. Interesting how all three seemed "corrupt". This seems to be a problem relating to NT4.0 (piece of junk OS). Trying to convince for a Linux replacement, but it looks like it may be Win2K.

Thanks for your info. Please keep those ideas flowing as Oracle's document was 20 pages of a whole lot of fixes, but no answer to what is actually wrong.

Questions on TAR:
<Note:145452.1> Database Crash with errors in alert.log, ORA-00206, ORA-00202, on metalink. Please check.

Are all control files / database files residing on the local server or across a network?
What type of disk system are you using?

EXTRACTED the notes from metalink below: fact: Oracle Server - Enterprise Edition

     fact: MS Windows NT 4.0
     symptom: Database crashes
     symptom: ORA-00206: error in writing control file (name) block
(num), #
     blocks (num)
     symptom: ORA-00202: control file: (name)
     symptom: ORA-27072: skgfdisp: I/O error
     symptom: OSD-04008: WriteFile() failure, unable to write to file
     symptom: O/S-Error: (OS 33) The process cannot access the file
because
     another process has locked a portion of the file
     symptom: Using multiple control files
     cause: Oracle working with expected behavior. Somehow the CONTROL
FILES got
     corrupted and Oracle could not read the blocks. The problem could
be attributed
     to incomplete writes happening to the controlfiles during the time

of crash.

All the best,
Kyle

"Svend Jensen" <Svend_at_OracleCare.Com> wrote in message news:3D04E7E5.9050700_at_OracleCare.Com...
> Kyle wrote:
>
> > Hi all,
> >
> > I think I have the answer to what is making this instance crash, but the
> > solid why answer escapes me. The instance was crashing with an OS33 and
a
> > bunch of other ora errors I have seen before when a Virus scanner was
> > running. This is not the case now.
> >
> > You'll see a snipet of the log below, but basically the box is as
follows.
> > I have a trace file somewhere....
> >
> > Oracle 8.1.7.0 running atop of NT 4.0 SP6.
> > No hot backups, no virus scanner, no nothing, only Oracle, no DOS copys
> > running either.
> >
> > What I think caused this was:
> >
> > I have four redo log file groups that were not mirrored at 20MB in size.
I
> > came along and did something I have done on dozens of databases before
> > (admitedly, not NT4.0). I added a second member of identical size to
each
> > log group, thus mirroring the logs. I switched through them all and
they
> > all became "In Use".
> >
> > After the crash and the database restarted I could still cycle through
the
> > logs. None of them seemed to be marked Invalid, etc. If a log was
> > corrupt, this would be indicated. As said, I can cycle through them
with
> > "SWITCH LOGFILE". I doubt very much my control files are corrupt as it
> > occurs in various control files. Also, I would expect errors on start
up
> > that one was out of sync. If all of them were out of whack I would also
> > expect an error when I forced a checkpoint. This has not been the case.
> >
> > Also, if an external program were locking the control file, a database
> > bounce would not free it. Oracle must be doing something. Out of the
five
> > or six crashes, the control file has differed.
> >
> > I am about to add the log members again to see if this is the case, but
I
> > would very much appreciate feedback as to if anyone else has had a
similar
> > experience.
> >
> > I am a Unix person and know little about NT, especially NT4.0
> >
> > The log follows,
> > Thanks very much.
> >
> > Wed Jun 05 16:34:51 2002
> > Thread 1 advanced to log sequence 32
> > Current log# 4 seq# 32 mem# 0: E:\ORADATA\AIGFAS\REDO04.LOG
> > Current log# 4 seq# 32 mem# 1: E:\ORADATA\AIGFAS\REDO04B.LOG
> > Wed Jun 05 16:34:51 2002
> > ARC0: Beginning to archive log# 3 seq# 31
> > ARC0: Completed archiving log# 3 seq# 31
> > Wed Jun 05 18:33:26 2002
> > Errors in file e:\admin\aigfas\bdump\aigfasCKPT.TRC:
> > ORA-00206: error in writing (block 3, # blocks 1) of controlfile
> > ORA-00202: controlfile: 'E:\ORADATA\AIGFAS\CONTROL02.ORA'
> > ORA-27072: skgfdisp: I/O error
> > OSD-04008: WriteFile() failure, unable to write to file
> > O/S-Error: (OS 33) The process cannot access the file because
> > another process has locked a portion of the file.
> >
> > Wed Jun 05 18:33:27 2002
> > Errors in file e:\admin\aigfas\bdump\aigfasCKPT.TRC:
> > ORA-00221: error on write to controlfile
> > ORA-00206: error in writing (block 3, # blocks 1) of controlfile
> > ORA-00202: controlfile: 'E:\ORADATA\AIGFAS\CONTROL02.ORA'
> > ORA-27072: skgfdisp: I/O error
> > OSD-04008: WriteFile() failure, unable to write to file
> > O/S-Error: (OS 33) The process cannot access the file because
> > another process has locked a portion of the file.
> >
> > Wed Jun 05 18:33:27 2002
> > CKPT: terminating instance due to error 221
> > Instance terminated by CKPT, pid = 114
> > Dump file e:\admin\aigfas\bdump\aigfasALRT.LOG
> >
> > ____________________________________
> > Kyle Southam
> > Systems Architect
> > www.gadgets.co.nz
> >
> >
> >
> >
>
> Is your E: disk run out of space? Controlfiles extent dynamicly as
> logfile history is added. (max_log_history in create database statement)
> and some init.ora parameters.
>
> rgds
>
> Svend Jensen
>
Received on Mon Jun 10 2002 - 16:52:39 CDT