cancel
Showing results for 
Search instead for 
Did you mean: 

ORA-63999 with crash of instance in 11.2.0.2

volker_borowski2
Active Contributor
0 Kudos

Hi,

had some 15 minutes playtime today and tried one of my favorite tests on new upgraded 11g DB:

Delete and recover a datafile while instance keeps running.

Well that was a short drift this time, because after i deleted my testfile and forced the checkpoint,

the instance did crash instantly. Is this new expected behavior in 11.2 ?


Wed Apr 13 14:51:47 2011
Archived Log entry 1528 added for thread 1 sequence 1906 ID 0x181062ce dest 1:
Wed Apr 13 14:52:02 2011
Beginning global checkpoint up to RBA [0x773.12a.10], SCN: 38535020
Errors in file /oracle/TRN/saptrace/diag/rdbms/TRN/TRN/trace/TRN_ckpt_14942332.trc:
ORA-63999: data file suffered media failure
ORA-01116: error in opening database file 5
ORA-01110: data file 5: '/oracle/TRN/sapdata1/TRNusr_1/TRNusr.data1'
ORA-27041: unable to open file
IBM AIX RISC System/6000 Error: 2: No such file or directory
Additional information: 3
Additional information: 4
Additional information: 4194432
Errors in file /oracle/TRN/saptrace/diag/rdbms/TRN/TRN/trace/TRN_ckpt_14942332.trc:
ORA-63999: data file suffered media failure
ORA-01116: error in opening database file 5
ORA-01110: data file 5: '/oracle/TRN/sapdata1/TRNusr_1/TRNusr.data1'
ORA-27041: unable to open file
IBM AIX RISC System/6000 Error: 2: No such file or directory
Additional information: 3
Additional information: 4
Additional information: 4194432

>>> CKPT (ospid: 14942332): terminating the instance due to error 63999

Wed Apr 13 14:52:02 2011
System state dump requested by (instance=1, osid=14942332 (CKPT)), summary=[abnormal instance termination].
System State dumped to trace file /oracle/TRN/saptrace/diag/rdbms/TRN/TRN/trace/TRN_diag_9699420.trc
Dumping diagnostic data in directory=[cdmp_20110413145202], requested by (instance=1, osid=14942332 (CKPT)), summary=[abnormal instance termination].
Instance terminated by CKPT, pid = 14942332

Wed Apr 13 14:54:08 2011
Starting ORACLE instance (normal)

While the usual errors ORA-01116 and ORA-01110 which did result in a "datafile offline" in earlier releases

this ORA-63999 seems to be new. At least I did not face it before.

Has anybody who has MCOD installations already played around with that.

I think this is definately wrong behavior, that the instance goes completely down for loss of a single datafile.

Correct behavior should be to set the file offline and keep the instance up and open.

Comments ? Thoughts ?

Please restrain from suggestions how to recover this system. I am completely aware about that.

This thread is for discussing the side effect of the crash described.

Volker

Accepted Solutions (0)

Answers (2)

Answers (2)

volker_borowski2
Active Contributor
0 Kudos

Thank you both for feedback.

Looks indeed like this one will be fixed soon.

As the SAP-metalink account is currently no more accessible,

this information is of great help.

Volker

stefan_koehler
Active Contributor
0 Kudos

Hello Volker,

i guess you misunderstood. The bug #7691270 is already fixed in 11.2.0.2 and the new behaviour is the wanted one.

So after you have applied patchset 11.2.0.2 - it is expected, that the instance crashes in case of some I/O error. You can control the new behaviour with the parameter "_datafile_write_errors_crash_instance".

So if you want the old one - just set the parameter to FALSE.

Regards

Stefan

volker_borowski2
Active Contributor
0 Kudos

Ooops,

ideed I misunderstood this. I was not able to read the ML article, as the

ML User from SAPnet does not work any more.

Looks like Oracle does not like this type of shared user accounts and has disabled it.

Thanks for clearification.

Volker

stefan_koehler
Active Contributor
0 Kudos

Hello Volker,

the instance did crash instantly. Is this new expected behavior in 11.2 ?

I read a blog about that topic some time ago, but i am not able to find it anymore. Yes this behaviour is "new", but i am not quite sure when it was introduced (in 10.2.0.5, 11.1.x or 11.2.x).

I also can remember that this "new" behaviour can be controlled by an event or an underscore parameter (not quite sure anymore).

In the past the instance only crashed if a system data file was missing ... but it has changed.

Regards

Stefan

0 Kudos

May be because of this (11.2.0.2):

Bug 7691270 Crash the DB in case of write errors (rather than just offline files)

> If datafilewrite_errors_crash_instance = TRUE (default) then any write to a datafile which fails due to an IO error causes an instance crash.

stefan_koehler
Active Contributor
0 Kudos

Hello Fidel,

thansk for the completion - with the name of the hidden parameter i was able to find the blog in my RSS feed reader

http://oraclue.com/2010/10/21/oracle-11-2-0-2-patch-set-new-feature-and-change/

http://oradbastuff.blogspot.com/2010/09/11202-11g-release-2-patch-set-2.html

Regards

Stefan

0 Kudos

Thanks,

I also read it recently and I could not find it.

After looking a little I found my initial source (last one was MOS):

http://blogs.oracle.com/UPGRADE/2011/03/new_hidden_parameters_in_oracl.html