RE: Strange problem

From: Mark W. Farnham <mwf_at_rsiz.com>
Date: Sun, 30 Apr 2017 23:35:46 -0400
Message-ID: <04be01d2c22c$05992d80$10cb8880$_at_rsiz.com>



Two cases I have seen too old for me to remember the exact error numbers, but with similar symptoms working on restart and then intermittently not working:  

  1. A file that has autoextended past the addressable length and an extent past the addressable length is trying to be flushed by dbwr. But I *thought* they fixed that circa V10 so that even if the beginning of a file’s next autoextend was within range but the size of the extension would put it past the addressable size it would either refuse to extend to only extend to the maximum addressable length. You might be able to decode the data block addresses to see.
  2. A particular type of removable drive unfortunately had pegs sticking into a narrow aisle such that someone walking down the aisle could bump them and put a that drive into a read only state at the hardware level.

#1 was slimy because the file would read at start time (apparently the addressing for the tail block was in a bigger address space than dbwr used at the time.) IF memory serves we were lucky and the extent was an index, so we were able to drop that index and fill the tablespace with dummy tables until we were sure all the space not actually writable would never be written until we had time to move everything out of that table space.
 

#2 would have been hilarious if it had not been so difficult to diagnose, and resulted in a customer added cardboard flange protecting the buttons.
 

It’s probably something completely different, but I think it IS likely that dbw0 is somehow losing write authority. IF this were RAC a wild guess would be instances started by different owners and the one trying to write doesn’t have the right owner and/or group authority.  

mwf  

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Mladen Gogala Sent: Sunday, April 30, 2017 4:22 PM
To: oracle-l
Subject: Strange problem  

Hi!

I am having a strange problem with an Oracle 12.1.0.2 instance on SUN Solaris 11.3 (SPARC):

ORA-01110: data file 7: '/oradata/DEV/datafiles/data01b.dbf' 2017-04-28 23:05:11.723000 -04:00
Flush retried for xcb 0x5c0e98a90, pmd 0x5c3faa538 Errors in file /app/oradev/diag/rdbms/oradev/DEV/trace/DEV_pmon_465049.trc: ORA-00372: file 7 cannot be modified at this time ORA-01110: data file 7: '/oradata/DEV/datafiles/data01b.dbf'
2017-04-28 23:05:18.756000 -04:00
Flush retried for xcb 0x5c0e98a90, pmd 0x5c3faa538 Errors in file /app/oradev/diag/rdbms/oradev/DEV/trace/DEV_dbw0_465083.trc: ORA-00372: file 7 cannot be modified at this time ORA-01110: data file 7: '/oradata/DEV/datafiles/data01b.dbf' USER (ospid: 465083): terminating the instance due to error 372 System state dump requested by (instance=1, osid=4295432379 (DBW0)), summary=[abnormal instance termination]. System State dumped to trace file /app/oradev/diag/rdbms/oradev/DEV/trace/DEV_diag_465073_20170428230518.trc 2017-04-28 23:05:23.820000 -04:00
Instance terminated by USER, pid = 465083 2017-04-29 10:58:13.293000 -04:00
Starting ORACLE instance (normal) (OS id: 44563) CLI notifier numLatches:29 maxDescs:1355 All SGA segments were allocated at startup

When I restart the instance all files are available. The instance is patched with the January 2017 PSU, soon to be patched with April 2017 PSU. Here are the relevant versions:

SunOS sun02 5.11 11.3 sun4v sparc sun4v  

-bash-4.4$ sqlplus / as sysdba  

SQL*Plus: Release 12.1.0.2.0 Production on Sun Apr 30 16:15:43 2017  

Copyright (c) 1982, 2014, Oracle. All rights reserved.    

Connected to:
Oracle Database 12c Enterprise Edition Release 12.1.0.2.0 - 64bit Production With the Partitioning, OLAP, Advanced Analytics and Real Application Testing options

The underlying file system is, of course, ZFS. When the instance is restarted, all the files are available and there is nothing in the V$RECOVER_FILE. I opened SR with Oracle, but as this is a development system, they are taking their time. Has anyone seen this before and, if yes, what would be the solution? The system crashes during nightly batches, which load a lot of data and are write intensive.        

-- 
Mladen Gogala
Oracle DBA
Tel: (347) 321-1217


--
http://www.freelists.org/webpage/oracle-l
Received on Mon May 01 2017 - 05:35:46 CEST

Original text of this message