ORA-600 [kcidr_io_check_common_2] When Using Raw Devices on 11.2.0.2 (문서 ID 1269346.1)

In this Document

Symptoms

Changes

Cause

Solution

References

---

This document is being delivered to you via Oracle Support's Rapid Visibility (RaV) process and therefore has not been subject to an independent technical review.

Applies to:

Oracle Database - Enterprise Edition - Version 11.2.0.2 and later
HP-UX Itanium

***Checked for relevance on 26-Jul-2012***

Symptoms

On HP Itanium, when trying to create a new 11.2.0.2 database or upgrade an existing database to 11.2.0.2 and the database has controlfiles, datafiles, and/or redo log files located on raw devices, we get errors like:

ALTER DATABASE MOUNT
Errors in file
/u02/oracle/app/oracle/diag/rdbms/demo/demo/trace/demo_ckpt_20205.trc:
ORA-00210: cannot open the specified control file
ORA-00202: control file: '/dev/vg11/rcontrol_01.dbf'
...
Errors in file
/u02/oracle/app/oracle/diag/rdbms/demo/demo/trace/demo_m001_20219.trc
(incident=192165):
ORA-00600: internal error code, arguments: [kcidr_io_check_common_2], [1],
[/dev/vg11/rcontrol_01.dbf], [0], [1], [0], [], [], [], [], [], []
ORA-00210: cannot open the specified control file
ORA-00202: control file: '/dev/vg11/rcontrol_01.dbf'
...

Doing a tusc of the startup, eg. with:

$ tusc -fao /tmp/tusc.out sqlplus "/ as sysdba"
SQL> startup mount

we can see permissions errors raised when opening the block device:

open("/dev/vg11/control_01.dbf", O_RDONLY|O_NDELAY|0x800, 060750)
--> ERR#13 EACCES

Note the difference in the file names - rcontrol_01.dbf is the character device whereas control_01.dbf (without the leading "r") is the corresponding block device.

Changes

No changes to the environment, but the 11.2.0.2 database software was just upgraded or installed

Cause

This is the same issue as described in non-published bug:10107681, which was closed as a OS/vendor problem.

From 11.2.0.2 the fix for bug:9956769 is present to add protection to the volumes - from this bug, we have the following:

If you do not want to mount disks used by Oracle ASM on a separate mount
point, then on servers where the system administrator has explicitly granted
to the Oracle Grid Infrastructure installation owner read/write access to any
/dev/rdisk/* files for use by Oracle ASM, you must enable read permissions on
the corresponding block device. You do not need to change the ownership of
the block device.

Solution

Possible solutions are:

1) Relocate database files to filesystem
The problem is only hit when using raw devices (note that the issue is also present when using ASM as it affects eg. disk discovery)

 - or -

2) Explicitly give read permissions on the block devices
This should be done for all database files (controlfiles, datafiles, redo logs) located on raw devices, eg.:

# chmod a+r /dev/vg11/control_01.dbf

- or -

3) Apply HPUX patch PHCO_41479

References

BUG:10252972 - ORA-600 OPENING CONTROLFILE DURING UPGRADE FROM 11.2.0.1.0 TO 11.2.0.2.0 ON RAW
BUG:10331267 - UNABLE TO USE RAW DEVICES WITH 11.2.0.2