Sunteți pe pagina 1din 5

Problem

Oracle won't start after linking odm library

Error Message
SQL> startup
ORA-24324: service handle not initialized
ORA-01041: internal error. hostdef extension doesn't exist

Solution
1. check odm library link to make sure it is linked correctly.
To link odm library, follow the procedures below.
cd $ORACLE_HOME/product/<version>/lib
mv libodm10.so libodm10.so.old
ln -s /usr/lib/sparcv9/libodm.so libodm10.so
where< version> is Oracle version.

If you are using a CFS shared file system for binaries, perform these tasks from one system. If you
are using local file systems, perform these steps on each cluster system.
Refer to the Storage Foundation for Oracle RAC Installation and Configuration Guide for the loca
tions of the Veritas libraries specific to your version of Oracle.

2. create odm directory in /dev

3. mount /dev/odm by running the command below.


mount -F odm /dev/odm /dev/odm

4. check the contents of /dev/odm once mounted


# ls -l /dev/odm
total 0
crw-rw-rw- 1 root root 291, 3 Sep 19 20:35 cluster
crw-rw-rw- 1 root root 291, 0 Sep 19 20:35 ctl
crw-rw-rw- 1 root root 291, 4 Sep 19 20:35 fid
crw-rw-rw- 1 root root 291, 2 Sep 19 20:35 ktrace
crw-rw-rw- 1 root root 291, 1 Sep 19 20:35 stats

Or you can just run the command below without mounting odm in step 3
/etc/init.d/odm start
This command loads odm module and mounts up /dev/odm.

6. Startup Oracle
Check Oracle alert log to verify that ODM being used is Veritas.
The above odm library link issue would not occur if during Veritas software installation, a Veritas
Storage Foundation for Oracle license is entered
# gabconfig –a (output port d)
Issue:
Getting ODM error upon reboot:
ODM ERROR V-41-1-0-0-1 `/dev/odm' cannot be mounted, as you are not licensed to run ODM
"/sbin/rc2.d/S980odm start" FAILED

Solution:
Either install the SFORA key to enable ODM or disable ODM on reboot if ODM is not being
used:
# mv /sbin/rc2.d/S980odm /sbin/rc2.d/s980odm

Problem
Whilst mounting the Oracle Disk Manager (ODM) device on Solaris 11.2, the mount command
may fail with “Device busy” status if an incorrect license has been configured. ODM device is
usually mounted automatically as part of boot process.

Solution
WORKAROUND:
Apply correct license, remove /dev/odm mount point, recreate it and mount ODM device
mount –F odm /dev/odm /dev/odm

Another work-around is rebooting the system and then mounting /dev/odm.

Applies To
Storage Foundation & High Availability Solutions (SFHA) 6.0.5, 6.1.1 & 6.2
Solaris 11.2 (SPARC)

Readme:
Date: 2008-12-20
OS: HPUX
OS Version:11.31
Packages:VRTSodm GA RP2
Etrack Incidents: 1451946, 1453996

Errors/Problems Fixed:
1451946 : odm_lockpages() can deadlock with multiple pglists
1453996 : odmmkfile/qiomkfile should fail for non-root users

Fixes Applied for Products:


VRTSodm
1. Install and Uninstall Instructions for the ODM patch:
----------------------------------------------------
1.1. To install the ODM 5.0-RP2 patch:
a) The ODM 5.0(GA) must be installed before applying this patch.

b) To verify the Veritas Oracle Disk Manager level, execute:


# swlist -l product | egrep -i 'VRTSodm'
VRTSodm 5.0.31.0 Veritas Oracle Disk Manager

c) All pre/corequisite patches must be installed. The Kernel patch requires a system reboot for bot
h installation and removal.

d) To install the patch, execute the following command:


# swinstall -x autoreboot=true -s PHKL_39130

If the patch is not registered, you can register it using the following command:
# swreg -l depot
The depot is the absolute path where the patch resides.

1.2. To remove the ODM 5.0-RP2 patch:


a) Execute the following command:
# swremove -x autoreboot=true PHKL_39130

IMPORTANT : READ BEFORE INSTALLATION & USE:


1. There is a panic in ASYNC I/O codepaths in HPUX 11.31. See etrack incident 848725. HP prov
ided a patch for ic148. Patch your machine before you install VRTSvxvm package. See Installa
tion steps below for how to put this patch.

2. After installing VRTSvxfs package, copy the following binaries


- fsclustadm, qiomkfile, qioadmin, qiostat, vxfsckd, vxgetmsg
from,
/usr/lbin/fs/vxfs5.0
to,
/usr/sbin
This is fixed after 5.0_11.31-2006-11-15a drop was delivered. Please see etrack incident 820774.
Fix will be part of next drop.

3. After installing VRTSodm package and after every reboot, the ODM module is not getting load
ed automatically.
To check the module state fire :
kcmodule -v odm
To Load the odm module explicitly :
kcmodule -v odm=loaded
This behaviour is due to VRTSvxvm's dmp et.al modules issue. Please see etrack incident 859475
4. After installing VRTSodm package and after every reboot,
/dev/odm is not getting mounted automatically. You need to mount it explicitly using the follow
ing command.
mount -F odm /dev/odm /dev/odm
This behaviour is also due to VRTSvxvm's dmp modules issue. Please see etrack incident 859475

5. umount of a filesystem on which odm files sit may return EBUSY (see etrack incident 856896.
Please use "vxumount -o force " in that case.

Also note that you need to unmount (or force unmount) all VxFS file-systems
on which odm files are sitting, before going to unmount /dev/odm.

6. There is a cache-coherency issue in FDD (a module in VRTSvxfs), due to which you may not
see the data written through ODM files in clones or shared-mmaps.
(See etrack incidents 843809 and 857983)

INSTALLATION STEPS:
- Copy the vm_vasm patch to /usr/conf/mod/ to a
freshly installed ic148 perf machine
- Install VRTSvlic
- Install VRTSvxvm
- Install VRTSvxfs
- Install VRTSodm
Known Issues
============
Smartsync Option not supported for ODM 5.0 on HP-UX 11iv3

The Beta releases of SFORA 5.0 and SFRAC 5.0 on HP-UX 11iv3 will not support ODM (Oracle
Disk Manager) to be used in the smartsync mode of operation. ODM will be mounted in the
'nosmartsync' mode by default.
Using ODM in the smartsync mode of operation might lead to unexpected behavior for the appli
cations.

Problem
ORA-17500: ODM err:ODM ERROR V-41-4-2-199-28 No space left on device

Solution
The issue:
This error can occur when trying to extend an Oracle ODM file if the file is a Quick I/O file that
was converted to an ODM file.
A converted Quick I/O file can have a space reservation set on the file that will prevent extending
the file to a size larger than the reservation.

To resolve the issue:


Use the following example procedure to check and remove a space reservation.
1. Check for a space reservation on the file using the getext command, for example:
# /opt/VRTS/sbin/getext test
test: Bsize 1024 Reserve 1024 Extent Size 0
In the above example, the file "test" has a space reservation (Reserve) of 1024 file system blocks.
(Note: on older versions of VERITAS File System the command path might be /usr/lib/fs/bin
/getext.)

2. If the ODM file in question has a space reservation, use the setext command to clear the reser
vation:
# /opt/VRTS/sbin/setext -r 0 test
(Note: on older versions of VERITAS File System the command path might be /usr/lib/fs/vxfs/bin
/setext.)

3. Verify the removal of the space reservation using getext.


# /opt/VRTS/sbin/getext test
test: Bsize 1024 Reserve 0 Extent Size 0
Note that Reserve is now zero and the space reservation is gone.

S-ar putea să vă placă și