How To clean up after a Failed Oracle Clusterware

主题:	Real Application Clusters 10g CRS Install on Windows Is Successful But Configuration Assistants Fail
	:	注释:310791.1	类型:	PROBLEM
	上次修订日期:	06-JUN-2006	状态:	PUBLISHED

In this Document
  
  
  
  

---

Applies to:

Oracle Server - Enterprise Edition - Version: 10.1.0.0 to 10.1.0.4
Microsoft Windows 2000
Microsoft Windows Server 2003 (64-bit Itanium)
Microsoft Windows Server 2003
Cluster configuration assistants that run as part of the CRS install of 10g Real Application Clusters on Windows are failing following a successful install of the CRS binaries.

Symptoms

IT HAS RECENTLY COME TO THE ATTENTION OF ORACLE SUPPORT THAT ONE REASON THESE ASSISTANTS MAY FAIL IS IF THE INSTALLATION IS DONE USING TERMINAL SERVICES.

TO AVOID THAT ISSUE SPECIFICALLY, BE SURE TO USE EITHER VNC OR TERMINAL SERVICES IN CONSOLE MODE TO PERFORM THIS INSTALLATION.  YOU CAN INVOKE TERMINAL SERVICES IN CONSOLE MODE WITH THE FOLLOWING COMMAND:

mstsc -v:servername /F –console

======================================================================================

The installActions.log is showing errors such as:

PRIF-10: failed to initialize the cluster registry

Configuration assistant "Oracle Private Interconnect Configuration Assistant" failed


Step 4: Starting up CRS stack on all nodes
node1 service OracleCSService in improper STOPPED state, err(997)
node2 service OracleCSService in improper STOPPED state, err(997)

OR

Step 4: Starting up CRS stack on all nodes
node1 service OracleCSService in improper STOPPED state, err(0)
node2 service OracleCSService in improper STOPPED state, err(997)

OR

Step 4: Starting up CRS stack on all nodes
oracletest1 failed to startup service OracleEVMService, err(1053)

OR

Step 1:  checking status of CRS cluster
Step 2:  configuring OCR repository
ignoring upgrade failure of ocr(-1073740972)
failed to configure Oracle Cluster Registry with CLSCFG, ret -1073740972
after installing the patch 3555863 the configuration failes with the same
problem

Changes

***

Solution

Section I: Common Causes and Corresponding Solutions:

A. The OCFS format is not recognized on one or more of the remote cluster nodes

If you are using OCFS for your OCR and Voting disk partitions, then you will probably need to:

1. Leave the OUI window in place
2. Reboot the 2nd node (& any additional nodes)
3. Retry the assistants

OR

B. There is a timing issue with start of the OracleCSService:

1. Leave the OUI window in place
2. Try starting the OracleCSService manually on all nodes (you may also need to manually start the OracleObjectService)
3. Retry the assistants

OR

C. You are on 2003 box and Automount of new drives is not enabled:

For RAC on Windows 2003, you are required to issue the following commands on all nodes:

diskpart
automount enable

Note: Please refer to Note 254611.1 "Shared Partitions Errors in RAC Configuration on Windows 2003" for more information
on running these commands.

If you are already failing at the configuration assistants and this has not yet been run on all nodes in the cluster, you will need to clean up your CRS install, issue this command on all nodes, reboot all nodes, and start the CRS install again.

OR

D. You have entered a ' . ' in one of the node names during CRS install.

We cannot accept a ' . ' in a node name, if you have done so, we suggest changing this to an ' _ ' or ' - '

Clean up and reinstall CRS

OR

E. Node1 failed to startup service OracleEVMService, err(1053)

This is a special case ...

Starting the EVM service on Node 1 failed because we couldn't find dll: MSVCP60.DLL
This was in two places on Node 2 (including c:\winnt\system32\) butonly in one place on Node 1 ... in the OpenManage software tree.
Copying MSVCP60.DLL to c:\winnt\system32 should allow EVM to start.

However, a reinstall of CRS is still needed.

OR

F. 

Step 1:  checking status of CRS cluster
Step 2:  configuring OCR repository
ignoring upgrade failure of ocr(-1073740972)
failed to configure Oracle Cluster Registry with CLSCFG, ret -1073740972
after installing the patch 3555863 the configuration failes with the same
problem

You must DIRECTLY grant Administrator group to the user under which you are logged in to the servers in this cluster.  All clusternodes should be logged in as the same user.  If you assign this user to a group, then grant the group Administrator privileges, this is not enough.  The Administrator privilege must be granted explicitly and directly to the user.

Specifically for the case where you encounter these errors:

Step 4: Starting up CRS stack on all nodes
node1 service OracleCSService in improper STOPPED state, err(0)
node2 service OracleCSService in improper STOPPED state, err(997)

A check in the ocssd.log file:  located in $CRS_HOME\crs\log
shows you messages that make reference to a fully qualified private hostname
The way to fix this will be to add an alias to the HOSTS file on each node that allows resolution of the Private IP
whether or not the hostname is fully qualified ... as such:

  node1-priv  node1-priv.us.oracle.com

*****

In all of the above cases, clean up the failed CRS install following instructions in companion

NOTE:341214.1 - How To Remove and Reinstall Oracle 10gR2 Clusterware Install on Windows

Note:

Although these cluster configuration assistants are labelled 'optional' and the installActions.log does a good job of showing how to run these assistants manually, Oracle Support strongly recommends that a clean CRS install be performed before proceeding with the rest of your Real Application Clusters configuration on Windows.

References

- Shared Partition Errors in RAC Configuration on Windows 2003
- How To Remove and Reinstall Oracle 10gR2 Clusterware Install on Windows

Keywords

''   ''   ''   ''   ''   ''   ''   ''