Oracle® Fail Safe

3.1 Oracle Fail Safe Compatibility When Installed on Windows NT

This release of Oracle Fail Safe is compatible with the software shown in the following table when Oracle Services for MSCS is installed on Microsoft Windows NT Enterprise Edition version 4.0 (Service Pack 5 or 6a), and Oracle Fail Safe Manager is installed on one of the following operating systems:

• Microsoft Windows NT version 4.0 (Service Pack 5 or 6a)

• Microsoft Windows 98

• Microsoft Windows 2000

• Microsoft Windows XP

3.2 Oracle Fail Safe Compatibility When Installed on Windows 2000

This release of Oracle Fail Safe is compatible with the software shown in the following table when Oracle Services for MSCS is installed on Microsoft Windows 2000 Advanced Server or Datacenter Server, and Oracle Fail Safe Manager is installed on one of the following:

• Microsoft Windows NT version 4.0 (Service Pack 5 or 6a)

• Microsoft Windows 98

• Microsoft 2000

• Microsoft Windows XP

5.1 Reboot Twice After Upgrading from Oracle Fail Safe Release 2.1.3 to Oracle Fail Safe Release 3.3.1

After a new installation is complete, Oracle Universal Installer will instruct you to reboot the system. This reboot will update the system path to include the Oracle Fail Safe release 3.3.1 directory.

If you are upgrading from Oracle Fail Safe release 2.1.3, then after the first reboot, run Verify Cluster in Oracle Fail Safe Manager to update the Oracle Fail Safe database resource DLL. Then reboot each cluster node a second time, so the cluster services can pick up the new resource DLL.

5.2 MSCS Cluster Administrator Displays Problems with Fail-Safe Resource Types

Sometimes, after completing an Oracle Fail Safe installation, you see problems with the fail-safe resource types (such as databases) in MSCS Cluster Administrator. MSCS Cluster Administrator denotes the problem by displaying an Oslash symbol (Ø) over the resource type name.

If this occurs, do the following:

1. If you forgot to reboot the cluster nodes after installing Oracle Fail Safe, do so now.

2. Make sure that the PATH environment variable includes the Oracle Services for MSCS path. (In the Command Prompt window, enter PATH. The Oracle Services for MSCS path (<Oracle_Home>\fs\fssvr\bin) should be included. If it is not included, add it, and then reboot the nodes on which the Oracle Services for MSCS path is missing.

3. Make sure that the Oracle Fail Safe resource DLL, FsResOdbs.dll, is installed in <Oracle_Home>\fs\fssvr\bin.

   If the resource DLL is not there, reinstall Oracle Fail Safe.

4. Use Oracle Fail Safe Manager to verify the cluster (on the Troubleshooting menu, select Verify Cluster), then reboot each cluster node, one at a time. The Verify Cluster command automatically verifies registration of Oracle resource DLLs. You may not need to reboot all cluster nodes. After you reboot one node, check MSCS Cluster Administrator to see if the Oslash symbol has been removed from the resource type names. If the Oslash symbol is gone, you do not need to reboot all cluster nodes.

5.3 Silent Installation Is Case-Sensitive

The Oracle Fail Safe silent installation is case-sensitive. You must enter all command-line syntax exactly as shown in the Oracle Fail Safe Installation Guide (for example, the -responseFile parameter must be entered using all lowercase characters, except for the F, as shown). You must enter the file specification using the case shown in Windows Explorer (for example, C:\Ofs\Silent_Install\OfsProducts.rsp).

5.4 Do Not Run Windows Event Viewer During Oracle Services for MSCS Installation

If you attempt to install Oracle Services for MSCS while the Windows Event Viewer is running, the installation procedure may return an error indicating that it cannot copy the fsus.dll file (fsja.dll on Japanese systems) and asking if you want to retry, ignore, or cancel. You can continue the installation if you stop the Windows Event Viewer and then click Retry in the Oracle Fail Safe installation Error window.

6.1 DCOM Security Changes on Windows 2000

If the list of Access permissions for the default setting of DCOM is empty, the SYSTEM and INTERACTIVE accounts have implicit access rights. In Windows 2000, if a user or group is added to the list, the implicit rights no longer apply; rights are granted only to explicitly named users or groups. In Windows NT, the SYSTEM and INTERACTIVE accounts retain their implicit rights.

During the installation of Oracle Services for MSCS, the Oracle Fail Safe user account is added to the default access permissions list. If the access list is empty, the SYSTEM account is automatically added, but not the INTERACTIVE account.

After installing Oracle Services for MSCS, if there are problems opening hyperlinks in an HTML file or problems using Microsoft Outlook or Outlook Express, use the dcomcnfg.exe tool to add the necessary users to the default access permissions list.

See Microsoft Support article Q274696 for more information.

6.2 Oracle Services for MSCS and Microsoft Cluster Server Can Run Under Different Accounts

When your cluster was first configured and Microsoft Cluster Server was first installed, it was installed under a Windows user account.

Oracle Services for MSCS runs as a Windows service under a domain user account (not the system account) that has Administrator privileges on all cluster nodes. When you install Oracle Services for MSCS, you are instructed to provide a user name and password combination for a user account that has the required privileges. This account does not have to be the same account under which Microsoft Cluster Server was installed.

Is Alive polling of Oracle Fail Safe resources is performed using the Windows account that Microsoft Cluster Server is running under, not the account under which Oracle Services for MSCS is running.

7.1 Oracle Fail Safe Manager Requires a User Name and Password When Run on Windows 2000 or Windows XP

You must supply a user name and password when you log on to Oracle Fail Safe Manager running on a Windows 2000 or Windows XP system.

If you log on to Oracle Fail Safe Manager running on any other operating system, you are not required to specify the user name and password if your user account has appropriate privileges.

9.1 Default Oracle Intelligent Agent Is Stopped and Restarted When Database Is Shut Down

The default Oracle Intelligent Agent incorrectly discovers fail-safe databases on the node where the default Oracle Intelligent Agent is running and maintains a connection to the database. (The default Oracle Intelligent Agent listens on the node's host address and therefore should not discover fail-safe databases because they use a virtual address.) Therefore, when a fail-safe database is taken offline in normal or transactional mode using Oracle Fail Safe Manager, Oracle Fail Safe shuts down the default Oracle Intelligent Agent prior to shutting down the database. Oracle Fail Safe restarts the default Oracle Intelligent Agent after the database shutdown operation is complete.

9.2 Create Sample Database

Oracle Fail Safe includes a Create Sample Database command that installs a preconfigured sample database on a cluster disk specified by the user. The sample database has limited functionality and is intended only for testing purposes and for use with the online Oracle Fail Safe Tutorial; it should not be used for production.

The following table shows the releases of the sample databases included with Oracle Fail Safe release 3.3.1, and the releases of the Oracle database that can be tested with the sample database.

The sample database files can be installed only on a node where the appropriate release of the Oracle database server is installed. For example, if Oracle database server release 8.1.7 is installed, then only the sample database files for 8.1.5 will be installed during the Oracle Fail Safe installation.

There are known limitations in using the sample databases with releases other than the release for which they were originally created. When an 8.1.5 sample database is installed for use with Oracle database server release 8.1.7, the following errors are returned when attaching to the database:

ERROR:
ORA-06533: PLS-908: The stored format of SYS.STANDARD is not supported by this release.

Error accessing packing DBMS_APPLICATION_INFO
ERROR:
ORA-06533: PLS-908: The stored format of SYS.STANDARD is not supported by this release.

The sample database feature provided by Oracle Fail Safe is to be used for testing only; to create a database for production, use the Oracle Database Configuration Assistant or create the database manually.

9.3 User Name for Database Must Be SYS

To ensure the success of all Oracle Fail Safe release 3.2.1 and release 3.3.1 operations, the database user name must be "SYS" unless operating system authentication is used. If operating system authentication is used, then Oracle Fail Safe does not use the SYS account.

10.1 Add Oracle HTTP Server to Group Operation Fails

If an Add Oracle HTTP Server to Group operation for Oracle release 8.1.7 fails due to a problem with the default httpd.conf file, the following errors are returned during the clusterwide operation:

FS-10382: NTCLU-221 : Bringing resource <Oracle-HTTP-Server-name> online
** ERROR : FS-10726: Resource <Oracle-HTTP-Server-name> is in a failed state
** ERROR : FS-10012: Failed to bring the cluster resource <Oracle-HTTP-Server-name> online. Check the Windows NT event log for messages
** ERROR : FS-10782: The Oracle HTTP Server resource provider failed to bring resource <Oracle-HTTP-Server-name> online

** ERROR : FS-10497: Starting clusterwide rollback of the operation

To fix this problem, open the default httpd.conf file located in <Oracle_Home>/apache/apache/conf/httpd.conf and insert a comment character (#) in front of the commands that include the configuration for Apache JServ 1.1 and the Oracle configuration file for custom settings. After you insert the comment characters, the httpd.conf file will include lines similar to the following:

# Include the configuration for Apache JServ 1.1

# include "C:\Oracle\Ora81\Apache\Jserv\conf\jserv.conf"

# Include the Oracle configuration file for custom settings

# include "C:\Oracle\Ora81\Apache\Apache\conf\oracle_apache.conf"

11.1 Inconsistent NAMES.DEFAULT_DOMAIN Parameter Might Cause Oracle Reports Server to Fail

The value for the NAMES.DEFAULT_DOMAIN parameter in the sqlnet.ora file must be the same across all Oracle homes in the cluster. If it is not, you may experience problems with Oracle Reports Server resources. For example, an Add Oracle Reports Server to Group operation might fail or a Verify Group operation might fail if the group contains an Oracle Reports Server.

11.2 Incorrect Message When You Verify a Group Containing an Oracle Reports Server

When you verify a group that contains an Oracle Reports Server, the following message may be displayed when the service name for the Oracle Reports Server exists and is correct:

FS-10693: Oracle Reports Server <report-name> does not exist in <Oracle_Home>\Network\Admin\Tnsnames.ora on node <node-name>.
Do you want to create it with default port 9100?

This message is displayed each time you verify the group that contains the Oracle Reports Server regardless of how you respond to the error message query.

13.1 Validate the formsweb.cfg file After Adding an Oracle Forms Load Balancer Server Resource

After adding an Oracle Forms Load Balancer Server resource to a group, validate the MetricServerPort and ServerPort parameters in the formsweb.cfg file, which is used by the Forms CGI executable file. The file is located at <Oracle_Home>\Forms60\formsweb.cfg. The MetricServerPort parameter that you specified should have the same value as the RequestPort startup parameter specified when you added the Oracle Forms Load Balancer Server resource. The ServerPort parameter should have the same value as the Port startup parameter for Oracle Forms Server.

14.1 Add Oracle MTS to Group Operation Fails

If the Add Oracle MTS Service to Group operation fails, it may be due to a problem with the value for LAST_HOME in the Windows registry. If you attempt to add an Oracle MTS Service to a group and the LAST_HOME entry for HKEY_LOCAL_MACHINE/SOFTWARE/ORACLE/ALL_HOMES is not equal to the highest Oracle home number on your system, then the Add Oracle MTS Service to Group operation will fail and you will receive the following messages:

FS-10778: The Oracle MTS Service resource provider failed to configure the cluster resource <resource-name>
FS-10618: Failed to change the configuration for Windows service <service-name>
0x424: The specified service does not exist as an installed service

For example, if you have HOME0, HOME1, and HOME2 on your system, the value for LAST_HOME must be 2 for the Add Oracle MTS Service to Group operation to work. To fix the problem, change the value of LAST_HOME to match the highest Oracle home number on the local system and rerun the Add MTS Service to Group Wizard.

16.1 Discovery of Virtual Hosts with Oracle Intelligent Agent 8.1.7

To discover an Oracle Fail Safe virtual host using Oracle Intelligent Agent 8.1.7:

1. In Oracle Enterprise Manager, right-click on Nodes.

2. Select Discover Nodes.

3. Enter the virtual host name without the domain name.

If you enter the virtual host name with the domain name, you will get an error. This restriction applies only to the first time you discover a virtual host. Subsequent times, entering the virtual host name with or without the domain name will not return an error.

16.2 Partial Support of JobOut Subdirectory

Oracle Intelligent Agent release 8.1.7 deposits its jobs output files into a directory called JobOut. For highly available Intelligent Agents, the JobOut directory is a subdirectory under the agent's ConfigPath directory on the cluster disk. The Intelligent Agent requires the JobOut subdirectory to run jobs.

When creating an Oracle Intelligent Agent and adding it to a group, Oracle Fail Safe creates the JobOut subdirectory on the cluster disk. However, when verifying a group with a highly available Intelligent Agent in it, Oracle Fail Safe does not verify that the JobOut subdirectory exists. In addition, when changing a highly available Intelligent Agent's cluster disk, Oracle Fail Safe does not create a JobOut subdirectory on the new disk, nor does it remove the JobOut subdirectory on the old disk.

16.3 Fail-Safe Database Discovered as Standalone Database on Physical Nodes

When you view standalone databases on physical nodes, you may see fail-safe databases that are online on the physical node being discovered. This problem appears on systems running Windows NT version 4.0 Service Pack 5 or 6a.

If there is an Intelligent Agent resource in the group, then databases in the group will not be discovered under the physical nodes. The exception to this case is when the database uses more than one virtual address. If the first virtual address used in the listener.ora and tnsnames.ora files for the database is not selected for use by the agent, then the databases in the group will be discovered under the physical nodes.

16.4 Default Intelligent Agent of Physical Node Is No Longer Restarted by Oracle Fail Safe

Prior to Oracle Fail Safe release 3.1, the default Intelligent Agent of a physical node was restarted during the Add Database to Group and Remove Database from Group operations. Starting with Oracle Fail Safe 3.1, the default Intelligent Agent of a physical node is no longer automatically restarted. If you need an updated list of databases under the physical nodes, restart the default agent manually.

19.1 References to Microsoft Cluster Server (MSCS)

Prior to the introduction of Windows 2000, Microsoft Corporation referred to Microsoft Cluster Service as Microsoft Cluster Server. Because Oracle Fail Safe works with both Microsoft Cluster Service and Microsoft Cluster Server, both products are referred to as Microsoft Cluster Server in the Oracle Fail Safe documentation.

19.2 Warning Written to Windows Event Log When Database Polling Is Disabled

If you disable database (Is Alive) polling, then every hour the resource DLL will write a warning event to the Windows event log to indicate that database polling is disabled. (You disable database polling using Oracle Fail Safe Manager as follows: Select the database from the tree view, click the Database tab, set the Database Polling field to disabled, and then click Apply.)

19.3 Changing the IP Addresses of Cluster Nodes in a Windows Cluster

A Windows cluster system is typically composed of two nodes and two networks: a public network, which is used by clients of the cluster, and a private network called the cluster interconnect, which is used internally by the cluster hardware and software to monitor the state of the two nodes during their operation.

Thus, each node has at least two IP addresses, one on the public network and one on the private network (for example, 192.168.10.1 could be the IP address on the public network and 10.10.0.1 could be the IP address on the private network).

These addresses are specified in the Microsoft Cluster server software.

The following steps describe how to change the IP addresses of cluster nodes in a Windows NT cluster. The steps assume that there are two clusters nodes: Node A and Node B. The following steps are performed on Node A unless otherwise noted.

1. Open Oracle Fail Safe Manager and take all of the groups offline except the Cluster Group. (In the tree view, right-click a group, then choose Take Offline. Repeat until all groups except the Cluster Group have been taken offline.)

2. Close Oracle Fail Safe Manager.

3. Open MSCS Cluster Administrator and check that all of the groups are offline except the Cluster Group. Then follow these steps:

   1. In the tree view, choose the Resources folder.

   2. Right-click Oracle Services for MSCS (if you are using release 3.2.1 or later) or right-click FailSafe Service (if you are using a release prior to release 3.2.1) and choose Take Offline.

   3. Right-click a group and choose Move Group. Repeat until all groups are moved to Node B.

4. Change the Cluster IP address (for Node A) in MSCS Cluster Administrator, as follows.

   1. Select the Resources folder.

   2. Right-click Cluster IP Address.

   3. Select Properties.

   4. Change the IP Address.

   5. Click OK.

5. For each group, change its IP address in MSCS Cluster Administrator, as follows:

   1. Select the Resources folder.

   2. Right-click the Network Name <IP address> for a group and select Properties.

   3. Change the network Name on the General tab and change the IP address Name on the Parameters tab.

   4. Click Apply.

   5. Repeat steps b through d for all of the groups.

6. Change the IP Address of the network adapter on Node A. (For instructions, see Microsoft Article ID: Q230356 - Changing the IP Address of Network Adapters in Cluster Server available on the Microsoft Corporation Web site.)

7. Reboot Node A.

8. After Node A reboots, open MSCS Cluster Administrator.

9. Right-click a group and choose Move Group. Repeat until all groups are moved to Node A.

10. Change the IP Address of the network adapter on Node B.

11. Reboot Node B.

12. After Node B reboots, update the host file on both Node A and Node B (located in winnt/system32/drivers/etc/host).

13. Update the DNS Server if necessary.

14. Open MSCS Cluster Administrator and make sure that the Cluster Group and all resources are online.

15. Open Oracle Fail Safe Manager and then follow these steps:

    1. Make sure that all resources are online.

    2. Select the cluster from the tree view, and on the Troubleshooting menu, choose Verify Cluster. Make sure that no warnings are returned due to a problem with host name to IP address resolution.

    3. Select a group from the tree view, and on the Troubleshooting menu, choose Verify Group. This will report a number of errors in Oracle Net configuration files and Windows services and will ask if you want the errors to be fixed. Accept all fixes and rerun the Verify Group operation until no more fixes are suggested and no warnings are produced.

    4. Repeat step 15c for each group on the cluster.

    5. Test manual failover of each group on both nodes and then test connectivity from clients. (To test manual failover of a group, right-click the group in the Oracle Fail Safe Manager tree view, then choose Move Group.)

19.4 Using the Verify Commands

If errors are returned when you run one of the Verify commands (Verify Cluster, Verify Group, or Verify Standalone database), fix the errors and then rerun the Verify command. Repeat this process until the verify operation runs error-free.

19.5 Information on Moving the Initialization Parameter File

To move the initialization paramater (init.ora) file for a database that has been configured for high availability, follow these steps:

1. Move the initialization parameter file to the desired location.

   • If you are moving the initialization parameter file from a cluster disk to a local disk, make sure that you place a copy of the initialization parameter file in the same location on each cluster node.

   • If you are moving the initialization parameter file from one cluster disk to another, make sure that the disk to which you move the initialization parameter file is not being used by a resource in another group.

2. Select the database from the Oracle Fail Safe Manager tree view, then click the Database tab.

3. In the Parameter File box, enter the new specification for the initialization parameter file, then click Apply.

4. In the tree view, select the group that contains the database for which you have moved the initialization parameter file, then on the Troubleshooting menu, select Verify Group.

5. Using Oracle Fail Safe Manager, place the database offline and bring it back online to verify that there are no problems created by the new initialization parameter file location.

19.6 Oracle Fail Safe Tutorial

Lesson 12 of the Oracle Fail Safe Tutorial demonstrates how to add an Oracle Reports Server and Oracle HTTP Server to a group. You may find when you run the report in a Web browser, that the image associated with the Report does not load. This is a known problem for which there is no workaround. However, it does not affect failover for the Oracle Reports Server or Oracle HTTP Server.

19.7 Support for Archive Logs

In Oracle8i release 8.1.7 and later, you can define an archive log destination as a service name instead of a disk location.

For example:

log_archive_dest_1='LOCATION=f:\<dir_name> MANDATORY REOPEN=5' 
log_archive_dest_2='SERVICE=standby OPTIONAL REOPEN=120' 

The example shows typical entries that are allowed in Oracle8i release 8.1.7 for archive log support:

• The first entry specifies a file system destination. Oracle Fail Safe will check this location to make sure that the disk used is a cluster disk and accessible to all cluster nodes.

• The second entry specifies a TNS service name entry of "standby." Oracle Fail Safe will not automatically update the tnsnames.ora file on all cluster nodes. You must manually edit or add the service name entry to the tnsnames.ora file on all cluster nodes.