< Previous  | TOC | Index | Next  >

Upgrading from 4.1 MP2 to 5.0

The procedure in the following sections assumes that you have a working SF Oracle RAC 4.1 MP2 cluster.

Overview of Upgrade tasks

You will have to perform the following tasks:

• Pre-Upgrade steps
• Upgrade steps - invoking the installer to upgrade to SF 5.0 Oracle RAC

  ---

    Note   When invoking the installer, use the -rsh option if you have not configured SSH communications between systems. SSH is the default.

  ---

• Post-Upgrade steps

Pre-Upgrade steps

1. Upgrade the OS on all cluster nodes:
   • For SLES9, upgrade to SLES9 SP3.
   • For RHEL4, upgrade to RHEL4 Update 3.
2. After the OS upgrade, reboot the nodes. On reboot, all the modules will load and Oracle resources will come up.
3. Log in as root user to one of the nodes in the cluster.
4. Make a backup of VCS configuration file, main.cf as follows:

   # cp /etc/VRTSvcs/conf/config/main.cf \

   /etc/VRTSvcs/conf/config/main.cf.41mp2

5. If you are using Oracle 9i, stop any gsd processes that may be started. Log in as Oracle user and enter:

   # $ORACLE_HOME/bin/gsdctl stop

6. Stop all Oracle instances in the cluster.
   • If Oracle RAC instances are under VCS control, offline the corresponding service group:

     # hagrp -offline <ora_grp> -any

   • If Oracle RAC instances are managed by CRS (Oracle 10g only), stop them using the CRS commands.
7. Stop the Oracle listener process. If it is configured under VCS control, enter:

   # hares -offline <listener-res> -sys <sys1>

   # hares -offline <listener-res> -sys <sys2>

8. If you are using Oracle 10g, stop CRS on each node in the cluster.
   • If you have configured CRS under VCS with the cssd-resource (ideal case), use the following command to stop CRS on all the nodes:

     # hares -offline cssd-resource -sys <sys1>

     # hares -offline cssd-resource -sys <sys1>

   • If you have not configured CRS under VCS (non-ideal case), stop CRS on all the nodes using:

     # /etc/init.d/init.crs stop

9. For Oracle 10g, take offline the ocr and vote volumes or file systems.
10. Take offline all other VCS groups that depend on VxFS file systems or VxVM disk groups, whether local or CFS.
11. Stop all applications that use VxFS or VxVM but that are not under VCS control. Unmount the corresponding file systems and deport the corresponding disk groups.
12. Make sure VCS is running.
13. Set to 0 the AutoStart attribute for all VCS resources for Oracle RAC instances and the cssd resource if it is configured under VCS.

Upgrade steps

1. Insert the disc containing the Veritas SF Oracle RAC software in a disc drive connected to one of the nodes for installation.
2. Log in as root on one of the systems for installation.
3. Enter:

   # mount /mnt/cdrom

4. If you are using the Symantec product installer, navigate to the directory containing the installation program. If you are using the installsfrac script, skip to step 8.
   • For RHEL 4.0 on x86_64:

     # cd /mnt/cdrom/rhel4_x86_64

   • For SLES9 on x86_64:

     # cd /mnt/cdrom/sles9_x86_64

5. Start the product installer:

   # ./installer

6. Enter "i" to install or upgrade a product.
7. From the displayed list of products to install, choose Veritas Storage Foundation for Oracle RAC. Skip to step 11 to continue the installation.
8. Navigate to the directory containing the installation program:
   • For RHEL 4.0 on x86_64:

     # cd /mnt/cdrom/rhel4_x86_64/storage_foundation_for_oracle_rac

   • For SLES9 on x86_64:

     # cd

     /mnt/cdrom/sles9_x86_64/storage_foundation_for_oracle_rac

9. To install Storage Foundation for Oracle RAC5.0, you must invoke installsfrac using the option that corresponds to your configuration:
   • To install/upgrade using ssh (default):

     # ./installsfrac galaxy nebula

   • To install/upgrade using rsh:

     # ./installsfrac -usersh galaxy nebula

10. The installer begins by presenting a copyright message.and checks that the systems are ready for installation. At the conclusion of these initial checks, and before installing any rpms, the installer:
    • Creates a log called installsfrac-<some_unique_string> on each system in the directory /var/tmp.
    • Specifies the utilities it used to communicate with the remote systems; typically this is ssh -x and scp.
    • Verifies ssh or rsh accessibility based on the mode you selected while invoking the installer.
    • Verifies the existing licenses and prompt for the following:

      Do you want to enter another license key for galaxy?

      [y,n,q](n)

      If necessary, answer "y" and enter the license keys at the prompt.

    • Checks the existing rpms installed and lists the rpms that it will replace on each system during upgrade.
11. When the installer prompts you to confirm whether to proceed with the upgrade. Answer "y" to proceed. You are prompted:

installsfrac is now ready to upgrade SFRAC.

All SFRAC processes that are currently running must be stopped.

Are you sure you want to upgrade SFRAC? [y,n,q](y)

12. When you select "y" in the previous step, the installer stops CVM and CFS agent, updates LLT and GAB configurations. The installer displays:

Stopping CVM and CFS Agents ...........................Done

Updating gabtab on galaxy ............................ Done

Updating gabtab on nebula ............................ Done

installsfrac must now make configuration updates and stop the cluster before upgrading SFRAC rpms

Are you ready to begin the Veritas Storage Foundation for Oracle RAC upgrade at this time? [y,n,q] (y)

13. If you select "y" in the previous step, the installer backs up the previous configuration files (types.cf and main.cf), updates types.cf and freezes the cvm service group. It also checks for all open volumes to verify whether VM is in use or not. The installer displays:

    Backing up types.cf and main.cf ......................... Done

    Freezing group cvm ...................................... Done

    Updating types.cf file .................................. Done

    Checking galaxy for encapsulated bootdisk ... Not encapsulated

    Checking galaxy for open volumes ........................ None

    Checking for patch(1) rpm on galaxy... version 2.5.9 installed

    Veritas Storage Foundation for Oracle RAC 5.0 Installation Program

    Checking nebula for encapsulated bootdisk ... Not encapsulated

    Checking nebula for open volumes ........................ None

    Checking for patch(1) rpm on nebula ...version 2.5.9 installed

14. The installer stops processes, uninstalls 4.1 SF Oracle RAC rpms and installs 5.0 SF Oracle RAC rpms.

    Stopping SFRAC: 100%

    Shutdown completed successfully on all systems

    Uninstalling SFRAC: 100%

    Uninstall completed successfully on all systems

    Installing SFRAC: 100%

    Installation completed successfully on all systems

Post-upgrade steps

1. After the installation is completed, installer updates main.cf on all the cluster nodes.

   Updating VCS 5.0 main.cf on galaxy ....................... Done

   Updating VCS 5.0 main.cf on nebula ....................... Done

2. Installer verifies Persistent Naming configuration on all the nodes for network interfaces used for LLT configuration (SLES9 only).

   Verifying that all NICs have PERSISTENT_NAME set correctly on galaxy:

   For VCS to run correctly, the names of the NIC cards must be boot persistent.

   Press [Enter] to continue:

   CPI WARNING V-9-0-0 No config file for NIC with MAC address 00:04:23:ac:2f:49 (present name eth3) found!

   CPI WARNING V-9-0-0 No config file for NIC with MAC address 00:04:23:ac:2f:48 (present name eth2) found!

   CPI WARNING V-9-0-0 No config file for NIC with MAC address 00:11:43:33:02:de (present name eth1) found!

   PERSISTENT_NAME is not set for all the NICs. You need to set them manually before the next reboot.

   Press [Enter] to continue:

3. Installer gives information about Volume Manager default disk group configuration.

   None of the target systems can be configured because VxVM is already installed.

   Use the vxdctl(1M) command to configure the default disk group after the installation is completed.

   Press [Enter] to continue:

4. For each system, the installer prompts you to verify the fully qualified hostnames. For example:

   ...

   Is the fully qualified hostname of system "galaxy" =

   "galaxy.example.com"? [y,n,q,b,?] (y) y

   Querying fully qualified domain name of host "galaxy" .. ok

   Is the fully qualified hostname of system "nebula" =

   "nebula.example.com"? [y,n,q,b] (y) y

5. When you are prompted about enabling Storage Foundation Management Sever, answer "n".

   ...

   Enable Storage Foundation Management Server Management?

   [y,n,q,b] (y) n

6. The installer now configures VEA, gridnode, actionagent, and Storageagent for all nodes in the cluster.

   Verify the install systems Fully Qualified Hostnames.

   Configuring VEA in STANDALONE mode on "galaxy" ..... ok

   Configuring gridnode on "galaxy" ................... ok

   Registering gridnode on "galaxy" ................... ok

   Configuring actionagent on "galaxy" ................ ok

   Registering actionagent on "galaxy" ................ ok

   Stopping StorageAgent on "galaxy" .................. ok

   Registering StorageAgent on "galaxy" ............... ok

7. When you are prompted to start SF Oracle RAC processes, accept the default "y" by pressing Enter.

   Do you want to start Veritas Storage Foundation for Oracle RAC

   processes now? [y,n,q,b] (y)

   Starting SFRAC: 100%

   Installation completed successfully on all systems

8. Installer configures CFS agents, unfreezes and brings online the cvm group.

   Configuring CFS agents:

   Confirming VxVM configuration daemons startup .. All systems Enabled

   VxVM on galaxy is being upgraded; you cannot reconfigure the default disk group.

   Unfreezing cvm ......................................... Done

   Onlining ClusterService Group on galaxy ................ Done

   Onlining cvm Group on galaxy ........................... Done

   Onlining cvm Group on nebula ........................... Done

   VxVM on nebula is being upgraded; you cannot reconfigure the default disk group.

9. At this point, all SF Oracle RAC components should be up and all service groups should be online. All eight gab ports must be up. The CSSD (Only for 10g) and Oracle resources would not be online.
10. If you are running Oracle9i, skip to step 14. If you are running Oracle 10g, replace the SKGXN libraries in $CRS_HOME and $ORACLE_HOME by using the following steps:

    ---

      Note   It is very important to perform this step when upgrading from SF Oracle RAC 4.1 MP2 to 5.0 (Oracle 10g only) because, when you reach this step, the VCSMM driver (which is running currently) is the 5.0 version, while the Oracle binaries are still linked with the old vcsmm libraries (4.1 MP2 libraries). If you skip this step and try to start the CSSD resource in the following steps, CRS may reboot all the nodes.

    ---

    For replacing the libraries on $CRS_HOME:

    # mv $CRS_HOME/lib32/libskgxn2.so

    $CRS_HOME/lib32/libskgxn2.so.old

    # cp /usr/lib/libvcsmm.so $CRS_HOME/lib32/libskgxn2.so

    # mv $CRS_HOME/lib/libskgxn2.so

    $CRS_HOME/lib/libskgxn2.so.old

    # cp /usr/lib64/libvcsmm.so $CRS_HOME/lib/libskgxn2.so

    For replacing the libraries on $ORACLE_HOME

    # mv $ORACLE_HOME/lib32/libskgxn2.so

    $ORACLE_HOME/lib32/libskgxn2.so.old

    # cp /usr/lib/libvcsmm.so $ORACLE_HOME/lib32/libskgxn2.so

    # mv $ORACLE_HOME/lib/libskgxn2.so

    $ORACLE_HOME/lib/libskgxn2.so.old

    # cp /usr/lib64/libvcsmm.so $ORACLE_HOME/lib/libskgxn2.so

11. Set to 1 the AutoStart attribute of all VCS resources for Oracle RAC instances and the CSSD resource.
12. Bring the CSSD resource online (only for Oracle 10g).

    # hares -online cssd -sys galaxy

    # hares -online cssd -sys nebula

13. Confirm that CRS is online (only for Oracle 10g). Enter:

    $CRS_HOME/bin/crs_stat -t

14. Import any VxVM disk group not under VCS control. Also mount any VxFS file system not under VCS control.
15. If you are going to upgrade Oracle, skip this step. If you are not going to upgrade Oracle, relink the existing Oracle binaries with the SF 5.0 Oracle RAC libraries (because they are currently linked with 4.1 MP2 libraries). To relink Oracle, bring up the installer menu:

    # cd /opt/VRTS/install

    # installsfrac -configure galaxy nebula

    Navigate to Install and Relink menu item. Choose the Oracle version from the menu according to the existing version on your cluster and follow the instructions to relink Oracle. On the subsequent screens, enter the Oracle username, Oracle group name, location of ORACLE_BASE, CRS_HOME (only for 10g), and ORACLE_HOME. Confirm your response in the verification screen. The installer copies the SF 5.0 Oracle RAC libraries to CRS_HOME (only for 10g) and ORACLE_HOME to relink the Oracle binaries.

16. Before you proceed to the Oracle upgrade process, you must shutdown these Oracle resources. Stop CRS and bring Oracle resources offline.
17. Upgrade Oracle if required. To upgrade to Oracle 10g Release 2 (10.2.0.2):
    • Install Oracle. See Installing Oracle 10g RAC.
    • Migrate the database. See Upgrading and migrating Oracle software for applying patchsets as required to your current version of Oracle on your specific Linux platform.