Upgrade > Upgrade > Before the upgrade

Before the upgrade

Review and perform the tasks in this section before beginning the upgrade.

Important Read the TLS hardening during upgrade section in SA 10.51 upgrade before you start the upgrade.

Uninstalling All CORD patches

Before performing the upgrade, CORD patches (for example, 10.21 patch) must be uninstalled from all servers being upgraded.

For a Single Core host (no distributed core components), the SA Installer can automatically remove any CORD patches you have installed. However for a core with distributed components or a Multimaster Mesh with CORD patches installed, you must manually uninstall the patch from all core servers.

In a Multimaster Mesh, the patch must be uninstalled from all cores, before starting the upgrade. Rollback the patch on the secondary cores first, then the primary core, to make sure that the primary core is at a higher version than the secondary cores.

Note Satellites can be upgraded at a later time, after the cores have been already upgraded to the new version. If the cores were TLS hardened, you must use the upgrade media to rollback the patch on the satellites. After the rollback, the services are not restarted automatically and should not be started manually either (because connections to the cores are not possible due to TLS restrictions). At this point, you may start the upgrade on the satellite.

Checking whether CORD patches are applied

You can check whether patches are applied on your core servers using the following methods:

  • On any core server, by checking the install.inv file
  • On cores with distributed components, by running a Health Check Monitor test

Method 1: Check the install.inv file

On each core server, run the following command:

grep -A 2 opsware_patch /var/opt/opsware/install_opsware/inv/install.inv

If the command gives no output, then there are no CORD patches installed on the server.

For example, on a Single Host core that has a patch installed, the output would look similar to the one below:

%opsware_patch
build_id: opsware_60.0.70014.0
state:  complete
--
%opsware_patch_sql
build_id: opsware_60.0.70014.0
state:  complete

Method 2: Run a Health Check Monitor test

For cores with distributed components, you can run the SA Core Health Check Monitor (HCM) to verify that there are no CORD patched applied. To verify that all systems have had the patch removed, run the following command:

# /opt/opsware/oi_util/bin/run_all_probes.sh

Usage:

# /opt/opsware/oi_util/bin/run_all_probes.sh run check_opsware_version hosts="[<user*>@]<system>[:<password>] [[<user@>]<system>[:<password>]]..." [keyfile=<keyfiletype>:<keyfile>[:<passphrase>]]

Where:

Health Check Monitor Arguments

Argument

Description

<system>

Name/IP of a reachable SA Core server.

<user*>

Optional user to access the remote system. The user needs to have sudo permission. Default is "root".

Note This option is available only since 10.22.
<password> Optional password on the specified <system>.

<keyfiletype>

SSH keyfile type (rsa_key_file or dsa_key_file).

<keyfile>

Full path to the SSH keyfile. Specifies the file containing the current server’s SSH private key. Passwordless login with keyfile is supported only for the root user.

<passphrase>

Optional pass-phrase for <keyfile>.

You must specify all servers hosting core components in the current core (hosts="<system>[:<password>]). There are a number ways to specify login credentials for those hosts. For example, if you were using passwords, the full command would be like this:

# /opt/opsware/oi_util/bin/run_all_probes.sh \
run check_opsware_version hosts="192.168.172.5:s3cr3t \ 192.168.172.6:pAssw0rd"

The hostnames and passwords, of course, should be replaced with your actual values.

Correct output looks similar to this:

Verify base version consistent on all systems: SUCCESS

Verifying patch versions...

*** 192.168.172.5: NO PATCHES INSTALLED

*** 192.168.172.6: NO PATCHES INSTALLED

Verify consistent patch versions: SUCCESS

If the script is successful and it shows that no patches are installed as above, you can proceed with the upgrade.

If the script succeeds but there are patches installed, the output will look similar to this:

Verify base version consistent on all systems: SUCCESS

Verifying patch versions...

*** 192.168.172.5: opsware_60.0.70014.0

*** 192.168.172.6: opsware_60.0.70014.0

Verify consistent patch versions: SUCCESS

In this case, do not proceed with the upgrade without first uninstalling the patches.

For more detailed information about the The SA Core Health Check Monitor (HCM), see the SA 10.51 Administer section.

Removing CORD patches

If there are patches installed on your core(s), you can follow the procedure below to rollback the patches.

If you attempt to upgrade a server that has a CORD patch still installed, the software will abort with the following message:

ATTENTION: This system contains a version of Opsware that has been patched - upgrading or uninstalling Opsware is not permitted until this patch has been removed. Please use the following program to remove this patch from *all* core systems before attempting the upgrade:
<distribution>/opsware_installer/uninstall_patch.sh
Failure to remove the patch from all systems before beginning the upgrade may cause severe damage to the core.
Exiting Opsware Installer.

Removing CORD patches from a standalone or single-host core

The SA Installer will automatically roll back any applied CORD patches when invoked on a standalone or single-host core. No manual intervention is required.

Removing CORD patches from first and secondary cores in a multimaster mesh

CORD patches must be uninstalled on one core at a time. If the core has distributed components, you can simultaneously uninstall the CORD patches from all machines in that core that host core components.

Satellite CORD patches, however, cannot be uninstalled in at the same time as the uninstallation of core server CORD patches.

Important The patches must be removed in the following order:

1. Remove any applied CORD patches from the Secondary Core(s) (one core at a time)

2. Remove any applied CORD patches from the First (Primary) Core

The CORD patch may be removed by using the CORD patch media or the SA 10.51 media.

  • Removing patches using the patch installation media

    If the CORD installation media is available, it is recommended to rollback the patch using it. For more details, see the Release Notes for the installed patch version.

  • Removing patches using the SA 10.51 installation media – uninstall_patch.sh

    The following procedure must be applied on each of the core servers, starting with the secondary cores and after the rollback completes successfully, continuing with the primary core.

    On a core:

    1. On each server hosting SA components:

      • From the SA 10.51 media, run the uninstall patch script:

        <distro>/opsware_installer/uninstall_patch.sh –-force_run

        where <distro> is the full path to the distribution media. For example:

        /<mountpoint>/hpsa-primary/disk001

      • If this is a patched system, the following will be displayed:

        You are about to remove an Opsware patch. All core services must be running to successfully perform this operation.
        Continue (Y/N)?

        Press Y to begin the patch uninstall. The script will display progress of the uninstallation and a success message upon completion. If the CORD patch uninstall is not successful, contact your HPE Support Representative.

        All core services on the target machine must be running to remove a patch otherwise the patch uninstall process will fail.

    2. After the uninstall_patch script has completed successfully, on the Model Repository node run the following:

      <distro>/opsware_installer/uninstall_patch_db.sh –r /var/opt/opsware/install_opsware/resp/resp.<Timestamp> -–force_run

      where <distro> is the full path to the distribution media. For example:

      /<mountpoint>/hpsa-primary/disk001

      The response file specified with the –r option is created on each core server at install time.

      Note If the core has a remote database, no commands need to be run on them.

  • Removing patches using the SA 10.51 installation media – the Management Console

    To rollback patches on a core you can also use the Management Console tool. You need to invoke it once for each core, starting with the secondary cores and, after the process completes successfully, on the primary core. The tool will rollback both the component patch (opsware_patch) and the database patch (opsware_patch_db).

    To run this tool, you will need the CDF file of the core. See the "Core Definitions Files" section in the SA 10.51 Install section for details.

    From the installation media, invoke the following command:

    <distro>/opsware_installer/hpsa_mgmt_console.sh –c <path_to_CDF_file>

    where <distro> is the full path to the distribution media. For example:

    /<mountpoint>/hpsa-primary/disk001

    Select the Remove patch option in the utility menu and follow the instructions.

    For more details, see the "SA Management Console" section in the SA 10.51 Install section.

Removing CORD patches from a satellite

Note To prevent failures during the component startup (due to incompatible communication protocols between satellites and cores), after the rollback, the services are not restarted automatically and should not be started manually. The upgrade can be started while the components are stopped.

The following procedure must be applied on each of the satellite servers, which are running SA components.

  • From the SA 10.51 media, run the uninstall patch script:

    <distro>/opsware_installer/uninstall_patch.sh –-force_run

    where <distro> is the full path to the distribution media. For example:

    /<mountpoint>/hpsa-sat_base/disk001 (if the OS Provisioning components are not installed)
    /<mountpoint>/hpsa-sat_osprov/disk001

  • If this is a patched system, the following will be displayed:

    You are about to remove an Opsware patch. All core services must be running to successfully perform this operation.
    Continue (Y/N)?

    Press Y to begin the patch uninstall. The script will display progress of the uninstallation and a success message upon completion. If the CORD patch uninstall is not successful, contact your HPE Support Representative.

Send Help Center feedback