Warning This document applies to reverting from SGC 3.0.0.x to an older SGC 3.0.0.x release. It cannot be used to revert to a release prior to SGC 3.0.0.0.

This section describes the process required to perform an online manual revert. During this process the cluster remains in service and provided that the connected TCAP stacks are using the ocss7.sgcs connection method, calls will fail over from one SGC to another as required.

Note Failover cannot be guaranteed to be 100% successful as failover for any given dialog or message is highly dependent on timing. For example, a message queued in an SGC at the SCTP layer will be lost if that SGC is terminated prior to transmission.

Manual Revert Procedure

  1. Backup the cluster.

  2. Prepare the replacement nodes. Note that each node must conform to the recommended installation structure. Either:

    1. Use the previous installation of the SGC version being reverted to.

    2. Restore a backup of the previous installation of the SGC version being reverted to.

    3. Create a fresh installation of the SGC version being reverted to.

    4. Ensure that the current installation supports the version being reverted to. This can be verified by running the CLI command display-info-nodeversioninfo and verifying that the target version is listed in the supportedFormats column.

    5. Do not copy configuration files yet.

  3. Issue the CLI command: start-revert: target-format=$TARGET_CLUSTER_FORMAT. This command:

    • Checks pre-requisites and places the cluster into REVERT mode.

    • Converts live cluster data to the specified target data format.

    • Saves SGC configuration in the target data format (sgc.dat).

      While in this mode:

    • Calls continue to be processed.

    • Older SGC versions may join the cluster, provided that they are compatible with the target cluster version.

    • Configuration changes will be rejected.

  4. Ensure that the current cluster format is set to the target format. This can be verified by running the CLI command display-info-clusterversioninfo and ensuring that the currentClusterFormat column is set to the target version.

  5. Copy configuration (config/* and var/*) from the original cluster members to the replacement cluster members. This ensures that in the event of cluster failure the first node to restart initializes the correct configuration and not an empty configuration. This step may only be carried after once the cluster is in REVERT mode as prior to this time the configuration files may be saved in a new format not understood by the target nodes.

  6. Revert the first cluster member:

    1. Stop the original node: $ORIGINAL_SGC_HOME/bin/sgc stop

    2. Verify that the original node has come to a complete stop by checking its logs and the process list.

    3. Start the replacement node: $REPLACEMENT_SGC_HOME/bin/sgc start

    4. Verify that the replacement node has started and successfully joined the cluster. The CLI command display-info-nodeversioninfo can be used to view the current cluster members.

    5. Wait for 2-3 minutes to allow the cluster to redistribute shared data amongst all of the members.

  7. Repeat the previous step for each of the remaining cluster members. This must be performed one node at a time.

  8. Issue the CLI command: complete-revert. This checks pre-requisites, then performs the actions required to leave REVERT mode.

  9. Verify that the cluster has completed the revert. The CLI commands display-info-nodeversioninfo and display-info-clusterversioninfo may be used to verify this.

Rolling Back An In-Progress Manual Revert

  • Before start-revert was issued:

    • (Optional) Delete the installation directories for the (unused) replacement cluster members.

  • After complete-revert:

    • This is an upgrade operation, not a rollback.

  • After start-revert and before complete-revert:

    1. For every cluster member that is running the replacement SGC version, ONE AT A TIME:

      1. Stop the SGC: $REPLACEMENT_SGC_HOME/bin/sgc stop

      2. Verify that the node has come to a complete halt by checking its logs and the process list.

      3. Start the original SGC: $ORIGINAL_SGC_HOME/bin/sgc start

      4. Verify that the original SGC has started and successfully joined the cluster. The CLI command display-info-nodeversioninfo can be used to view the current cluster members.

      5. Wait for 2-3 minutes to allow the cluster to redistribute shared data before proceeding to the next node.

    2. Once all nodes are running the original pre-revert version, complete the rollback by issuing the abort-revert CLI command.

Previous page Next page