TPVM Upgrade Workflow

Procedure

  1. Perform validations on user input for the device IP, firmware-host, and tpvm-image.
    1. The device IP is a registered device with the minimum supported SLX version and with associated TPVM configuration. It must be one of the EFA HA peers managing the device.
    2. Ensure that the firmware-host is registered prior to the TPVM upgrade.
    3. The tpvm-image is validated later during the SLX TPVM upgrade.
  2. Read the current TPVM configuration and operational data (including TPVM version and IP address) from the device, and then perform the following validations. TPVM configuration will be pushed to the device in the node replacement case.
    1. If TPVM is neither configured nor installed, then the TPVM configuration persisted in the EFA DB is pushed to the device and TPVM instance is installed. This operation supports the node replacement RMA case.
    2. If TPVM configuration from the device differs from the persisted EFA configuration, then the device‘s configuration has priority, and the EFA DB is updated.
    TPVM Configuration Special Handling for RMA Node Replacement Case
    • When the TPVM configuration interface management IP is set to “dhcp", the TPVM IP address must remain the same. This is due to a dependency on EFA deployment where the active node is expecting the peer node to be configured with a specific IP address. The peer node IP cannot be changed without restarting EFA HA cluster daemons on the active node.
    TPVM Configuration Special Handling for All Cases
    • You must re-apply the trusted-peer configuration on the node where it was applied previously. It exists on only one of the nodes in the EFA HA cluster. The appropriate node is identified and the trusted peer configuration is pushed to the correct node during TPVM upgrade or node replacement.
  3. Issue the appropriate SLX command to the device to upgrade or install the TPVM.
    1. Issue the tpvm upgrade command to the device. The device stops and takes a snapshot to roll back in case of failure. The device downloads the TPVM image and upgrade the TPVM instance. The TPVM starts after the upgrade of the TPVM instance, and the existing TPVM configurations are programmed on the running TPVM instance.
    2. In a node replacement case, the TPVM configuration is pushed to the device in the previous step. The tpvm deploy command is issued to the device. No TPVM snapshot is needed because the replacement switch is typically a new switch with no TPVM configured.
  4. Redeploy EFA on the upgraded or installed TPVM node with the current EFA version from the active node. Allow the redeployed peer node to rejoin the EFA HA cluster.
    1. A new deployment strategy is driven from the current active EFA node to redeploy the current EFA version only on the newly upgraded or installed TPVM peer node. Because the active EFA instance remains operational during the deployment, the peer node can rejoin without disrupting the active EFA instance.