Plixer ML EngineΒΆ

Review/complete the recommended upgrade preparations, and then follow these steps to upgrade a Plixer ML Engine deployment to the latest version, (UPGRADE_VER should be replaced with the target version for the upgrade, e.g., 19.5.0):

  1. SSH to the Plixer ML Engine VM (i.e., the host used for management/deployment) as the plixer user.

  2. Download the installer for the latest version:

    curl -o plixer-machine-learning-update.run https://files.plixer.com/scripts/plixer-machine-learning/release/UPGRADE_VER/plixer-machine-learning-update.run

  3. Download the checksum file and validate the integrity of plixer-machine-learning-update.run:

    curl -o plixer-machine-learning-checksums.txt https://files.plixer.com/scripts/plixer-machine-learning/release/UPGRADE_VER/plixer-machine-learning-checksums.txt
cat plixer-machine-learning-checksums.txt
sha256sum plixer-machine-learning-update.run

  4. Set the correct permissions for the installer:

    chmod +x plixer-machine-learning-update.run

  5. Run the installer as the plixer (recommended) user:

    STAGE="release"
VERSION="UPGRADE_VER"
./plixer-machine-learning-update.run

After the installer script completes running, setup.sh will automatically be run to pull in any configuration changes and redeploy pods with new images.

Note

If any changes were previously made to pxi-settings.yaml, azure.tfvars, aws.tfvars, or vsphere.tfvars, it will be retained even if the upgrade package includes a newer version of the file. The updated file will instead be saved with a .dpk-dist extension, and any necessary edits should be migrated before it is used to overwrite the old configuration/tfvars file.

Once the upgrade process is complete, wait for the rke2-server service to restart. This sequence can be monitored by running:

journalctl -xeu rke2-server -f

Additional notes for Plixer ML Engine upgrades from v19.4.0 to v19.5.0

  • Plixer Scrutinizer 19.6.0 includes new management/configuration functions for the Plixer ML Engine, requiring all attached engine deployments to also be upgraded from v19.4.0 to v19.5.0 or higher.

  • After Plixer Scrutinizer is upgraded to v19.6.0, all previous settings related to attached Plixer ML Engine deployments will be reset. Engines will need to be re-registered (but not re-deployed) via the Plixer Scrutinizer web interface before being upgraded to v19.5.0.

  • When upgrading from v19.4.0 to v19.5.0, setup.sh --reconfigure will automatically be run (instead of setup.sh as described above) to initiate the new configuration process and collect all required information (including the authentication token generated by Plixer Scrutinizer).

  • If the Plixer ML Engine is deployed as a standalone VM, new Docker images will be downloaded (may take several minutes) after the package updates. This step is skipped for cloud deployments.