Upgrading FM

Description

This guided setup allows you to upgrade an existing Dataiku Cloud Stacks for Azure. It assumes you had followed the guided setup example to build your initial setup.

Steps

Warning

For any upgrade to Fleet Manager version 12.6.0 or higher, it is required to previously stop the virtual machine hosting Fleet Manager, or the upgrade process could fail.

Stop Fleet Manager server

• Go into the resource group into which the deployment was made. We call it <resource-group>

• Find the machine hosting Fleet Manager. Its name should be <resource-group>-instance

• Click on its name, the instance blade opens

• Click on the Stop button.

• Wait for the machine the reach the state Stopped (deallocated)

Backup Fleet Manager’s data disk

• Find the data disk, its name should be <resource-group>-instance-data-disk

• Click on its name, the volume blade opens

• Click on + Create snapshot

• Choose an identifiable name, for instance fm-backup-YYYYMMDD, and click on Review+Create

• Click on Create

• Wait for the deployment to finish, and click on Go to resource

• Click on Properties in the left menu and make a note of Resource ID value

Delete the existing server

• Go back to the instance of step Stop Fleet Manager server

• Click on Delete

• On the blade opening from the righ hand side, select OS disk and Data disks, do not select the network resources

• Tick the deletion disclaimer then click on Delete at the bottom of the blade

• Wait for the resources (machine and disks) to disappear from the resource group even after multiple refreshes of the resources list

Create the new stack

• Follow the guided setup example to deploy the new version of Fleet Manager, populate the Snapshot field with the snapshot Resource ID previously noted.

Re-provision your DSS

After upgrading Fleet Manager, the DSS instances will be seen as unresponsive. A re-provisioning of the DSS instances will be needed for the status displayed in Fleet Manager to be updated

Troubleshooting

PostgreSQL related error messages

If you are troubleshooting a non-responsive Fleet Manager after an upgrade, you might want to observe the logs displayed by sudo journalctl -u fm-setup. If you see the message Postgres server cannot be upgraded because it was not stopped properly. Please consult documentation. or PostgreSQL upgrade failed, it is likely the machine hosting Fleet Manager was not properly stopped before the upgrade. You can fix it by upgrading to an intermediate version first. Follow these instructions:

• Replay steps Stop Fleet Manager server, Backup Fleet Manager’s data disk and Delete the existing server

• Replay Create the new stack but use the following button instead

• Restart the upgrade process