Upgrading FM

Description

This guided setup allows you to upgrade an existing Dataiku Cloud Stacks for Google Cloud. 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

• In the deployment manager, click on your deployment. We call it <deployment>

• Unwind the list of resources then click on instance-<deployment>

• Click on Manage resource at the top right

• In section Network interfaces, find the Primary internal IP address and make a note of it. We call it <fm-ip-address>

• Then, click on the vertical dots More actions menu at the top right, then click on Stop

• Wait for the instance the reach the state Stopped

Backup Fleet Manager’s data disk

• Go back to the list of resources in the deployment manager

• Click on data-<deployment>

• Click on Manage resource at the top right

• Click on Create snapshot

• Give it an identifiable name, for instance fm-backup-YYYYMMDD, make a note of it, then click on Create

• Wait for the snapshot to reach status Ready for use

Delete the existing deployment

• Go back to the deployment in the Deployment Manager

• Click on Delete

• Keep option Delete <deployment> and all resources created by it selected

• Click on Delete All

Create the new stack

• Follow the guided setup example to deploy the new version of Fleet Manager, but add new elements to the properties section:

  • privateIpAddress:<fm-private-ip>

  • snapshot:global/snapshots/<snapshot-name> or snapshot:projects/<project-name>/global/snapshots/<snapshot-name> if the snapshot is in a different project

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 step Stop Fleet Manager server above

• Make sure you still have the snapshot ID of the last working version of Fleet Manager

• Replay step Delete the existing deployment above

• Replay step Create the new stack above but use the following template URLs instead

gsutil cp gs://dataiku-cloudstacks/templates/fleet-manager/12.5.2/fleet-manager-instance.jinja .
gsutil cp gs://dataiku-cloudstacks/templates/fleet-manager/12.5.2/fleet-manager-instance.jinja.schema .

• Restart the upgrade process

DSS machines seem unresponsive

In case the DSS machines seem unresponsive in the FM UI following the upgrade, reprovision the different DSS machines for them to be able to communicate again with FM.