[3.12] - January 31, 2022

This document provides an overview of new features, improvements, and bug fixes in the Fortanix Confidential AI 3.12 release.

Improvements

• Improved the form validation for training algorithm parameters that now supports default values, integer validation, and required vs optional fields
• Usage metrics in Fortanix DSM allow you to proactively manage usage by visualizing metrics in the Fortanix DSM Dashboard, detecting changes in activity, and configuring notifications that alert you when usage approaches a threshold.

Bug Fixes

• Fixed form validation for the "Neighbors" parameter for the k-Nearest Neighbors algorithm when creating a training flow.
• This release sets the following time-outs for managed cluster state transitions:
  • Added 15-minute timeouts for managed cluster "Creating" and "Starting" states.
  • Removed "Nonexistent" / "NotFound" state from ManagedClusterStatus (these were not used in the database or the REST API).
  • Added array bounds checks during JSON parsing of Azure Kubernetes Service results.
  • Reworked on some error reporting logic.
• CSV dataset preparation now prevents the user from accidentally using the same column as both feature and target.
• Fixed an issue where network errors could lead to misbehavior in session handling, leading to Fortanix Confidential AI users ending up on the Fortanix CCM dashboard.

Known Issues

• Provisioning a new cluster will fail or be struck randomly.

Limitations

Fortanix has a fair usage policy in this Early Customer Access Program. Hence, Fortanix has limited the resources one can create per account. Therefore, it is expected to observe a resource creation failure message once you have reached the max limit.

To report an issue/bug, visit https://support.fortanix.com/hc/en-us/requests/new.