Release 14.5

New in ZEDEDA Cloud Release 14.5

Article publish date: (1/29/2025)

ZEDEDA Cloud versions

  • ZEDEDA GUI: 14.5.2
  • ZEDEDA API: 14.5.0

New features 

  • Edge node resiliency: Support for edge node clusters (Early Access Program).
    • This release introduces the ability to cluster 3 EVE nodes, expanding resiliency for edge deployments. While existing support for RAID and ZFS ensures protection against disk failures, this new functionality extends protection to node failures. Nodes leverage clustered storage to ensure data availability across the cluster. Applications are load-shared, with automatic failover in the event of a node failure. Once a failed node is recovered or replaced, applications and data are automatically restored and rebalanced.
    • A new top-level menu, Edge Node Clusters, has been introduced to support this functionality. Additionally, various screens (for example: applications, network instances, volumes...) now include a toggle to enable deployment on a cluster instance.
    • Edge node clusters rely on a new EVE-OS variant that integrates clustered storage and the cluster control plane. To participate in the Early Access Program, please contact ZEDEDA.

  • New container-mode ‘reduced isolation’ for ARM64.
    • This release introduces a new container-mode for deploying containers on ARM64 platforms. Reduced isolation containers do not provide the added security of a lightweight virtual machine wrapper around the workload. While it reduces isolation, it provides containers direct access to hardware accelerators such as GPUs, NPUs, and TPUs. This is relevant for use-cases where it is not possible to virtualize or PCI passthrough the accelerator to a virtualization layer such as the Nvidia Jetson family of devices. Reduced isolation is only supported on ARM64 based hardware. The minimum EVE LTS version for this functionality is 13.4.

  • Edge-AI: Support for Nvidia Jetson platforms
    • EVE now supports running container workloads in ‘reduced isolation mode’ on Nvidia Jetson Xavier and Orin based devices. EVE includes Nvidia libraries and firmware aligned with Jetpack that using CDI (Container Device Interface) can be exposed to container workloads. Library/firmware support aligned with Jetpack 5.1.3 (starting EVE 13.4 LTS) and Jetpack 6.0 (currently master branch only) is provided.
    • Initial support requires building EVE from source. Pre-built images will soon be published on docker-hub. 13.4 LTS images can be built using ‘platform=nvidia’. Master branch images can be built using ‘platform=’nvidia-jp5’ or ‘platform=’nvidia-jp6’.
    • See more details in GitHub at How to use EVE-OS on a NVIDIA Jetson platform

  • Interface ordering for virtual and app-direct interfaces (BETA).
    • This release introduces explicit control over interface ordering for virtual and app-direct interfaces in workloads. It is currently a BETA feature and not recommended for production deployment.
    • Enabling Send Interface Order, available at the project level, will ensure the interface order over both virtual and direct interfaces is maintained as per the app definition.
      • Note that workload dependencies apply. For most modern Linux distributions running systemd with the default predictable naming enabled, the OS names interfaces according to their PCI order. This will ensure alignment with the interface order configured in the app definition. However, if your workload uses net.ifnames=0 (a legacy naming scheme, e.g., eth0, eth1,..), the kernel may ignore PCI addresses and label interfaces in an unpredictable way based on driver probing or other timing factors. In such cases the application definition defined order may not match the interface names/ordering in the guest OS.
    • More details can be found in GitHub at lf-edge app connectivity interface order.
    • The minimum EVE version for this capability is EVE 13.8.0

  • Switch Networks with multiple ports.
    • Support for switch network instances with multiple ports has been added. Using adapter labels, multiple ports can be associated to a single switched network instance. This enables redundant connectivity into an L2 segment and enables connecting multiple external end-devices into a single switch instance. EVE uses spanning tree (STP) to prevent loops and BPDU guards to prevent end-devices from participating in STP. The minimum EVE LTS version for this functionality is 13.4.

    • The UI chatbot is now available on every page. When you click the Z icon, the chatbot opens so you can enter your question. You’ll get a summary answer, and links to the Help Center for more step-by-step details if applicable.



      Z-Buddy-Splash Screen-2.gif

Enhanced features 

  • Hardware model enhancements:
    • Edge node resources such as CPU, memory, and disk are now displayed in a graphical resource card. This change is visible when onboarding a new node, on the edge node basic info page and when viewing a hardware model in the marketplace.
    • Marketplace models now allow omitting cpu/mem/disk resource details enabling creation of generic models.

  • The top-level menu item of Cluster Instances has been renamed to Cluster Orchestration to better reflect its purpose.

  • UI drawer enhancements for quick views and action menus.
    • Drawer component for Project Quick View
    • Drawer component for Models Quick View
    • Drawer component Action menu improvements
    • Drawer component Edge Node improvements

  • UI Network screen changes for IP assignment type field:
    • Previously called DHCP: Manual, Client, Passthrough
    • Now called IP Address Configuration: Static IP, DHCP Client, None

  • App policy enhancement for network assignments in deployment projects.
    • Previously when deploying an app instance with multiple network interfaces through a deployment project app-policy, the Network Instance Assignment Method had to be ‘default’ or ‘tag’ for all interfaces. Now mixing and matching of default and tag is allowed.

  • NTP server configuration enhancements ensure that an edge-node and its applications are appropriately time synchronized is key for correct operation. This release adds additional configuration capabilities:
    • The ability to define multiple NTP servers in networks and network instances.
    • It is now possible to use FQDN names for NTP servers. Note that for application workloads that receive their NTP server IP information through DHCP from a network-instance configuration, the FQDN to IP address resolution will only occur at application boot time.
    • Networks configured for ‘DCHP Client’ now allow static configuration of the NTP servers. By default the statically configured server(s) will be added to any DHCP received NTP servers. Use the ‘Override NTP Server’ option to ignore any DHCP provided NTP servers.
    • The minimum EVE version for these new capabilities is EVE 14.0.

Fixes (resolutions to customer reported bugs)

  • You can disable two factor authentication (2FA) using a time-based one-time password (TOTP)
  • When configuring a network with DHCP=manual, the option for specifying a network-range was offered. This was erroneous and has been removed. This does not affect Network Instances where a range is supported.
  • Under certain conditions the Jobs status and end-date may be incorrect. This has been corrected.
  • Changes to cluster orchestration URL in the ZEDEDA GUI.
  • App instance creation no longer fails when it shares a network instance with another app instance. The app policy is now checking if a non-default network instance already exists on the node.
  • Edge node ‘Last seen’ data retention has been extended to 3 years.

Known issues

Windows VMs in kubevirt-eve will have undefined memory usage as shown in the App Instance page, under the Utilization graphs.


Doc updates 

Was this article helpful?
2 out of 2 found this helpful