UCC SMI Release Notes, Release 2026.03.1.08

First Customer Ship (FCS)

23-July-2026

End of Life (EoL)

23-July-2026

End of Software Maintenance (EoSM)

21-Jan-2028

End of Vulnerability and Security Support (EoVSS)

21-Jan-2028

Last Date of Support (LDoS)

31-Jan-2029

Hardware reliability

UPF deployment resource assignment

You can assign CPU cores, NUMA nodes, memory, and network interfaces to each UPF VM, allowing you to optimize performance and resource utilization for your workloads.

All VMs on a node must use custom resource assignment when this option is enabled.

API experience

OpenTelemetry integration for observability

You can gain real-time insight into CNDP application health and performance by integrating OpenTelemetry, allowing collection of metrics and logs for unified monitoring and faster troubleshooting across CNDP-based deployments.

Grafana is upgraded to v12.

Software reliability

Upgrade of OS to Ubuntu 24

This release recommends upgrading the SMI base image and Cluster Manager to Ubuntu 24. Additionally, it is recommended to update the Inception Server to the latest SMI disk ISO and refresh the container images in smi-app, smi-library, smi-build, smi-shared, smi-incubator, and related components.

You must upgrade the CEE to 2026.03.1.08 release along with base image upgrade and Docker to 29.x.

API experience

Kafka output integration for CEE

You can stream alert, performance, and health data directly from CEE to your Kafka clusters, enabling real-time integration with external operational systems.

This feature allows you to map data streams to Kafka topics in JSON format and configure secure connectivity, operating alongside existing northbound methods.

Licensing process

MSG SMI Smart Agent dynamic product identity support

You can now enable smart licensing activation only after the MSG SMI Smart Agent retrieves and validates the product identity from a remote endpoint.

This feature improves deployment flexibility and reliability when product identity is not available at installation.

Ease of setup

ULB 5G control-plane IPAM integration

You can select host interfaces for pod and service traffic with Cilium and enable IPAM to assign additional IP addresses to secondary pod network interfaces, supporting workloads that require multiple network connections in complex Kubernetes deployments.

Software reliability

Rust-based CDL deployment

You benefit from improved CPU and memory efficiency by enabling the CDL Rust implementation.

The Rust-based CDL container delivers the same APIs and workflows as the Go-based version and can be activated with a configuration flag, allowing seamless upgrades and uninterrupted operations.

Software reliability

Cloud Shield event-based rule framework for CNDP

You gain unified event visibility and automated incident response across all CNDP layers, enabling faster troubleshooting and root cause analysis.

This feature ingests, correlates, and acts on events from multiple sources, with system insights displayed in a centralized web dashboard.

Upgrade

Optimized SMI rolling upgrade configuration for multi-server deployments

You can customize rolling upgrade timeouts, including post-uncordon wait, drain grace period, and drain timeout, to match your cluster workloads.

This feature gives you more flexibility and control, helping upgrades proceed reliably and according to your operational requirements.

Software reliability

Strict password policy for host OS and Ops Center

You can enhance password security across Host OS and ops-center by enabling a strict password policy.

Once enabled for a cluster, you must use passwords that meet strengthened requirements, ensuring only compliant passwords are accepted during deployment and for any subsequent changes.

Hardware reliability

Mellanox 200G NIC firmware tuning enhancement

This feature allows you to flexibly adjust Mellanox 200G NIC firmware parameters by using a generic key/value model that enables advanced NIC optimization and supports high-scale deployments.

You can specify settings at the node or resource level, and the system automatically validates and applies them during cluster synchronization. The existing configurations remain unchanged unless you add new parameters.

Hardware reliability

IPM control-plane power saving

You can reduce idle power usage on Kubernetes control-plane nodes by enabling coordinated power management policies that allow CPUs and network interfaces to enter deeper energy-saving states, while your critical workloads remain unaffected.

Software reliability

Thanos Query frontend ingress for external metrics access

You can securely access complete and deduplicated CNDP cluster metrics from external clients through the Thanos Query Frontend. Each external metrics query is authenticated and aggregated across all Prometheus replicas, resolving data gaps and providing consistent results.

Ease of setup

Central CEE observability stack

This feature lets you centrally collect, store, and visualize KPI and telemetry metrics from your network, using a unified observability stack.

You gain accessible, historical metric data with integration to your dashboards, making network performance monitoring and validation simpler and more comprehensive.

NOTE: This feature is a custom development available exclusively to specific customer environments. Please contact your Cisco account representative for further details regarding your specific deployment.

Ease of setup

Support for separate installation of Longhorn and custom addons

CNDP now supports separate installation of Longhorn and custom addons outside the cluster manager package. This reduces package size, shortens offline image loading time, and helps avoid unused addon images consuming cluster resources.

You can deploy only the addons you need, making cluster setup faster, more flexible, and easier to manage across addon lifecycles.

Ease of setup

Data Layer Manager

This release introduces Data Layer Manager (DLM), a unified data services package for cloud-native Network Functions. DLM bundles key data services, including Vault, Redis/Valkey, MinIO, Postgres/pgEdge, Kafka, and CDL, into a single independently managed package with dedicated Ops-Center configuration and lifecycle management.

You can simplify deployment, versioning, scaling, and upgrades of data layer services while supporting isolated, multi-instance, and geo-redundant deployments for 5G core Network Functions.

Improved progress reporting for ordered deployment

Previous behavior: When ordered Helm deployment was enabled, the deployment progress in the SMF shell/session only showed 0% at the start and 100% at completion. Intermediate progress updates were not displayed, making it difficult to monitor deployment status or troubleshoot during long-running operations.

New behavior: With this update, ordered deployment now explicitly manages progress reporting. The CLI displays incremental, readiness-based progress updates throughout the ordered deployment flow, similar to standard deployments. Users can now see real-time progress (e.g., 0% → 10% → 40% → ... → 100%) during ordered deployment.

Customer impact: This change improves visibility into deployment status, making it easier to monitor and troubleshoot long-running deployments in real time.

Improved system shutdown handling during ordered deployment

Previous behavior: With ordered deployment enabled, issuing a system mode shutdown command during the deployment process did not take effect until all charts were fully deployed and the process reached 100%. Users could not stop the system or interrupt the deployment mid-way, which caused delays—especially when pods failed to come up or deployments stalled.

New behavior: Cancellation checks have been added to the ordered deployment feature. Now, a system mode shutdown command is observed immediately, even while the ordered deployment flow is waiting between priority groups or during resource readiness polling. This allows users to stop the system ops-center at any point in the deployment process.

Customer impact: Users can now promptly interrupt and shut down the system during ordered deployment, enabling faster troubleshooting and recovery in case of deployment issues or stalled pods.

Extended SW-RAID support for extra-storage volumes

Previous behavior: The Software RAID (SW-RAID) implementation only supported the configuration of boot and data RAID arrays. It did not provide a mechanism to provision and mount multiple standalone (non-RAID) extra storage volumes on bare-metal nodes.

New behavior: The SW-RAID implementation has been extended to support multiple extra-storage volumes. Operators can now define standalone disks (raid-level none) to be formatted and mounted automatically at boot. This is particularly useful for applications like MinIO that manage their own data redundancy and require direct access to multiple NVMe drives.

Customer impact: Customers can now leverage additional NVMe storage capacity for applications that do not require RAID-level redundancy.

Planning requirement: Disk allocation follows a strict order: Boot RAID (2 drives) is allocated first, followed by RAID volumes, and finally direct-disk (standalone) volumes.

Deployment constraint: Extra-storage volumes must be configured before the initial bare-metal installation. These volumes are provisioned during OS initialization; therefore, adding or modifying extra-storage configurations requires a full node re-installation (re-deployment).

Configuration rule: All mount-path configurations must start with /mnt/.

CPU isolation for topology-aware NRI policy

Previous behavior: The topology-aware NRI policy did not explicitly configure availableResources. As a result, logical CPU 0 could be included in the CPUs available for NRI-managed Kubernetes workloads. On K8sKVM nodes, the policy could also consider CPUs outside the Kubernetes NUMA domain.

New behavior: The topology-aware NRI policy now explicitly configures availableResources. Logical CPU 0 is excluded from the NRI-managed CPU set and remains available for base operating system and host housekeeping activities, similar to the BalloonsPolicy. On K8sKVM nodes, availableResources is further restricted to CPUs in the Kubernetes NUMA domain.

Customer impact: One fewer logical CPU is available to NRI-managed workloads on each node that uses the topology-aware NRI policy. No configuration change is normally required; however, customers with tightly sized CPU reservations should ensure that capacity calculations do not include CPU 0. On K8sKVM nodes, sufficient CPUs must be available in the Kubernetes NUMA domain because NRI will not allocate CPUs from the KVM NUMA domain.

CSCws36917

Support abort functionality for tac-debug-pkg collection

CSCwt68685

NSS Pods Running as Root User Violates Least Privilege Principle

CSCwu41967

Invalid request: Admin action make-unconfigured-good is not supported on pass through controllers.

CSCwv39084

VM Pods move to pending state

Cisco UCS C220 M8

4.3(6.260017) 6.0(2.260044)

Cisco UCS C220 M7

4.3(6.260017) 6.0(2.260044)

Cisco UCS C220 M6

4.3(6.260017) 6.0(2.260044)

Cisco UCS C220 M5

4.3(2.260007)

smi-install-disk.24.04.0-20260707.qcow2.SPA.tgz

The application-level POD ISO image signature package for use with bare metal deployments. This package contains the base ISO image as well as the release signature, certificate, and verification information.

24.04.0-20260707

smi-install-disk.24.04.0-20260707.iso.SPA.tgz

The application-level POD QCOW image signature package for use with bare metal deployments. This package contains the QCOW image as well as the release signature, certificate, and verification information.

24.04.0-20260707

cee-2026.03.1.08.SPA.tgz

The SMI Common Execution Environment (CEE) offline release signature package. This package contains the CEE deployment package as well as the release signature, certificate, and verification information.

2026.03.1.08

cluster-deployer-2026.03.1.08.SPA.tgz

The SMI Deployer image signature package for use with bare metal deployments. This package contains the Deployer image as well as the release signature, certificate, and verification information.

2026.03.1.08

dlm-1.1.12.4.SPA.tgz

The Data Layer Manager (DLM) offline release signature package.

This package contains the data store deployment files, including the release signatures, security certificates, and relevant verification information.

1.1.12.4

longhorn-2026.03.1.08.SPA.tgz

The Longhorn add-on offline release signature package.

2026.03.1.08

NED package

The NETCONF NED package. This package includes all the yang files that are used for configuration.

ncs-6.4.8.3-cisco-smi-nc-1.1.2026.03.1.08.tar.SPA.tgz

ncs-6.4.8.3-cisco-cee-nc-1.1.2026.03.1.08.tar.SPA.tgz

ncs-6.1.14-cisco-smi-nc-1.1.2026.03.1.08.tar.SPA.tgz

ncs-6.1.14-cisco-cee-nc-1.1.2026.03.1.08.tar.SPA.tgz

NSO

6.4.8.3

6.1.14

Microsoft Windows

Open a command line window and type the following command:

> certutil.exe -hashfile <filename.extension> SHA512

Apple MAC

Open a terminal window and type the following command:

$ shasum -a 512 <filename.extension>

Linux

Open a terminal window and type the following command:

$ sha512sum <filename.extension>

OR

$ shasum -a 512 <filename.extension>

Note:      <filename> is the name of the file. <extension> is the file type extension (for example, .zip or .tgz).

SMI documentation

Subscriber Microservices Infrastructure

Service request and additional information

Cisco Support