CEE Log Forwarding on Fluent Worker Pods

When you enable CEE log forwarding, all logs are transferred to the configured nodes unless the filters do not match. CEE already supports the filters to drop logs from selected namespaces, pods, and pod annotations.

The new filters on Fluent worker pods that intake the logs from each node reduce the volume of logs being forwarded.

CEE supports the following new filters for log forwarding that are configurable in Logging Config mode:

• Retain logs by namespaces using the logging worker keep-namespace-logs CLI

• Retain logs by pods using the logging worker keep-pod-logs CLI

• Drop logs by Linux OS services using the logging worker drop-os-service-logs CLI

• Remove log metadata keys using the logging worker remove-keys [ keys ] CLI

Cilium Configuration

The Cilium open-source project provides networking and security capabilities for Kubernetes clusters. It is used as a networking and security addon for Kubernetes, replacing or augmenting the default Kubernetes networking components.

When you install or uninstall Cilium on the existing cluster without force-vm redeployment, each node is rebooted so that the pods will be controlled by Cilium.

You can enable Cilium in chaining mode with Calico using the following CLI command in Cluster Configuration mode:

The following CLI command is deprecated in this release:

Cluster Synchronization using Netplan

You can trigger an additional cluster synchronization phase on the SMI cluster manager for Netplan updates. The Netplan sync phase helps with optimizing cluster sync and minimizing the time taken for route additions or updates from the complete cluster sync.

You can configure the Netplan sync phase using the following CLI command in Cluster Configuration mode:

Error Handling Logs in Cluster Deployer

The cluster deployer is enhanced to support debug logging in kubeadm init for errors such as cluster synchronization failure due to misconfiguration. A debug message captures the error logs in /var/tmp/kubeadm_out.log to access the setup and retrieve the kubeadm init logs. You can view the error messages during cluster sync.

Kubernetes Version Upgrade

With this release, you can upgrade the Kubernetes version from 1.27 to 1.28.

Default Setting: Enabled – Always On

Managing Custom Grafana Dashboards

The custom Grafana dashboards in CEE Ops Center are made persistent in this release. This enhancement ensures that the dashboards are not lost during pod restart, node shutdown, or any upgrade.

The Grafana dashboards are stored in the /mnt/stateful_partition/data/cee-global/data-postgres-x directory. The valid users of the dashboards are either local users or TACACS users who have access to CEE Ops Center.

You can configure the grafana enable-basic-auth { true | false } CLI to enable or disable basic authentication. When enabled, you can perform all CRUD operations on the dashboards with the existing Grafana HTTP API.

Sequential Deployment of Helm Charts in Ops-Center

This feature provides NF users the flexibility to manage the deployment of helm charts by introducing chart priority groups.

It allows users to control the order in which the helm charts are deployed, ensuring that pods related to high-priority charts are available before deploying the lower-priority charts.

This feature introduces the following new CLI commands:

• helm ordered_deployment enable true —Use this command to enable sequential deployment.

• helm ordered_deployment chart chart_name priority_group priority_id —Use this command to order the helm charts.

  Prior to executing this command, you must enable sequential deployment.

SMI Cluster Deployer on RHEL 8.9

You can install the inception deployer on Red Hat Enterprise Linux (RHEL) to enable Linux agnostic deployment. The supported RHEL version is 8.9. This feature is an extension of the inception server installation using Ubuntu 20.04 LTS.

Updated Default Value for Pod Eviction Toleration Time

The default value for default-unreachable-toleration and default-not-ready-toleration CLI commands is updated from 300 seconds to 30 seconds. This change will enable pods to move faster to other nodes during K8s pod rescheduling when the node is unreachable or not ready.

Updated Versions for Third Party Software

SMI supports updated versions for the following third party software in this release:

• Calico—3.27.0

• Confd—7.7.14

• Containerd—1.7.13

• Docker—24.0.9

• Helm—3.14.1

• Prometheus—2.48.1

Default Setting: Enabled – Always On

UCS M7 Server Support

SMI Bare Metal supports the UCS C220 M7 server with dual socket for On-Premise and Private 5G deployments.

The Cisco UCS C220 M7 Rack Server is a versatile general-purpose infrastructure and application server. This high-density, 1RU, 2-socket rack server delivers industry-leading performance and efficiency for a wide range of workloads, including virtualization and bare-metal applications.

For more information, see the UCC SMI Deployment Guide.

UCS Server Health Check

SMI allows you to perform extended heath check and monitoring of the deployed UCS C-series servers. The existing health reporting system within the REST (Redfish) API of the UCS servers is leveraged to obtain information from multiple hardware platforms.

The server health check is a steadfast method to check and report the server health before executing a cluster-sync operation. This check reduces the disruptions caused by cluster-sync failures.

This feature supports the following new CLI commands:

• To force quit the cluster sync, you can override the health check sync failure using the following command:

  clusters cluster_name nodes node_name ucs-server ignore-health { false | true }

• To review the condition of the servers anytime, you can run the server-check synchronization phase using the following command:

  clusters cluster_name actions sync run sync-phase server-check