Network Observability Operator release notes

This release introduces the FlowCollectorSlice API to support hierarchical governance, allowing project administrators to independently manage sampling and subnet labeling for their specific namespaces.

This release introduces a new Service deployment model in the FlowCollector custom resource. This model provides an intermediate option between the Direct and Kafka models. In the Service model, the eBPF agent is deployed as a daemon set, and the flowlogs-pipeline component is deployed as a scalable service.

The health alerts feature, introduced in previous versions as a Technology Preview feature, is fully supported as health rules in the Network Observability Operator 1.11 release.

This release introduces enhanced visualization and filtering tools in the OKD web console.

This release includes eBPF-based DNS tracking to enrich network flow records with domain names.

This release introduces automatic integration between the Network Observability Operator and the Gateway API when a GatewayClass resource is created. This feature provides high-level traffic attribution for cluster ingress and egress traffic without requiring manual configuration of the FlowCollector resource.

With this release, functional data remains visible in the Overview and Topology views even if some background queries fail. This enhancement ensures that the scope and group drop-down menus in the Topology view remain accessible during partial service disruptions.

With this release, network flows from unknown sources are categorized into four distinct groups: external, unknown service, unknown node, and unknown pod.

The default performance of the Network Observability Operator is improved for new installations. The default value for cacheActiveTimeout is increased from 5 to 15 seconds, and the cacheMaxFlows value is increased from 100,000 to 120,000 to accommodate higher flow volumes.

With this release, the Network Observability Operator monitors the status of the LokiStack resource and reports errors or configuration issues. The Network Observability Operator verifies LokiStack conditions, including pending or failed pods and specific warning conditions.

With this release, functional data remains visible in the Overview and Topology views even if some background queries fail. This enhancement ensures that the scope and group drop-down menus in the Topology view remain accessible during partial service disruptions.

Network observability alerts might not trigger when an elevated sampling rate causes the volume to fall below the lowVolumeThreshold filter. This results in fewer alerts being evaluated or displayed.

When the DNSTracking feature is enabled in a "Loki-less" installation, the required metrics for DNS graphs are unavailable. As a consequence, you cannot view DNS latency and response codes in the dashboard.

Before this update, the chart tooltip date was not displayed as intended, due to a breaking change in a dependency. As a consequence, users experienced missing date information in the OKD web console plugin’s Overview tab chart, affecting data context.

Before this update, cluster information was not refreshed after scaling, causing a warning message to persist in large clusters, not updating with changes.

Before this update, some IPs declared by OVN-Kubernetes were not enriched, causing unenriched IPs like 100.64.0.x to not appear in Machines network. As a consequence, IPs not enriched caused the wrong network visibility for users.

Before this update, a race condition during Network Observability Operator startup could cause API discovery to fail silently. As a consequence, the operator could fail to recognize the OKD cluster, leading to missing mandatory ClusterRoleBinding resources and preventing components from functioning correctly.

Before this update, some network flow fields were missing translations during the IPFIX export process. As a result, exported IPFIX data was incomplete or difficult to interpret in external collectors.

Before this update, the link to create a FlowMetric custom resource incorrectly directed users to a YAML editor instead of the intended form view. Additionally, the editor was pre-filled with incorrect default values.

Before this update, virtual machine (VM) owner types incorrectly displayed a generic question mark (?) icon in the Topology view.

Before this update, many DNS "NXDOMAIN" errors were returned due to ambiguous URLs being used in network observability.