koku-metrics-operator:v4.4.1

Features & Capabilities:

• Query Prometheus to create reports specific to cost management.
• Package reports as a tarball and uploaded to cost management through console.redhat.com Ingress service.
• Create an integration in console.redhat.com.
• Persistent data storage.
• Collect all available metrics within Prometheus, up to 90 days.
• Air-gapped capable.
• Resource Optimization metrics collection.
• Virtual machine metrics collection.
• NVIDIA GPU and MIG (Multi-Instance GPU) metrics collection.

New in v4.4.1:

• (Bugfix) Fixed nvidia-gpu-pod-uptime-seconds metric to correctly report wall-clock pod uptime instead of GPU compute engine active time.
• Added nvidia-gpu-pod-utilization metric to track per-pod GPU utilization over time.
• (Bugfix) Fixed DCGM PromQL queries for clusters configured with honor_labels=true.
• (Bugfix) Fixed NVIDIA MIG GPU memory capacity query to correctly scope metrics to MIG GPU instances only.
• Updated dependencies.

What's Changed

• [COST-7234] generate v4.4.0 upstream bundle by @bacciotti in #906
• S390x: go test GitHub actions changes by @ShivangGoswami in #923
• fix yq download target typo in Makefile by @bacciotti in #905
• clarify upstream release order in docs by @bacciotti in #907
• Update stream by @djnakabaale in #933
• Add agentic docs by @djnakabaale in #916
• Bump the testing-framework group across 1 directory with 2 updates by @dependabot[bot] in #930
• Bump go.uber.org/zap from 1.27.1 to 1.28.0 by @dependabot[bot] in #931
• bump codecov and go versions by @djnakabaale in #940
• Fix gpu_pod_uptime to report actual pod uptime and add gpu_pod_utilization by @snecklifter in #938
• Bump the testing-framework group with 2 updates by @dependabot[bot] in #949
• [COST-7484] handle honor_labels=true in DCGM PromQL queries by @djnakabaale in #947
• [COST-7376] add whats new in v4.4.1 to CSV description by @bacciotti in #957

Full Changelog: v4.4.0...v4.4.1