2026-02-06
Weekly roundup of SRE, Cloud Native, and Infrastructure news.
Grafana Labs caps a breakout year with $400M+ ARR
Grafana Labs announced on February 3 that its annual recurring revenue has surpassed $400M, now supporting 7,000+ organizations including 70% of the Fortune 50. Ted Young, OpenTelemetry co-founder, joined Grafana Labs, and the company donated Beyla, its eBPF-based auto-instrumentation project, to the OpenTelemetry community. GrafanaCON 2026 is scheduled for April 20-22 in Barcelona.
CNCF survey: Kubernetes hits 82% production adoption
The 2025 CNCF Annual Cloud Native Survey published on January 20 shows 82% of container users now run Kubernetes in production. 68% of respondents deploy AI workloads on Kubernetes, up from 52% the previous year. Notably, the top challenge has shifted from technical to organizational — "cultural changes with the development team" is now the #1 barrier, cited by 47%.
GenAI reshapes observability: 85% adoption, but efficiency gains lag
Elastic's 2026 observability trends report published on February 4 reveals 85% of organizations now use GenAI for observability, projected to reach 98% within two years. Top use cases include automated correlation of telemetry signals (58%), root cause analysis (49%), and remediation automation (48%). However, only 14% report substantial efficiency gains so far. OpenTelemetry production adoption nearly doubled year-over-year from 6% to 11%, with 89% of production users considering vendor compliance critical.
Reclaiming underutilized GPUs with Kubernetes scheduler plugins
CNCF published a blog post on reclaiming underutilized GPUs using custom Kubernetes scheduler plugins. The ReclaimIdleResource plugin queries Prometheus for GPU utilization metrics from DCGM and preempts pods only when actual usage falls below a configured threshold. A related DaoCloud case study demonstrated 3x improvement in GPU utilization using HAMi across 10,000+ GPU cards.
Ingress NGINX Controller retires in March 2026
Kubernetes SIG Network announced that the Ingress NGINX Controller will be officially retired in March 2026. After retirement, no security patches, CVE fixes, or new releases will be provided. Compatibility with Kubernetes 1.35+ is expected to break. SIG Network recommends all users begin migrating to Gateway API immediately, the modern persona-driven replacement for the legacy Ingress resource.
Kubernetes Node Readiness Controller: declarative node scheduling gates
SIG Node introduced the Node Readiness Controller, a new project that provides declarative scheduling gates for Kubernetes nodes via a custom resource called NodeReadinessRule. The controller addresses limitations of the standard node Ready condition by allowing operators to define complex infrastructure dependencies — such as CNI plugin health, GPU driver availability, or DaemonSet readiness — before a node accepts workloads. Key features include continuous and bootstrap-only enforcement modes, dry-run simulation for safe fleet-wide rollout, and integration with Node Problem Detector for condition reporting. Currently at alpha v0.1.1.
Prometheus v3.9: Native Histograms are now stable
Prometheus v3.9.0 released on January 6 removes the experimental native-histogram feature flag entirely. Native Histograms are now a stable feature controlled via scrape_native_histograms config setting. Compared to classic histograms, native histograms offer higher resolution with 30%+ storage efficiency, atomic composite sample storage, and compatibility with OpenTelemetry exponential histograms. Starting from v4.0, scrape_native_histograms and send_native_histograms will both default to true.