Google makes sidecarless ambient networking generally available for GKE
Google Cloud has made ambient networking for GKE generally available, moving service-mesh functions from per-pod sidecars into shared node components and managed control-plane services. The design aims to reduce proxy overhead and restart-heavy maintenance while retaining mutual TLS, authorisation, routing and telemetry for Kubernetes workloads.
Google previews in-place migration to GKE Dataplane V2
Google Cloud has previewed in-place migration from eligible GKE Dataplane V1 clusters to Dataplane V2, avoiding a parallel-cluster rebuild while introducing eBPF and Cilium networking. The upgrade could reduce disruption for stateful and agent-heavy platforms, but eligibility, rollback, policy compatibility, observability and staged capacity testing remain critical before production adoption.
Google explains three overlapping bugs behind weeks of GKE API timeouts
Google Cloud has published a detailed account of three separate networking failures that produced similar Kubernetes API timeouts in GKE. The incident blocked around 800 developers from production deployment for weeks. Google traces the symptoms across DNS resolution, network address translation and firewall behaviour, highlighting why layered evidence matters during distributed-system diagnosis.
Google scales GKE networking for hundreds of thousands of concurrent AI-agent sandboxes
Google Cloud has detailed GKE networking changes for high-density agent sandboxes, including moving high-cardinality metadata from labels to annotations and filtering labels used for Cilium identities. Google reports 300 sandbox starts per second at sub-second latency, while Agent Substrate sustained 500 activations per second in a 200,000-agent benchmark with sub-500-millisecond suspend and resume.