Azure ACNS adds metrics filtering

Kubernetes network telemetry gets expensive fast. Every pod, service, and node is constantly emitting signals, and the usual tradeoff is ugly — collect everything and drown in cost, or sample aggressively and miss the thing that broke production. Microsoft’s May 1 update to Advanced Container Networking Services, or ACNS, is basically an attempt to break that tradeoff for Azure Kubernetes Service. The new part is that metrics filtering and network log filtering plus aggregation are now generally available for AKS, not just preview features. (techcommunity.microsoft.com) ### What is ACNS, exactly? ACNS is Microsoft’s add-on suite for AKS networking. It bundles observability, security, and performance features so operators can see container traffic, enforce network policy, and troubleshoot failures without bolt(techcommunity.microsoft.com)ned filtering lands on Cilium. (learn.microsoft.com) ### What changed this week? Two observability features crossed into general availability. One is container network metrics filtering. The other is container network log filtering and aggregation. Microsoft framed them as a pair because they solve the same operational problem from different sides — metrics cardinality on one end, log volume on the other. The company published the GA announcement on May 1, 2026. (techcommunity.microsoft.com) ### Why is metrics filtering a big deal? Because metrics blow up quietly. In a large cluster, every extra label, namespace, pod, and flow dimension creates more time series, and Prometheus-style systems charge you in storage, query latency, an(techcommunity.microsoft.com)ead of collecting everything and throwing half of it away later. (learn.microsoft.com) ### Why does “pre-ingestion” matter so much? Filtering after ingestion is like buying every grocery item in the store and then deciding at home what dinner should be. You already paid the bill. Pre-ingestion filtering means the noisy data never gets stored in the first place, which is why Microsoft is tying this release so (learn.microsoft.com)ams narrow collection to high-value workloads and signals. (techcommunity.microsoft.com) ### What about the logs side? Logs answer different questions than metrics. Metrics tell you that packet drops or DNS failures are rising. Logs help show which flows, workloads, or verdicts were involved. ACNS now lets teams filter the network(techcommunity.microsoft.com) Logs can be stored in Azure Log Analytics or locally on nodes, depending on how the cluster is set up. (techcommunity.microsoft.com) ### Who gets the most benefit? Big AKS estates. Small clusters can often brute-force observability costs for a while. Large multi-tenant clusters cannot. They hit the wall on cardinality, ingestion bills, and dashboards full of junk long befor(techcommunity.microsoft.com)ing permanent telemetry debt. (techcommunity.microsoft.com) ### Is there a catch? Yes — the most powerful filtering controls are tied to Cilium. Microsoft’s docs say broad metric collection works across supported Cilium and non-Cilium data planes, but source-level filtering by namespace, pod label, and metric type is available on Cilium clusters, not non-Cilium ones. So the full value proposition depends on how the AKS network stack is configured. (learn.microsoft.com) ### So what’s the real takeaway? This is not a flashy new dashboard. It is plumbing. But it is useful plumbing. Microsoft is trying to make AKS observability economically survivable at scale — keep the signals that help you debug production, drop the noise that only inflates the bill. For cloud teams already deep in Azure, that is a meaningful upgrade.