In many regulated industries, especially across India, not every Kubernetes cluster runs in the cloud. Banks & government systems often operate air-gapped Kubernetes clusters with no internet access. In these environments, SaaS-based observability assumptions does not hold good &Observability becomes core Kubernetes infrastructure, not a service.

This session explains why air-gapped K8s environments matter and outlines the challenges they introduce, including constrained scaling, offline upgrades, local image management, and storage-bound observability pipelines.

We then walk through a real-world journey of building K8s observability in an air-gapped setup using OpenTelemetry and self-hosted LGTM stack (Loki, Grafana, Tempo, Mimir). The talk covers in-cluster telemetry design, cardinality control, offline upgrades, autoscaling without internet and operating observability components as first-class Kubernetes workloads, followed by practical limitations and operational best practices.