// the find
cilium/hubble-ui
Observability & Troubleshooting for Kubernetes Services
Hubble UI is the browser-based service map for Cilium's Hubble observability layer, showing live L3/L4/L7 traffic flows between Kubernetes services as a dependency graph. It's a companion tool — you're not installing this standalone, you're getting it as part of a Cilium deployment. The target audience is platform engineers and SREs who are already running Cilium and want to see what's actually talking to what.
The eBPF data source is the real advantage: flows are captured at the kernel level, so there's no service mesh sidecar tax and no blind spots from uninstrumented workloads. The backend is a thin Go proxy over Hubble's gRPC Observer API with a custom binary protocol to the frontend, which keeps the browser from drowning in raw protobuf at high flow rates. The mock data layer in the backend is actually usable — it makes local development possible without a real cluster, which is rare for k8s tooling. CI includes a kind-based integration workflow with real Cilium deployed, so the tests aren't just unit fiction.
524 stars for a project this useful suggests it lives and dies with Cilium adoption — outside that ecosystem it doesn't exist. The README is essentially a pointer to the main Cilium docs; there's nothing here explaining what the UI actually shows or how to interpret it, which is a problem when you're debugging at 2am. The frontend is React/TypeScript but the backend vendor directory is enormous (Azure SDK vendored in a Kubernetes UI backend is a head-scratcher — likely pulled in transitively), making the dependency surface hard to audit. There's also no mention of historical flow retention or querying past events; the UI appears to be live-only, so it's useless for post-incident review unless you pair it with something like Loki.