Introduction

The Nebari LGTM Pack deploys the Grafana LGTM observability stack — Loki for logs, Grafana for dashboards, Tempo for traces, and Mimir for metrics — onto a Kubernetes cluster as a single Helm release, with optional Nebari integration for routing and Keycloak SSO.

Grafana arrives with its datasources already wired to the three backends, so there is no “now connect Loki to Grafana” step. Point telemetry at the push endpoints and it shows up.

                        ┌──────────────────────────┐
                        │        Grafana UI        │
                        │  Loki · Tempo · Mimir    │
                        │  datasources pre-wired   │
                        └────────────┬─────────────┘
                                     │
            ┌────────────────────────┼────────────────────────┐
            │                        │                        │
        ┌───┴────┐              ┌────┴────┐              ┌────┴────┐
        │  Loki  │              │  Tempo  │              │  Mimir  │
        │  logs  │              │ traces  │              │ metrics │
        └───┬────┘              └────┬────┘              └────┬────┘
            │ :3100                  │ :4317 / :4318          │ :8080
            └────────────────────────┴────────────────────────┘
                             push endpoints

What ships today

• Grafana with Loki, Tempo, and Mimir provisioned as datasources by uid (loki, tempo, mimir), plus four upstream Kubernetes dashboards and a Nebari Gateway Traffic dashboard.
• Loki in SingleBinary mode with filesystem storage, fed by a Promtail DaemonSet that scrapes container logs off every node.
• Tempo with OTLP receivers on gRPC :4317 and HTTP :4318.
• Mimir, monolithic by default (single-binary StatefulSet, one PVC, 30-day retention) and horizontally scalable on demand — see Mimir deployment modes.
• kube-state-metrics and prometheus-node-exporter for Kubernetes object and node metrics, annotated for scrape discovery.
• OpenTelemetry collector overrides that redirect a NIC-managed collector’s logs, traces, and metrics pipelines into this release — no GitOps edits required. See OpenTelemetry wiring.
• A NebariApp resource giving Grafana a hostname, TLS, Keycloak OAuth, and a landing-page tile. See Nebari integration.

Two extension points

This pack is designed to be built on rather than forked. Both extension points are soft dependencies — a pack that uses them works fine on a cluster where LGTM is not installed.

• Ship a dashboard — any pack can render a ConfigMap labeled grafana_dashboard, in its own namespace, and Grafana live-loads it within ~30s.
• Route telemetry — the collector override ConfigMap is picked up by NIC’s collector without either side writing to the other’s resources.

In this guide

• Getting started — install the chart and reach Grafana
• Deploying on Nebari — the GitOps path with Argo CD
• Nebari integration — hostname, Keycloak SSO, and the landing-page tile
• Local development — k3d and Tilt on your laptop

Guides

• Mimir deployment modes — monolithic vs. distributed, and why filesystem storage is invalid in one of them
• OpenTelemetry collector wiring — how telemetry reaches the backends on a NIC cluster
• Dashboards — what ships, and how another pack contributes one

Reference

• Configuration — the values this chart owns, plus the subchart pass-through map
• Architecture — components, data flow, and storage topology