Skip to main content
Grafana is a provider: connect it once and Nexus can reach the data sources behind it and query the dashboards your team has built. All of it comes through a single connection.

What it provides

Connecting Grafana lets Nexus discover and query the data sources it fronts: Each has its own page covering what it supports and how it’s set up. When Nexus cites a result from one of these data sources, the citation links into Grafana Explore with the same query and time range. You can rerun and refine it yourself.
These are the data source types Nexus discovers from Grafana today. If Grafana fronts a tool that isn’t listed here, connect that tool directly instead; most have their own page in this section.

Dashboards

Alongside the data sources it fronts, Grafana brings your dashboards, and Nexus uses them in two ways:
  • Read as images. Nexus renders a dashboard for the incident’s time window and reads the resulting image. It interprets the charts the way a responder glancing at the dashboard would: the spike, the dip, the line that breaks trend. It isn’t limited to the numbers behind a panel; it sees the panel.
  • Learned from. The queries built into your dashboards are a record of how your team investigates: which signals matter, how they’re filtered and grouped. Nexus learns from them, which sharpens how it queries the data sources behind Grafana, dashboard or not. See How telemetry works.
Reading dashboards depends on Grafana’s image renderer. Grafana Cloud includes it; on self-hosted Grafana, install the grafana-image-renderer plugin. Where it’s missing, Nexus sets your dashboards aside rather than trying to read them, though it still learns from their queries. Connect the dashboards your team reaches for during incidents, up to 100. Nexus queries the dashboards you selected, and your selection isn’t a hard boundary for learning: Nexus can also learn from other dashboards it can see.

Setup

What you’ll need:
  • The URL of your Grafana instance. It must use HTTPS and a publicly resolvable hostname; if your Grafana runs on a private network, run a proxy and connect through it instead.
  • A Grafana service account token. It must be non-expiring: Grafana suggests an expiry when you create a token, so choose no expiration. Otherwise the connection stops working when the token lapses.
Nexus only ever reads from Grafana, so grant the service account read access in whichever of these ways fits how you manage Grafana permissions:
  • Viewer basic role: assign the Viewer basic role to the service account. It grants read and query access to all data sources, and it’s the fastest to set up.
  • Fixed roles: assign the three fixed roles Dashboards Reader, Folders Reader, and Data sources Reader. The middle ground: scoped to just what Nexus reads, without maintaining a custom role.
  • Custom role: create a custom role with exactly the actions Nexus needs, then assign it to the service account:
To connect Grafana:
  1. From the Nexus telemetry settings, add a telemetry data source and choose Grafana.
  2. Enter your Grafana URL and service account token, then test the connection. If you’re connecting through a proxy, choose it here; you can also set a public URL so that links back into Grafana open somewhere your browser can reach.
  3. Once connected, Nexus discovers the data sources behind Grafana and lists your dashboards. Enable the data sources your team uses and select their most-used dashboards.
Requests to Grafana come from a fixed set of IP addresses, so if your instance sits behind an IP allowlist you can allowlist ours.
Every data source discovered from Grafana starts disabled, so you opt in deliberately. Turn on the ones your responders rely on.

Loki

Log queries, and metrics derived from logs.

Prometheus

Metrics for how your services behaved.

Tempo

Traces that follow a request across services.

Pyroscope

Profiles showing where CPU and memory went.

CloudWatch

AWS logs and metrics, through your Grafana connection.

Telemetry overview

How providers and data sources fit together.

How telemetry works

Routing, query planning, guidance, and memory.