Production Reality (Living Diagram)

The Reality panel compares production with your model

Reality turns your C4 diagram into a living view of what's actually running. Connect read-only production sources — Kubernetes, Datadog, New Relic — and Archyl discovers your real services and their dependencies, then reconciles them against your curated model. Your diagram stops being a snapshot and starts reflecting production.

Reality is read-only and non-destructive: scanning never changes your infrastructure, and it never edits your C4 model on its own. Everything discovered is a proposal you approve.

How it works

Reality keeps three planes cleanly separated:

  • Model — your curated C4 architecture (the intent). Only you change it.
  • Reality — a continuously refreshed mirror of what's observed in production (the facts).
  • Binding — the link between the two, which you confirm.

Scanners only ever write to the Reality and Binding planes. Promoting a discovered resource into your model is always an explicit click.

Step 1 — Connect a source in your organization

Credentials live in your organization's marketplace, so you configure each source once and any project can use it.

Go to Organization → Marketplace and connect one of:

Kubernetes

A read-only connection to your cluster's API.

  • API Server URL — e.g. https://10.0.0.1:6443
  • Read-only ServiceAccount Token — create a ServiceAccount bound to the built-in view ClusterRole, then generate a token
  • Skip TLS Verification — set to true only for self-signed dev clusters

Archyl never writes to your cluster — it only calls read endpoints (describe/list).

Datadog

  • API Key and Application Key (with APM read)
  • Sitedatadoghq.com, datadoghq.eu, us5.datadoghq.com, … (must match your account region)

Datadog reads its APM Service Map, including inferred external dependencies (databases, queues, third-party services).

New Relic

  • User API Key (NRAK-…) — used to read entities and relationships
  • Account ID (optional)
  • RegionUS or EU (must match your account)

New Relic reads APM applications and OpenTelemetry services plus their CALLS relationships.

Step 2 — Add the source to a project

Open a project diagram and click Reality in the toolbar (top right). The Reality panel opens on the right.

  1. Click Add source.
  2. Pick one of your connected integrations.
  3. Add the project scope — a Kubernetes namespace or a Datadog/New Relic environment (e.g. production). Leave it blank to include everything.
  4. Click Add & scan.

A project can have any number of sources — Kubernetes and Datadog, just one, or several.

Step 3 — Scan

Each source scans independently. Use the refresh icon on a source to re-scan it. Discovery is idempotent: re-scanning updates what's already there instead of duplicating it, and anything that disappeared from production is flagged (never deleted).

Observability sources can lag: services usually appear within minutes, but relationships are computed from trace analysis and may take 10–30 minutes of sustained traffic to populate.

Step 4 — Reconcile resources

The Resources tab groups what was discovered:

  • Unmodeled in production — running in production but not yet on your diagram. For each, you can:
    • Bind — link it to a suggested existing C4 element.
    • Promote — turn it into a new C4 element (see below).
    • Ignore — dismiss it (recoverable).
  • Drifted — bound to your model but its production state diverged (e.g. unhealthy, or vanished).
  • In your model — confirmed matches between production and C4.
  • Orphan — in your model but with no production counterpart.

Use the filter field to search by name, kind, or namespace.

Promoting a service into the container level

Promoting creates a new C4 container from a discovered service. Because containers live inside a system, open the system you want first (drill into the container level). If no system is open, Promote is disabled with a hint. The new container lands in the system you're viewing.

Step 5 — Reconcile connections

The Connections tab lists discovered dependencies (who calls whom). A connection is promotable once both of its endpoints are in your model (bound or promoted). Click Promote to draw it as a real relationship on your diagram. Promoted connections drop off the proposed list automatically.

The Reality lens on the diagram

While the Reality panel is open, the canvas turns on a live skin: C4 elements that are backed by a confirmed production resource show a small health dot (and an instance count) at the top-right. This stays non-destructive — it's an overlay, not an edit. Promoted services render as clean, normal nodes.

Safety and limits

  • Read-only — Reality never writes to your production systems and never auto-edits your C4 model. You approve every change.
  • Per-source isolation — scanning one source never affects another's data.
  • Reversible — bindings can be unbound, ignored items are recoverable, and promoted elements are normal C4 elements you can edit or delete.
  • Trust levels — relationships are tagged by origin: observed (from telemetry, the strongest signal) vs config-derived (from cluster config). Telemetry sources (Datadog, New Relic, Kubernetes + a service mesh) give the most complete dependency graph.