Learn

Architecture & pattern guides

Read the idea, build it on a Grid, prove it in a challenge. Short, practical guides — illustrated with RodGrid's own nodes — on system design, design patterns, flow modeling, and how to get the most out of the app.

All articles

From a picture to a model: governance and views

Record who owns each node and what it runs on in its Governance fields, then read the same Grid one view at a time.

Who owns this box?

Point at any node on your Grid and ask who gets called when it breaks. The drawing has no answer. It cannot say which environment the thing runs in either, what data it holds, or which decision put it there. A model is a Grid that can answer those questions, because the answers are recorded on the nodes.

Ten fields on the node

Select a node and look at the end of the Node Inspector, under Governance:

  • Free text: Ownership (the team or person who answers for the node), Runtime, and SLA / SLO.
  • Comma-separated lists: Protocols and ADR references.
  • One choice from a set: Environment (production, staging, development, testing, DR, sandbox), Data sensitivity (public to restricted, plus PII, PHI and PCI), Lifecycle (planned, active, deprecated, retired), Risk (low to critical) and Cost tier (low to high). Clicking the active choice clears it.

Every field is optional, and that matters: governance is a claim that someone is accountable, so on a node nobody owns, an invented owner is worse than a blank. Architecture nodes have this section, custom components included. Pattern roles, flow nodes and shapes do not.

NoteEditing governance metadata needs Premium on that Grid. Reading it is not gated: a collaborator on any plan sees what someone else recorded. The Views panel below sits behind the same Premium gate.

The plan that counts is the Grid owner's, and anyone who cannot edit these fields sees a read-only list of those that have a value, or no Governance section at all on a node with nothing recorded.

Views are a lens, never a second diagram

Because every catalog component knows which C4 levels it belongs to, the Views panel generates four views from the one Grid: Context, Container and Component, which are C4 levels, plus Deployment, which C4 treats as a supplementary diagram. There is no Code view. Open it with the Architecture views button in the top bar. Picking a view dims every element that does not belong to it, along with any edge touching a dimmed node, and the panel counts what is left: "2 of 3 elements in this view". Nothing is copied and nothing forks, so four views never become four drawings to maintain. The lens stays on when you close the panel; All elements clears it.

One Grid, read through four views

Container keeps all three bright. Context and Component keep the portal and the service, and dim Billing DB with its persists edge. Deployment keeps only Billing DB, and both edges dim. That last one surprises people, since the service obviously runs somewhere. Membership comes from each node's catalog entry, not from your title or your system, so read a view as a filter by component type. It does not merge boxes into one system the way a hand-drawn Context diagram would. What belongs at each level in the method itself is The C4 model: context, containers, components, code. To walk someone else through the Grid, use Presenting a Grid: walkthroughs and scenario overlays.

Three ways to get less out of it

  • Expecting a view to fix a vague Grid. The lens can only group components it can classify; custom and pattern nodes have no C4 level and stay dimmed in every view. A Grid built mostly from custom nodes (see Making the catalog yours: custom components) has little to show at any level, and a Grid focused on design patterns does not offer Views at all.
  • Redrawing the same boxes for each level. Generated views remove that work, so stop and draw the system once with catalog components: each already knows which C4 levels it belongs to, and the Views panel reads that. A diagram with its own boundaries, such as a Context diagram showing the system as one box, is still drawn separately.
  • Filling every field on every node. A guessed SLA reads exactly like a real one. Fill what someone will answer for.

Ten minutes on a Grid of yours

Take a Diagram Grid of a system you know; the inspector is hidden on Challenge Grids. Pick the three nodes that would page someone at 3am and fill Ownership, Environment and SLA / SLO on each. Then click through the four views and note which of the three stay bright at each one. The useful result is the node you could not fill in: a part of the system nobody owns, or one whose answers you still have to find. Without Premium on the Grid you cannot record the answers, but that list is still worth writing down.

Practice this