Skip to content

Your_Intent_Repository.md

Latest commit

 

History

History
74 lines (54 loc) · 3.15 KB

Your_Intent_Repository.md

File metadata and controls

74 lines (54 loc) · 3.15 KB
Raw
title Your intent repository

Your intent repository

Grid’s source of truth for what should exist is a Git repository you create and own — the intent remote Core Syncs into inventory.

The public grid-config repository is an example layout — learn from it, copy patterns, open PRs that improve samples. Do not fork it for production and forget ownership.

Recommended setup

  1. Create a new empty repository on the Git host you use (GitHub, GitLab, or another host that speaks Git). Private repos are expected. Where the username, token, and expiry go: Intent repository access.
  2. Copy a minimal tree from grid-config (one project, one cloud, one environment, one unit).
  3. Replace every REPLACE_* / lab CIDR / account-specific value.
  4. Point Grid at that remote with a clean URL (no token in the URL). Production HTTPS also needs a host username, a host-issued token, and that token’s expiry. SSH uses a deploy key instead. Full steps: Intent repository access.
GRID_GITOPS_REPO_URL=https://git.example.com/<org>/<your-infra-config>.git
GRID_GITOPS_BRANCH=main
GRID_GITOPS_PATH=          # empty = repo root (projects/… at top level)
GRID_GITOPS_SYNC_INTERVAL_SEC=600
  1. Sync from Admin → Sources or Infrastructure.

Canonical layout

projects/
  <project-slug>/
    <cloud>/                 # aws | gcp | azure | oracle | …
      <environment>/         # development | staging | production | custom
        <infra-type>/        # vpc | ec2 | gke | …
          <unit-name>.json   # one deployable unit
archive/                     # optional generated Terraform (exit buffer)

Examples from the public sample:

projects/demo-app/aws/production/vpc/example-vpc.json
projects/grid-labs/aws/development/ec2/grid-labs-vm.json

Full rules: Desired-state layout.

Ownership rules that keep GitOps sane

Do Don’t
PR JSON changes; review like application code Hand-edit the live GRID_CONFIG_ROOT clone and expect it to survive Sync
Keep secrets out of JSON (use cloud IAM / secret refs) Commit long-lived cloud keys
One unit ≈ one engine root (Terraform workspace or Kubernetes workload) Stuff unrelated stacks into a single giant JSON
Use metadata.dependsOn for cross-unit references Assume Grid merges dependent units into one apply

How Grid uses the repo

On Sync, Core:

  1. Fetches and hard-resets the working tree to origin/<branch> (tracked files).
  2. Discovers *.json units under the configured path prefix.
  3. Upserts infrastructure records (create / update / unchanged).
  4. Marks previously tracked paths that disappeared as stale — it does not destroy cloud resources.

Details: GitOps sync.

Related