kind testing, rset graph (#207)

Signed-off-by: Daniel Guns <danbguns@gmail.com>

Author: dgunzy

.github/workflows/live-tests.yml

@@ -0,0 +1,99 @@
+name: Live Cluster Tests
+
+# Runs the #[ignore]d live regression suite (tests/live_tests.rs) against
+# real kind clusters built by scripts/dev-clusters.sh. Catches what unit CI
+# can't: API discovery, watch/list/log wire behavior, and upstream
+# flux-operator condition-message formats drifting under flux9s.
+#
+# Weekly rather than per-PR: the failure mode being guarded is upstream
+# drift (new Flux/flux-operator releases), not PR regressions.
+
+on:
+  schedule:
+    # Mondays 06:00 UTC
+    - cron: "0 6 * * 1"
+  workflow_dispatch:
+    inputs:
+      clusters:
+        description: "Dev clusters to build (dev-clusters.sh mode)"
+        type: choice
+        options:
+          - ci # simple + legacy — everything the live suite asserts against
+          - simple
+          - legacy
+          - all
+        default: ci
+
+permissions:
+  contents: read
+
+concurrency:
+  group: live-tests
+  cancel-in-progress: false
+
+env:
+  KIND_VERSION: v0.30.0
+  # kind pulls images and starts controllers from cold — give the script's
+  # readiness waits CI-appropriate headroom (defaults are tuned for warm
+  # local Docker).
+  FLUX9S_DEV_DEPLOY_WAIT_TIMEOUT: 180s
+  FLUX9S_DEV_CRD_WAIT_RETRIES: "30"
+  FLUX9S_DEV_CRD_WAIT_SLEEP_SECONDS: "5"
+
+jobs:
+  live-tests:
+    name: Live regression suite (kind)
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@stable
+
+      - name: Cache cargo registry
+        uses: actions/cache@0057852bfaa89a56745cba8c7296529d2fc39830 # v4
+        with:
+          path: |
+            ~/.cargo/bin/
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: ${{ runner.os }}-cargo-live-${{ hashFiles('Cargo.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-cargo-live-
+            ${{ runner.os }}-cargo-
+
+      - name: Install kind
+        run: |
+          # ubuntu-latest ships a root-owned kind at /usr/local/bin — sudo so
+          # our pinned version deterministically replaces it.
+          sudo curl -fsSLo /usr/local/bin/kind \
+            "https://kind.sigs.k8s.io/dl/${KIND_VERSION}/kind-linux-amd64"
+          sudo chmod +x /usr/local/bin/kind
+          kind version
+          # kubectl, helm, and envsubst ship on ubuntu-latest runners
+          kubectl version --client
+          helm version --short
+
+      - name: Compile live tests (before spending cluster time)
+        run: cargo test --test live_tests --no-run
+
+      - name: Build dev clusters
+        run: ./scripts/dev-clusters.sh "${{ inputs.clusters || 'ci' }}"
+
+      - name: Run live regression suite
+        run: cargo test --test live_tests -- --ignored --test-threads=1 --nocapture
+
+      - name: Dump cluster diagnostics on failure
+        if: failure()
+        run: |
+          for ctx in kind-flux9s-simple kind-flux9s-legacy; do
+            kubectl config get-contexts -o name | grep -qx "$ctx" || continue
+            echo "════ $ctx ════"
+            kubectl --context "$ctx" get fluxinstance,resourcesets,resourcesetinputproviders,kustomizations,gitrepositories,ocirepositories -A 2>/dev/null || true
+            kubectl --context "$ctx" get events -n flux-resources --sort-by=.lastTimestamp 2>/dev/null | tail -30 || true
+            echo "──── flux-operator logs ────"
+            kubectl --context "$ctx" logs -n flux-system deploy/flux-operator --tail=50 2>/dev/null || true
+          done

CHANGELOG.md

@@ -7,7 +7,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Fixed
+- Graph view no longer flashes when the focused node is taller than the
+  viewport (e.g. a FluxInstance's workload group): the focus auto-scroll now
+  pins oversized nodes to their top edge instead of oscillating between
+  showing their top and bottom on alternating frames.
+
 ### Added
+- Graph view downstream discovery for ResourceSet and FluxInstance (#204):
+  their `status.inventory` now renders like a Kustomization's — Flux resources
+  produced by a ResourceSet become individual navigable nodes, Deployments and
+  other workloads get the workload group with live status, and arbitrary kinds
+  (Namespaces, CRDs, custom resources) aggregate into the resource group.
 - Controller pod log viewer (#192): `:logs` opens a submenu of the discovered
   Flux controller pods (`:logs <pod|prefix>` streams one directly). Logs are
   tailed and followed live in a bounded buffer; scrolling up pauses following
@@ -66,6 +77,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   mismatches are visible instead of just confusing.
 
 ### Internal
+- New live-cluster regression suite (`tests/live_tests.rs`, all `#[ignore]`d):
+  exact assertions against the dev kind clusters' planted fixtures — graph
+  inventory, describe events, the operator's step-failure message format, pod
+  log streaming, and legacy v1beta2 version-fallback discovery. Run locally
+  with `just test-live` after `./scripts/dev-clusters.sh ci`, or via the new
+  weekly/dispatch `live-tests.yml` workflow.
 - New generic `AsyncTask<K, T>` owns the request/dispatch/poll lifecycle for
   view fetches, replacing the five copy-pasted `*_pending`/`*_fetched`/`*_rx`
   field triplets and their hand-rolled trigger/poll methods.

Cargo.toml

@@ -180,6 +180,13 @@ path = "tests/field_extraction.rs"
 name = "trace_tests"
 path = "tests/trace_tests.rs"
 
+# Live-cluster regression tests — every test is #[ignore]d; run via
+# `just test-live` after `./scripts/dev-clusters.sh ci`, or by the
+# live-tests.yml workflow (weekly / manual dispatch).
+[[test]]
+name = "live_tests"
+path = "tests/live_tests.rs"
+
 [[test]]
 name = "snapshot_tests"
 path = "tests/snapshot_tests.rs"

docs/content/user-guide/_index.md

@@ -215,7 +215,11 @@ Visualize resource relationships and dependencies. Shows upstream sources and do
 The graph view displays:
 
 - **Upstream sources** (GitRepository, HelmRepository, etc.)
-- **Managed resources** (workloads, ConfigMaps, Services, etc.)
+- **Managed resources** (workloads, ConfigMaps, Services, etc.) — for
+  Kustomizations, HelmReleases, ResourceSets, and FluxInstances alike; a
+  ResourceSet's produced Flux resources appear as individual navigable nodes,
+  and arbitrary kinds (Namespaces, CRDs, custom resources) aggregate into
+  the resource group
 - **Resource groups** (aggregated by type)
 - **Workload groups** (aggregated workloads with status)
 

justfile

@@ -20,6 +20,11 @@ test:
 test-integration:
     cargo test --test crd_compatibility --test resource_registry --test model_compatibility --test field_extraction --test trace_tests --test graph_tests
 
+# Run live-cluster regression tests against the dev kind clusters
+# (build them first: ./scripts/dev-clusters.sh ci)
+test-live:
+    cargo test --test live_tests -- --ignored --test-threads=1
+
 # Run cargo-audit to check for CVEs (ignores unmaintained warnings)
 audit:
     # RUSTSEC-2026-0002 is currently pulled in transitively via ratatui 0.29's lru dependency.

scripts/dev-clusters.sh

@@ -3,8 +3,8 @@
 #
 # Creates three clearly-named clusters:
 #
-#   flux9s-simple   — Flux 2.8.x (current repo CRD baseline); day-to-day dev testing
-#   flux9s-stress   — Flux 2.5.x; 40+ resources of every kind for page-scroll / large lists
+#   flux9s-simple   — Flux 2.9.x + source-watcher; day-to-day dev testing (steps, AG, events)
+#   flux9s-stress   — Flux 2.7.x; 40+ resources of every kind for page-scroll / large lists
 #   flux9s-legacy   — Flux 2.2.x with OCIRepository/Bucket at v1beta2; tests the
 #                     version-fallback logic (flux9s tries v1 first, then v1beta2)
 #
@@ -16,6 +16,7 @@
 #   ./scripts/dev-clusters.sh simple       # build flux9s-simple only
 #   ./scripts/dev-clusters.sh stress       # build flux9s-stress only
 #   ./scripts/dev-clusters.sh legacy       # build flux9s-legacy only
+#   ./scripts/dev-clusters.sh ci           # build simple + legacy (live-test set)
 #   ./scripts/dev-clusters.sh all          # build all three clusters
 #   ./scripts/dev-clusters.sh delete       # delete all kind clusters and exit
 
@@ -37,7 +38,7 @@ STRESS_CLUSTER="flux9s-stress"
 LEGACY_CLUSTER="flux9s-legacy"
 
 FLUX_OPERATOR_CHART="oci://ghcr.io/controlplaneio-fluxcd/charts/flux-operator"
-FLUX_OPERATOR_VERSION="0.48.0"
+FLUX_OPERATOR_VERSION="0.55.0"
 DEPLOY_WAIT_TIMEOUT="${FLUX9S_DEV_DEPLOY_WAIT_TIMEOUT:-5s}"
 CRD_WAIT_RETRIES="${FLUX9S_DEV_CRD_WAIT_RETRIES:-5}"
 CRD_WAIT_SLEEP_SECONDS="${FLUX9S_DEV_CRD_WAIT_SLEEP_SECONDS:-3}"
@@ -302,7 +303,7 @@ run_build() {
 # ── cluster builders ───────────────────────────────────────────────────────────
 
 build_simple_cluster() {
-    header "Building  $SIMPLE_CLUSTER  (Flux 2.8.x — current repo CRD baseline)"
+    header "Building  $SIMPLE_CLUSTER  (Flux 2.9.x + source-watcher — current repo CRD baseline)"
 
     kind create cluster --name "$SIMPLE_CLUSTER" --wait 60s
     local ctx="kind-${SIMPLE_CLUSTER}"
@@ -311,6 +312,21 @@ build_simple_cluster() {
     apply_fluxinstance "$ctx" "${MANIFEST_DIR}/fluxinstance.yaml"
     wait_for_flux_crds "$ctx"
 
+    # source-watcher CRDs (simple cluster only — the medium/legacy instances
+    # don't run the source-watcher component)
+    info "Waiting for source-watcher CRDs …"
+    for crd in \
+        artifactgenerators.source.extensions.fluxcd.io \
+        externalartifacts.source.toolkit.fluxcd.io; do
+        local retries="$CRD_WAIT_RETRIES"
+        while ! kubectl --context "$ctx" get crd "$crd" &>/dev/null; do
+            retries=$((retries - 1))
+            [ $retries -le 0 ] && { warn "Timed out waiting for $crd – skipping"; break; }
+            sleep "$CRD_WAIT_SLEEP_SECONDS"
+        done
+        success "  $crd"
+    done
+
     apply_simple_manifests "$ctx" "flux-resources"
 
     success "Cluster $SIMPLE_CLUSTER is ready"
@@ -320,7 +336,7 @@ build_simple_cluster() {
 }
 
 build_stress_cluster() {
-    header "Building  $STRESS_CLUSTER  (Flux 2.5.x — stress/page-scroll testing)"
+    header "Building  $STRESS_CLUSTER  (Flux 2.7.x — stress/page-scroll testing)"
 
     kind create cluster --name "$STRESS_CLUSTER" --wait 60s
     local ctx="kind-${STRESS_CLUSTER}"
@@ -407,13 +423,19 @@ main() {
             run_build "$SIMPLE_CLUSTER" build_simple_cluster
             run_build "$STRESS_CLUSTER" build_stress_cluster
             ;;
+        # The live-test set (.github/workflows/live-tests.yml): the clusters
+        # tests/live_tests.rs asserts against, skipping the heavy stress cluster.
+        ci)
+            run_build "$SIMPLE_CLUSTER" build_simple_cluster
+            run_build "$LEGACY_CLUSTER" build_legacy_cluster
+            ;;
         all)
             run_build "$SIMPLE_CLUSTER" build_simple_cluster
             run_build "$STRESS_CLUSTER" build_stress_cluster
             run_build "$LEGACY_CLUSTER" build_legacy_cluster
             ;;
         *)
-            die "Unknown mode '$mode'. Use: both | simple | stress | legacy | all | delete"
+            die "Unknown mode '$mode'. Use: both | simple | stress | legacy | ci | all | delete"
             ;;
     esac
 
@@ -422,10 +444,10 @@ main() {
     echo "  Available clusters:"
     kind get clusters 2>/dev/null | sed 's/^/    • kind-/'
     echo
-    echo "  Day-to-day dev (Flux 2.8.x, current repo CRDs):"
+    echo "  Day-to-day dev (Flux 2.9.x, current repo CRDs):"
     echo "    kubectl config use-context kind-${SIMPLE_CLUSTER} && cargo run"
     echo
-    echo "  Page-scroll stress testing (Flux 2.5.x):"
+    echo "  Page-scroll stress testing (Flux 2.7.x):"
     echo "    kubectl config use-context kind-${STRESS_CLUSTER} && cargo run"
     echo "    → Press Ctrl+f / Ctrl+b to page through the list"
     echo

scripts/dev-manifests/fluxinstance-medium.yaml

@@ -8,8 +8,8 @@ metadata:
 spec:
   distribution:
     # Pin to a mid-series release band where OCIRepository/Bucket/Helm* are stable,
-    # but the cluster still represents an older CRD surface than the repo baseline.
-    version: "2.5.x"
+    # but the cluster still represents an older (yet supported) CRD surface than the repo baseline.
+    version: "2.7.x"
     registry: "ghcr.io/fluxcd"
   components:
     - source-controller

scripts/dev-manifests/fluxinstance.yaml

@@ -9,7 +9,7 @@ spec:
   distribution:
     # Pin to the current Flux release line represented by this repo's generated CRDs.
     # This keeps the "simple" dev cluster aligned with the latest API surface flux9s targets.
-    version: "2.8.x"
+    version: "2.9.x"
     registry: "ghcr.io/fluxcd"
   components:
     - source-controller
@@ -18,6 +18,8 @@ spec:
     - notification-controller
     - image-reflector-controller
     - image-automation-controller
+    # source-watcher provides ArtifactGenerator + ExternalArtifact (both watched by flux9s)
+    - source-watcher
   cluster:
     type: kubernetes
     multitenant: false

scripts/dev-manifests/simple/kustomize.yaml

@@ -62,3 +62,21 @@ spec:
   sourceRef:
     kind: Bucket
     name: terraform-state
+---
+# Deliberate failure case: the path doesn't exist in the repo, so this
+# Kustomization emits ReconciliationFailed Warning events on every retry —
+# feed for the :events view, describe Events section, and unhealthy filter.
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: broken-path-demo
+  namespace: flux-resources
+  labels: {app.kubernetes.io/managed-by: flux9s-dev}
+spec:
+  interval: 5m
+  retryInterval: 2m
+  path: ./does-not-exist
+  prune: true
+  sourceRef:
+    kind: GitRepository
+    name: flux-system