fix(nvsnap): repair install path - chart drift, disabled L2 fan-out, unbuildable server

src/compute-plane-services/nvsnap/.gitignore

@@ -22,8 +22,13 @@ dist/
 # Output of the go coverage tool
 coverage/
 *.out
-*.html
-!ui/dist/**.html
+# Scoped to the coverage artifact rather than a bare *.html: the broad
+# pattern silently swallowed ui/index.html, Vite's entry point, so
+# nvsnap-server could not be built from a clean checkout. The
+# !ui/dist/**.html negation that used to sit here was a workaround for
+# the same over-broad rule and is no longer needed.
+coverage.html
+*.cover.html
 
 # Go workspace
 go.work

src/compute-plane-services/nvsnap/deploy/helm/nvsnap/values.yaml

@@ -137,7 +137,7 @@ agent:
     # is no AppVersion fallback (see _helpers.tpl) — empty tag fails
     # the chart render with a clear error rather than templating a
     # known-broken image ref.
-    tag: "v0.1.3"
+    tag: "v0.2.32"
     pullPolicy: Always
 
   # nodeSelector is intentionally empty — GPU detection runs through

src/compute-plane-services/nvsnap/scripts/install-nvsnap.sh

@@ -227,6 +227,32 @@ else
     info "skipping cert-manager + webhook (--without-webhook)"
 fi
 
+# L2 per-capture PVC fan-out is off unless agent.l2.storageClass names an
+# RWX-capable class. Left empty the install still succeeds, but restore
+# fan-out silently degrades to the L3 peer cascade — a large throughput
+# difference that only shows up under multi-node fan-out, long after the
+# installer has printed a clean banner. Say so at install time.
+#
+# Reported, not auto-selected: picking the wrong class yields PVCs that
+# never bind, and the right choice depends on cluster topology. RWX
+# capability isn't exposed on the StorageClass API, so candidates are
+# matched on known RWX provisioners.
+if ! printf '%s\n' "${EXTRA_HELM_ARGS[@]}" | grep -q "agent.l2.storageClass="; then
+    rwx_re='smb\.csi|nfs\.csi|efs\.csi|filestore\.csi|azurefile|excelero|nvmesh|cephfs'
+    candidates=$(kubectl get storageclass -o jsonpath='{range .items[*]}{.metadata.name}{" ("}{.provisioner}{")"}{"\n"}{end}' 2>/dev/null \
+        | grep -iE "$rwx_re" || true)
+    echo "     WARNING: agent.l2.storageClass is unset — L2 per-capture PVC fan-out is DISABLED." >&2
+    echo "              Restore falls back to the L3 peer cascade (slower multi-node fan-out)." >&2
+    if [ -n "$candidates" ]; then
+        echo "              RWX-capable StorageClasses on this cluster:" >&2
+        echo "$candidates" | sed 's/^/                /' >&2
+        echo "              Enable with: --set agent.l2.storageClass=<name>" >&2
+    else
+        echo "              No RWX-capable StorageClass detected; L2 needs one provisioned first." >&2
+        echo "              Reference: deploy/k8s/nvcf-cluster-prep/storage-classes.yaml" >&2
+    fi
+fi
+
 # ─── 6. Helm install ───────────────────────────────────────────────────
 
 step "[6/7] helm install / upgrade nvsnap"

src/compute-plane-services/nvsnap/scripts/sync-versions.sh

@@ -27,6 +27,38 @@ source "$(dirname "${BASH_SOURCE[0]}")/versions.sh"
 
 DIRS=(deploy/k8s deploy)
 
+# Helm values files spell an image as split `repository:` / `tag:` fields
+# rather than one registry/name:tag token, so the substitution below can
+# never see them and the grep-based check below can never flag them. They
+# are handled separately by chart_tag()/set_chart_tag().
+CHART_VALUES=(deploy/helm/nvsnap/values.yaml)
+
+# Print the tag a chart values file pins for <image-name>, or nothing when
+# that repository isn't present. \042 and \047 are " and ' — spelled in
+# octal so this awk program survives shell quoting intact.
+chart_tag() {
+    awk -v name="$2" '
+        $1 == "repository:" { pending = ($2 == name) }
+        pending && $1 == "tag:" { gsub(/[\042\047]/, "", $2); print $2; exit }
+    ' "$1"
+}
+
+# Rewrite the tag a chart values file pins for <image-name>, preserving the
+# original indentation.
+set_chart_tag() {
+    local file="$1"
+    awk -v name="$2" -v ver="$3" '
+        $1 == "repository:" { pending = ($2 == name) }
+        pending && $1 == "tag:" {
+            match($0, /^[ \t]*/)
+            print substr($0, 1, RLENGTH) "tag: \"" ver "\""
+            pending = 0
+            next
+        }
+        { print }
+    ' "$file" > "$file.tmp" && mv "$file.tmp" "$file"
+}
+
 # image-name → version-var
 declare -A IMAGES=(
     [nvsnap-agent]="$NVSNAP_APP_VERSION"
@@ -49,6 +81,10 @@ for name in "${!IMAGES[@]}"; do
     for dir in "${DIRS[@]}"; do
         find "$dir" -name "*.yaml" -exec sed -i -E "${sed_re}" {} \;
     done
+    for f in "${CHART_VALUES[@]}"; do
+        [ -f "$f" ] && [ -n "$(chart_tag "$f" "$name")" ] || continue
+        set_chart_tag "$f" "$name" "$ver"
+    done
     echo "Synced ${name} -> ${new}"
 done
 
@@ -67,4 +103,28 @@ for name in "${!IMAGES[@]}"; do
     fi
 done
 
+# Same check for chart values. The grep above matches a combined
+# registry/name:tag token, which split repository:/tag: fields never form —
+# so without this loop a drifting chart tag passes verification silently.
+# That is exactly how the chart shipped nvsnap-agent v0.1.3 against an
+# NVSNAP_APP_VERSION of v0.2.32 (nvsnap#731).
+for f in "${CHART_VALUES[@]}"; do
+    [ -f "$f" ] || continue
+    for name in "${!IMAGES[@]}"; do
+        actual=$(chart_tag "$f" "$name")
+        [ -n "$actual" ] || continue
+        if [ "$actual" != "${IMAGES[$name]}" ]; then
+            echo "WARNING: ${f} pins ${name} tag ${actual}, expected ${IMAGES[$name]}" >&2
+            fail=1
+        fi
+    done
+    # The chart builds refs as <imageRegistry>/<repository>:<tag>, so a
+    # drifting registry breaks every image at once.
+    registry=$(awk '$1 == "imageRegistry:" { print $2; exit }' "$f")
+    if [ -n "$registry" ] && [ "$registry" != "$NVSNAP_REGISTRY" ]; then
+        echo "WARNING: ${f} imageRegistry is ${registry}, expected ${NVSNAP_REGISTRY}" >&2
+        fail=1
+    fi
+done
+
 exit $fail

src/compute-plane-services/nvsnap/ui/index.html

@@ -0,0 +1,13 @@
+<!DOCTYPE html>
+<html lang="en" class="dark">
+  <head>
+    <meta charset="UTF-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <title>NvSnap · GPU Capture &amp; Restore</title>
+    <link rel="icon" type="image/svg+xml" href="/favicon.svg" />
+  </head>
+  <body class="bg-[#0a0a0f] text-[#e4e4e7]">
+    <div id="root"></div>
+    <script type="module" src="/src/main.tsx"></script>
+  </body>
+</html>