0.3.1: drop premature VolumeSnapshotClass — depends on external snapshotter CRDs (deferred to phase 3)

templates/longhorn.yaml

@@ -23,21 +23,10 @@ parameters:
   fromBackup: ""
   # fsType: ext4 is the default; explicit so the contract is in Git.
   fsType: "ext4"
----
-# VolumeSnapshotClass — the API surface the Tower backend will
-# consume in Phase 3 to take snapshots of source PVCs before
-# spawn-env / Refresh ↓.
-#
-# `type: snap` (Longhorn's in-place snapshot) is what we want for
-# the fast Clone path: the snapshot is a CoW reference, not an S3
-# upload. `type: bak` is the slower S3-bound block backup, used
-# only by the daily DR layer.
-apiVersion: snapshot.storage.k8s.io/v1
-kind: VolumeSnapshotClass
-metadata:
-  name: longhorn-snapshot-class
-driver: driver.longhorn.io
-deletionPolicy: Delete
-parameters:
-  type: snap
 {{- end }}
+# NOTE: The CSI VolumeSnapshotClass (snapshot.storage.k8s.io/v1) lives
+# in Phase 3 of ADR 0003, alongside the Tower code that consumes it.
+# It depends on the external-snapshotter CRDs which Longhorn does not
+# ship — they're a separate install. Adding it here would fail the
+# initial sync on a fresh cluster. Phase 3 will land both together
+# (snapshotter CRDs + VolumeSnapshotClass + Tower VolumeClone path).