fix(volume-lookup): force application-defaults build on registry miss so nfs_prep pre-creates the NFS subdir

lookup('volume', role) read the lazily-populated canonical-volumes registry directly, so a lookup running before any consumer built that role (nfs_prep's pre-create loop for an early unpinned engine such as ollama) saw an empty registry and skipped the subdir; the swarm bind mount then failed with 'no such file or directory'. Force get_application_defaults() on a miss, guarded so the populated hot path skips the deepcopy. Adds two guard tests.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

plugins/lookup/volume.py

@@ -30,7 +30,7 @@ from typing import Any
 from ansible.errors import AnsibleError
 from ansible.plugins.lookup import LookupBase
 
-from utils.cache.applications import get_canonical_volumes
+from utils.cache.applications import get_application_defaults, get_canonical_volumes
 
 
 class LookupModule(LookupBase):
@@ -50,6 +50,14 @@ class LookupModule(LookupBase):
             raise AnsibleError("volume lookup: application_id must be non-empty")
 
         canonical = get_canonical_volumes(application_id)
+        if not canonical:
+            # The canonical-volumes registry fills lazily as a side effect of
+            # building the application defaults; force it on a miss so a lookup
+            # that runs before any consumer built this role (e.g. nfs_prep's
+            # pre-create loop) does not see an empty registry and silently skip
+            # the subdir. Guarded so the populated hot path skips the deepcopy.
+            get_application_defaults()
+            canonical = get_canonical_volumes(application_id)
 
         if len(terms) == 1:
             return [dict(canonical)]

tests/unit/plugins/lookup/test_volume.py

@@ -29,7 +29,7 @@ def _run(
     with patch(
         "plugins.lookup.volume.get_canonical_volumes",
         return_value=canonical if canonical is not None else {},
-    ):
+    ), patch("plugins.lookup.volume.get_application_defaults"):
         return LookupModule().run([application_id, name], variables={})
 
 
@@ -168,7 +168,7 @@ class TestVolumeLookup(unittest.TestCase):
         with patch(
             "plugins.lookup.volume.get_canonical_volumes",
             return_value=canonical,
-        ):
+        ), patch("plugins.lookup.volume.get_application_defaults"):
             result = LookupModule().run(["web-app-mattermost"], variables={})
         self.assertEqual(result, [canonical])
 
@@ -176,7 +176,7 @@ class TestVolumeLookup(unittest.TestCase):
         with patch(
             "plugins.lookup.volume.get_canonical_volumes",
             return_value={},
-        ):
+        ), patch("plugins.lookup.volume.get_application_defaults"):
             result = LookupModule().run(["web-app-no-meta"], variables={})
         self.assertEqual(result, [{}])
 
@@ -188,6 +188,32 @@ class TestVolumeLookup(unittest.TestCase):
         with self.assertRaises(AnsibleError):
             _run("web-app-mattermost", "", canonical={"data": {}})
 
+    def test_forces_defaults_build_on_registry_miss(self):
+        # Empty on the first read (lookup runs before any consumer built the
+        # role); populated after the forced build — the nfs_prep/ollama case.
+        populated = {"data": {"type": "volume"}}
+        with patch(
+            "plugins.lookup.volume.get_canonical_volumes",
+            side_effect=[{}, populated],
+        ) as gcv, patch(
+            "plugins.lookup.volume.get_application_defaults",
+        ) as gad:
+            result = LookupModule().run(["svc-ai-ollama"], variables={})
+        self.assertEqual(result, [populated])
+        gad.assert_called_once()
+        self.assertEqual(gcv.call_count, 2)
+
+    def test_populated_registry_skips_build(self):
+        # Already populated: no forced build, no wasted defaults deepcopy.
+        with patch(
+            "plugins.lookup.volume.get_canonical_volumes",
+            return_value={"data": {"type": "volume"}},
+        ), patch(
+            "plugins.lookup.volume.get_application_defaults",
+        ) as gad:
+            LookupModule().run(["web-app-mattermost"], variables={})
+        gad.assert_not_called()
+
 
 if __name__ == "__main__":
     unittest.main()