feat(instances): a VM's destruction says what goes, and keeps what a kept file needs

The Destroy tab was a sub-tab with one sentence and a link; its
confirmation page said only "This action is irreversible!" over a
"Remove Instance's data" switch. The tab now shows the state and
"Review destruction". The owner chose the confirmation layout from
three rendered options: the VM and its host, that its definition,
saved state and snapshot records go and that a running VM is forced
off, two plans for the files (delete the disk files, or keep them), the
VM's disks, what is always kept, the NVRAM choice, and Cancel.

Codex reviewed destroying; this is fixed:

- A disk file that cannot be deleted keeps the files below it, which
  it needs: they were deleted after it, leaving it broken. The files
  are deleted overlays first (split_disk_paths_by_use orders them by
  how many files each grows from), and if the files a kept one needs
  cannot be read, none of the rest goes. A message names each file
  kept so, and a failed deletion says the VM is removed.

instances/templates/instances/destroy_instance_form.html

@@ -1,38 +1,52 @@
 {% extends 'base.html' %}
 {% load i18n %}
-{% block title %}
-{% trans "Confirm Destroy" %}
-{% endblock title %}
+{% block title %}{% blocktrans with name=instance.name %}Destroy {{ name }}{% endblocktrans %}{% endblock title %}
 
 {% block page_heading %}
-{% trans "Destroy instance" %} {{ instance }}
+{% blocktrans with name=instance.name %}Destroy {{ name }}{% endblocktrans %} <small class="text-body-secondary fs-6">{% blocktrans with host=instance.compute.name %}on {{ host }}{% endblocktrans %}</small>
 {% endblock page_heading %}
 
 {% block content %}
-{% if can_delete %}
-    <div class="alert alert-danger">
-        {% trans "This action is irreversible!" %}
-    </div>
-    <form action="{% url 'instances:destroy' instance.id %}" class="form" method="post" role="form" id="delete_form">{% csrf_token %}
-        <div class="form-check form-switch">
-            <input class="form-check-input" type="checkbox" name="delete_disk" value="true" checked id="delete_disk">
-            <label class="form-check-label fw-bold" for="delete_disk">{% trans "Remove Instance's data" %}</label>
+<p class="small">{% trans "The VM's definition, saved state and snapshot records are removed. A running VM is forced off first; what the guest has not written to disk is lost." %}</p>
+<form action="{% url 'instances:destroy' instance.id %}" method="post" id="delete_form">{% csrf_token %}
+    <fieldset class="mb-3">
+        <legend class="h6">{% trans "What should happen to the files?" %}</legend>
+        <div class="row g-2">
+            <div class="col-12 col-md-6">
+                <label class="border rounded p-3 d-block h-100">
+                    <input class="form-check-input me-2" type="radio" name="delete_disk" value="true" checked>
+                    <strong>{% trans "Delete disk files" %}</strong>
+                    <span class="small d-block mt-2">{% trans "With them go this VM's external snapshot files, snapshot memory files and cloud-init seed." %}</span>
+                </label>
+            </div>
+            <div class="col-12 col-md-6">
+                <label class="border rounded p-3 d-block h-100">
+                    <input class="form-check-input me-2" type="radio" name="delete_disk" value="false">
+                    <strong>{% trans "Keep the disk files" %}</strong>
+                    <span class="small d-block mt-2">{% trans "They stay with their internal snapshot data; the snapshot records still go with the VM." %}</span>
+                </label>
+            </div>
         </div>
-        {% if instance.nvram %}    
-        <div class="form-check form-switch">
-            <input class="form-check-input" type="checkbox" name="delete_nvram" value="true" id="delete_nvram" checked>
-            <label class="form-check-label fw-bold" for="delete_nvram">
-                {% trans "Remove Instance's NVRAM" %}
-            </label>
-        </div>
-        {% endif %}
-        <button type="submit" class="btn btn-danger float-end" name="delete">
-            {% trans "Destroy" %}
-        </button>
-    </form>
-{% else %}
-    <div class="alert alert-danger" role="alert">
-        {% trans "You cannot destroy instance!" %}
+    </fieldset>
+    {% if instance.config_disks %}
+    <p class="small fw-semibold mb-1">{% trans "Disks" %}</p>
+    <ul class="small ps-3">
+        {% for disk in instance.config_disks %}
+        <li><strong>{{ disk.dev }}</strong> <span class="text-break">{{ disk.path|default:_("no file") }}</span></li>
+        {% endfor %}
+    </ul>
+    {% endif %}
+    <p class="small text-body-secondary">{% trans "A file another VM on this host uses, directly or under its own disk, is kept, and so is the template a disk was made from." %}</p>
+    {% if instance.nvram %}
+    <div class="form-check mb-3">
+        <input class="form-check-input" type="checkbox" name="delete_nvram" value="true" id="delete_nvram" checked>
+        <label class="form-check-label" for="delete_nvram">{% trans "Delete the NVRAM file (firmware settings). Untick to keep it." %}</label>
     </div>
-{% endif %}
-{% endblock content %}
+    {% endif %}
+    <div class="border-top pt-3 d-flex flex-wrap align-items-center gap-2">
+        <span class="small text-danger me-auto">{% trans "This cannot be undone." %}</span>
+        <a class="btn btn-secondary" href="{% url 'instances:instance' instance.id %}">{% trans "Cancel" %}</a>
+        <button class="btn btn-danger" type="submit" name="delete">{% trans "Destroy VM" %}</button>
+    </div>
+</form>
+{% endblock content %}

instances/templates/instances/destroy_tab.html

@@ -1,27 +1,16 @@
 {% load i18n %}
+{% load vm_states %}
 <div class="tab-pane" id="undefine" role="tabpanel">
-    <!-- Nav tabs -->
+    {% vm_state instance.status as state %}
     <div class="card">
-    <div class="card-header bg-transparent pb-0">
-    <ul class="nav nav-underline small" role="tablist" aria-label="Instance destroy menu">
-        <li class="nav-item">
-            <a class="nav-link active" href="#destroy" aria-controls="destroy" role="tab" data-bs-toggle="tab">
-                {% trans "Destroy Instance" %}
-            </a>
-        </li>
-    </ul>
-    </div>
-    <!-- Tab panes -->
-    <div class="tab-content card-body">
-        <div role="tabpanel" class="tab-pane active" id="destroy">
-            <p>{% trans 'This action starts remove instance process' %}</p>
+        <div class="card-body">
+            <p class="small text-body-secondary mb-2">{% trans "Destroy" %} · {{ state.0 }}</p>
             {% if can_delete %}
-                <a href="{% url 'instances:destroy' instance.id %}" class="btn btn-danger float-end">{% trans "Destroy" %}</a>
+            <a href="{% url 'instances:destroy' instance.id %}" class="btn btn-danger mb-2">{% trans "Review destruction" %}</a>
+            <p class="small text-body-secondary mb-0">{% trans "The next page shows what goes and lets you keep the disk files." %}</p>
             {% else %}
-                <button class="btn btn-danger disabled float-end" name="delete">{% trans "Destroy" %}</button>
+            <p class="mb-0">{% trans "Only the owners allowed to delete this VM and administrators can destroy it." %}</p>
             {% endif %}
-            <div class="clearfix"></div>
         </div>
     </div>
-    </div>
-</div>
+</div>

instances/tests/test_destroy.py

@@ -1,5 +1,5 @@
 from contextlib import contextmanager
-from unittest.mock import patch
+from unittest.mock import MagicMock, patch
 
 from computes.models import Compute
 from django.contrib.auth import get_user_model
@@ -64,6 +64,39 @@ class DestroyOrderTestCase(TestCase):
                 self.assertEqual(response.status_code, 200)
                 self.assertEqual(proxy.get_volume_by_path.called, deleted)
 
+    def test_a_disk_file_that_stays_keeps_the_files_it_needs(self):
+        # the overlay is not deleted: the file below it would leave it broken
+        volumes = {}
+
+        def volume(path):
+            vol = volumes.setdefault(path, MagicMock())
+            if path == "/pool/vm.snap1":
+                vol.delete.side_effect = libvirtError("Permission denied")
+            return vol
+
+        with patch("instances.models.wvmInstance") as wvm:
+            proxy = wvm.return_value
+            proxy.instance.isActive.return_value = False
+            proxy.split_disk_paths_by_use.return_value = (["/pool/vm.snap1", "/pool/vm.qcow2", "/pool/vm-data.qcow2"], [])
+            proxy.get_volume_by_path.side_effect = volume
+            proxy.backing_path.side_effect = {"/pool/vm.snap1": "/pool/vm.qcow2"}.get
+            response = self.client.post(reverse("instances:destroy", args=[self.instance.id]), {"delete_disk": "true"})
+        self.assertNotIn("/pool/vm.qcow2", volumes)
+        volumes["/pool/vm-data.qcow2"].delete.assert_called_once_with(0)
+        self.assertEqual([str(m) for m in get_messages(response.wsgi_request)][-1],
+                         "Disk /pool/vm.qcow2 was kept: a disk file that stays needs it")
+
+    def test_when_the_files_a_kept_one_needs_are_unknown_none_of_the_rest_goes(self):
+        with patch("instances.models.wvmInstance") as wvm:
+            proxy = wvm.return_value
+            proxy.instance.isActive.return_value = False
+            proxy.split_disk_paths_by_use.return_value = (["/pool/a.qcow2", "/pool/b.qcow2"], [])
+            proxy.get_volume_by_path.return_value.delete.side_effect = libvirtError("Permission denied")
+            proxy.backing_path.side_effect = libvirtError("pool unreachable")
+            response = self.client.post(reverse("instances:destroy", args=[self.instance.id]), {"delete_disk": "true"})
+        self.assertEqual(response.status_code, 302)
+        proxy.get_volume_by_path.assert_called_once_with("/pool/a.qcow2")
+
     def test_disks_are_deleted_after_the_undefine(self):
         proxy = self.destroy()
         proxy.get_volume_by_path.assert_called_once_with("/pool/own.qcow2")

instances/views.py

@@ -822,13 +822,25 @@ def destroy(request, pk):
                 proxy.delete(flags)
                 instance.delete()
 
-            # each disk has its own row: the VM is gone even if one is kept
+            # each disk has its own row: the VM is gone even if one is kept; a
+            # file that stays keeps the files below it, which it needs
+            kept = set()
             for path in to_delete:
+                if path in kept:
+                    messages.warning(request, _("Disk %(path)s was kept: a disk file that stays needs it") % {"path": path})
+                    continue
                 try:
                     with audit_vm(request, instance, f"delete volume {path}"):
                         proxy.get_volume_by_path(path).delete(0)
                 except libvirtError as err:
-                    messages.error(request, _("Disk %(path)s was not deleted: %(err)s") % {"path": path, "err": error_text(request, err)})
+                    messages.error(request, _("The VM is removed, but disk %(path)s was not deleted: %(err)s") % {
+                        "path": path, "err": error_text(request, err)})
+                    try:
+                        while (path := proxy.backing_path(path)) and path not in kept:
+                            kept.add(path)
+                    except libvirtError:
+                        # the files it needs are unknown: none of the rest goes
+                        kept.update(to_delete)
             for path in shared:
                 messages.warning(request, _("Disk %(path)s is used by another VM and was kept") % {"path": path})
         return redirect(reverse("instances:index"))

vrtManager/instance.py

@@ -1705,10 +1705,18 @@ class wvmInstance(wvmConnect):
         used_elsewhere = self.paths_used_by_other_domains(own)
         # a kept file keeps the files below it
         for path in [path for path in own if path in used_elsewhere]:
-            while (path := self._backing_path(path)) and path not in used_elsewhere:
+            while (path := self.backing_path(path)) and path not in used_elsewhere:
                 used_elsewhere.add(path)
+        def below(path):
+            """How many files the path grows from: an overlay is deleted before
+            the files it needs, so one that cannot be deleted keeps them."""
+            seen = set()
+            while (path := self.backing_path(path)) and path not in seen:
+                seen.add(path)
+            return len(seen)
+
         return (
-            [path for path in own if path not in used_elsewhere],
+            sorted((path for path in own if path not in used_elsewhere), key=below, reverse=True),
             [path for path in own if path in used_elsewhere],
         )
 
@@ -1739,7 +1747,7 @@ class wvmInstance(wvmConnect):
                 points.update(self._disk_sources(saved).values())
         files = list(tops)
         for path in tops:
-            while (path := self._backing_path(path)) in points and path not in files:
+            while (path := self.backing_path(path)) in points and path not in files:
                 files.append(path)
         grew = True
         while grew:
@@ -1748,14 +1756,14 @@ class wvmInstance(wvmConnect):
                 if top in files:  # reached by a branch in this pass
                     continue
                 branch, path = [top], top
-                while (path := self._backing_path(path)) in points | overlays and path not in files:
+                while (path := self.backing_path(path)) in points | overlays and path not in files:
                     branch.append(path)
                 if path in files:
                     files += branch
                     grew = True
         return files[len(tops):] + memory
 
-    def _backing_path(self, path):
+    def backing_path(self, path):
         """The backing file of a volume; None if it has none or is in no pool.
         Any other error is raised: a chain cut short would let a file a kept
         one needs count as deletable."""
@@ -2026,7 +2034,7 @@ class wvmInstance(wvmConnect):
             # do not keep each other, but a file in use keeps the files below it
             in_use = self.paths_in_use(own=left)
             for path in [path for path in left if path in in_use]:
-                while (path := self._backing_path(path)) and path not in in_use:
+                while (path := self.backing_path(path)) and path not in in_use:
                     in_use.add(path)
             left = [path for path in left if path not in in_use]
         return left

vrtManager/tests/test_cloud_init_seed.py

@@ -91,6 +91,7 @@ class DeleteWithDisksTestCase(unittest.TestCase):
         vm.get_disk_devices = lambda: [{"path": "/pool/web-1.qcow2"}]
         vm._snapshot_chain_files = lambda own: []
         vm.paths_used_by_other_domains = lambda own: set()
+        vm.backing_path = lambda path: None
         pool = MagicMock()
         if seed_in_pool:
             pool.storageVolLookupByName.return_value.path.return_value = f"/pool/cidata-{UUID}.iso"

vrtManager/tests/test_external_snapshots.py

@@ -611,7 +611,7 @@ class LeftBehindTestCase(unittest.TestCase):
     def test_a_file_another_vm_uses_is_not_reported(self):
         # on NFS libvirt could not remove it, but another VM uses it now
         with patch.object(wvmInstance, "paths_in_use", return_value={"/pool/a.one"}) as in_use, \
-                patch.object(wvmInstance, "_backing_path", return_value=None):
+                patch.object(wvmInstance, "backing_path", return_value=None):
             left = self.run_delete({"/pool/a.one", "/pool/vm.one.mem", "/pool/a.two", "/pool/a.qcow2"})
         self.assertEqual(left, ["/pool/vm.one.mem"])
         # their own backing files do not keep each other
@@ -622,7 +622,7 @@ class LeftBehindTestCase(unittest.TestCase):
         # file, and the in-use top's own backing reference does not count
         backing = {"/pool/a.one": "/pool/vm.one.mem"}
         with patch.object(wvmInstance, "paths_in_use", return_value={"/pool/a.one"}), \
-                patch.object(wvmInstance, "_backing_path", side_effect=backing.get):
+                patch.object(wvmInstance, "backing_path", side_effect=backing.get):
             left = self.run_delete({"/pool/a.one", "/pool/vm.one.mem", "/pool/a.two", "/pool/a.qcow2"})
         self.assertEqual(left, [])
 
@@ -774,6 +774,9 @@ class DestroyFilesTestCase(unittest.TestCase):
             backing[top] = "/pool/a.one"
         if extra:
             backing[extra[0]] = extra[1]
+        # the template's own chain, longer than the VM's files
+        template = ["/templates/base.qcow2"] + [f"/templates/base{n}.qcow2" for n in range(1, 10)]
+        backing.update(zip(template, template[1:]))
         if other_overlay_on:
             backing["/pool/other.qcow2"] = other_overlay_on
 
@@ -798,10 +801,10 @@ class DestroyFilesTestCase(unittest.TestCase):
         missing = libvirtError("no volume")
         missing.err = (libvirt.VIR_ERR_NO_STORAGE_VOL,)
         p.get_volume_by_path = MagicMock(side_effect=missing)
-        self.assertIsNone(p._backing_path("/elsewhere/base.qcow2"))  # in no pool
+        self.assertIsNone(p.backing_path("/elsewhere/base.qcow2"))  # in no pool
         p.get_volume_by_path = MagicMock(side_effect=libvirtError("connection lost"))
         with self.assertRaises(libvirtError):
-            p._backing_path("/pool/a.one")
+            p.backing_path("/pool/a.one")
 
     def test_the_snapshot_chain_and_memory_go_but_not_the_template(self):
         delete, keep = self.split()
@@ -822,6 +825,13 @@ class DestroyFilesTestCase(unittest.TestCase):
         )
         self.assertEqual(keep, [])
 
+    def test_an_overlay_is_deleted_before_the_files_it_needs(self):
+        # after a revert the branch a.two comes before a.one: if it cannot be deleted, a.one is still there
+        delete, _ = self.split(top="/pool/a.reverted")
+        for overlay, below in (("/pool/a.two", "/pool/a.one"), ("/pool/a.reverted", "/pool/a.one"),
+                               ("/pool/a.one", "/pool/a.qcow2")):
+            self.assertLess(delete.index(overlay), delete.index(below), (overlay, below))
+
     def test_a_snapshot_file_off_the_disks_chain_stays(self):
         # a disk detached since: its snapshot overlay grows from a file the
         # VM's disks do not