Hold a unit an administrator installed whatever its state, and a packaged unit only when the machine uses it (hq ADR 0103)

2026-09-22 19:52:37 +02:00
parent d3f2595968
commit 40e8ea9fda
3 changed files with 103 additions and 5 deletions
@@ -29,6 +29,9 @@ type machine struct {

 type fakeUnit struct {
 	active, enabled string
+	// fragment is where systemd loads the unit from; empty means /etc/systemd/system, where an
+	// administrator installs one.
+	fragment string
 }

 // systemctl answers as systemd does for the units the machine has, and "not-found" for any other.
@@ -41,8 +44,18 @@ func (m *machine) systemctl(args []string) (string, error) {
 	switch args[0] {
 	case "show":
 		if !ok {
+			if len(args) > 2 && strings.Contains(args[2], "FragmentPath") {
+				return "FragmentPath=\n", nil
+			}
 			return "LoadState=not-found\nActiveState=inactive\nType=simple\n", nil
 		}
+		if len(args) > 2 && strings.Contains(args[2], "FragmentPath") {
+			from := u.fragment
+			if from == "" {
+				from = "/etc/systemd/system/" + unit
+			}
+			return "FragmentPath=" + from + "\n", nil
+		}
 		return "LoadState=loaded\nActiveState=" + u.active + "\nType=simple\nRemainAfterExit=no\n", nil
 	case "is-enabled":
 		if !ok {
@@ -617,7 +630,8 @@ func TestAUnitAPackageOnlyShipsIsNotFound(t *testing.T) {
 	// the private network never came up. Found is what the machine runs.
 	dir := t.TempDir()
 	m := &machine{containers: map[string]*fakeContainer{},
-		units: map[string]*fakeUnit{"wg-quick@mesh0.service": {active: "inactive", enabled: "disabled"}}}
+		units: map[string]*fakeUnit{"wg-quick@mesh0.service": {active: "inactive", enabled: "disabled",
+			fragment: "/usr/lib/systemd/system/wg-quick@.service"}}}
 	report, state := applyAdopted(t, adopted(t, untaken("mesh-wireguard.overlay-up"),
 		`{"id":"mesh-wireguard.overlay-up","type":"service","unit":"wg-quick@mesh0.service","state":"running","boot":"enabled"}`),
 		store.State{}, m, dir)
@@ -904,3 +918,43 @@ func TestAVolumeTheRuntimeCannotBeAskedAboutStopsTheContainer(t *testing.T) {
 		t.Fatalf("a volume the runtime could not be asked about did not stop the container: %v", err)
 	}
 }
+
+func TestAUnitSomebodyInstalledIsHeldWhateverStateItIsIn(t *testing.T) {
+	// A predecessor's unit under /etc, deliberately stopped and disabled: starting it would put
+	// back a service somebody took down on purpose (novox/hq ADR 0103).
+	dir := t.TempDir()
+	m := &machine{containers: map[string]*fakeContainer{},
+		units: map[string]*fakeUnit{"hello.service": {active: "inactive", enabled: "disabled",
+			fragment: "/etc/systemd/system/hello.service"}}}
+	report, state := applyAdopted(t, adopted(t, untaken("hello-web.unit"),
+		`{"id":"hello-web.unit","type":"service","unit":"hello.service","state":"running","boot":"enabled"}`),
+		store.State{}, m, dir)
+	if o := outcomeOf(report, "hello-web.unit"); o.Action != "held" {
+		t.Fatalf("a unit an administrator installed was not held: %+v", o)
+	}
+	if m.did("systemctl start") || m.did("systemctl enable") {
+		t.Errorf("a unit somebody had stopped and disabled was started: %v", m.asked)
+	}
+	if _, ok := state.HeldAt("hello-web.unit"); !ok {
+		t.Error("the hold was not recorded")
+	}
+}
+
+func TestAPackagedUnitTheMachineUsesIsStillHeld(t *testing.T) {
+	// The predecessor's own service from a package, running: not the mesh's to restart.
+	dir := t.TempDir()
+	conf := filepath.Join(dir, "hello.conf")
+	m := &machine{containers: map[string]*fakeContainer{},
+		units: map[string]*fakeUnit{"nginx.service": {active: "active", enabled: "enabled",
+			fragment: "/usr/lib/systemd/system/nginx.service"}}}
+	report, _ := applyAdopted(t, adopted(t, untaken("hello-web.conf", "hello-web.unit"),
+		`{"id":"hello-web.conf","type":"file","path":"`+conf+`","content":"x\n"},
+		 {"id":"hello-web.unit","type":"service","unit":"nginx.service","state":"running","boot":"enabled",
+		  "restart-on":["hello-web.conf"]}`), store.State{}, m, dir)
+	if o := outcomeOf(report, "hello-web.unit"); o.Action != "held" {
+		t.Fatalf("a packaged unit the machine runs was not held: %+v", o)
+	}
+	if m.did("systemctl stop") || m.did("systemctl restart") {
+		t.Errorf("the predecessor's service was restarted: %v", m.asked)
+	}
+}