watcher: a name removed from the targets leaves the state (closes #223)

At startup, before the first check, Run removes from the loaded state
the domain, hostname and certificate entries of names no longer in
DNSWATCHER_TARGETS, takes those names off each port entry's list of
names and removes a port entry left with none, so the dashboard,
/api/v1/status and the startup notification count only configured
names. A configured domain's own records, saved as a hostname entry
under its name, are kept. Nothing is notified. Each port check, next to
the removal of stale port entries, now also removes the certificate
entries for an address a name no longer resolves to, except while none
of its nameservers answered, as port entries already were.

Model: opus-5-5

internal/state/state.go

@@ -298,6 +298,27 @@ func (s *State) GetDomainState(
 	return ds, ok
 }
 
+// DeleteDomainState removes a domain state entry.
+func (s *State) DeleteDomainState(domain string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	delete(s.snapshot.Domains, domain)
+}
+
+// GetAllDomainNames returns the names of all domain state entries.
+func (s *State) GetAllDomainNames() []string {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	names := make([]string, 0, len(s.snapshot.Domains))
+	for name := range s.snapshot.Domains {
+		names = append(names, name)
+	}
+
+	return names
+}
+
 // SetHostnameState updates the state for a hostname.
 func (s *State) SetHostnameState(
 	hostname string,
@@ -321,6 +342,28 @@ func (s *State) GetHostnameState(
 	return hs, ok
 }
 
+// DeleteHostnameState removes a hostname state entry.
+func (s *State) DeleteHostnameState(hostname string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	delete(s.snapshot.Hostnames, hostname)
+}
+
+// GetAllHostnames returns the names of all hostname state entries,
+// which include each apex domain's own records.
+func (s *State) GetAllHostnames() []string {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	names := make([]string, 0, len(s.snapshot.Hostnames))
+	for name := range s.snapshot.Hostnames {
+		names = append(names, name)
+	}
+
+	return names
+}
+
 // SetPortState updates the state for a port.
 func (s *State) SetPortState(key string, ps *PortState) {
 	s.mu.Lock()
@@ -383,6 +426,27 @@ func (s *State) GetCertificateState(
 	return cs, ok
 }
 
+// DeleteCertificateState removes a certificate state entry.
+func (s *State) DeleteCertificateState(key string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	delete(s.snapshot.Certificates, key)
+}
+
+// GetAllCertificateKeys returns all certificate state keys.
+func (s *State) GetAllCertificateKeys() []string {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	keys := make([]string, 0, len(s.snapshot.Certificates))
+	for k := range s.snapshot.Certificates {
+		keys = append(keys, k)
+	}
+
+	return keys
+}
+
 // checkDataDirWritable creates the data directory if needed, then writes
 // and removes the temp file that Save uses. It runs at startup so that an
 // unwritable directory stops the process, instead of the process running

internal/watcher/export_test.go

@@ -107,6 +107,11 @@ func (w *Watcher) MaybeSendTestNotification(ctx context.Context) {
 	w.maybeSendTestNotification(ctx)
 }
 
+// CleanupRemovedTargets exports cleanupRemovedTargets for testing.
+func (w *Watcher) CleanupRemovedTargets() {
+	w.cleanupRemovedTargets()
+}
+
 // CheckAllPorts exports checkAllPorts for testing.
 func (w *Watcher) CheckAllPorts(ctx context.Context) {
 	w.checkAllPorts(ctx)

internal/watcher/removed_test.go

@@ -0,0 +1,224 @@
+package watcher_test
+
+import (
+	"maps"
+	"slices"
+	"testing"
+
+	"sneak.berlin/go/dnswatcher/internal/state"
+	"sneak.berlin/go/dnswatcher/internal/watcher"
+)
+
+// TestRemovedTargetsLeaveTheState loads a state saved while a domain
+// and a hostname now removed from the configuration were still in it,
+// and runs the removal that Run does before the first check. The
+// removed names' domain, hostname and certificate entries are gone,
+// the configured names' are kept, and nothing is notified. Nothing is
+// looked up: the watcher has no resolver.
+func TestRemovedTargetsLeaveTheState(t *testing.T) {
+	t.Parallel()
+
+	const (
+		removedDomain = "example.com"
+		removedHost   = "www.example.com"
+	)
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+
+	// The state a check of all four names saves, each name at ip1.
+	for _, name := range []string{domain, removedDomain} {
+		deps.state.SetDomainState(name, &state.DomainState{
+			Nameservers: []string{nsA},
+		})
+	}
+
+	for _, name := range []string{domain, host, removedDomain, removedHost} {
+		deps.state.SetHostnameState(name, saved(
+			map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{"A": {ip1}}),
+			},
+		))
+		deps.state.SetCertificateState(
+			ip1+":443:"+name, &state.CertificateState{Status: "ok"},
+		)
+	}
+
+	err := deps.state.Save()
+	if err != nil {
+		t.Fatalf("saving the state: %v", err)
+	}
+
+	err = deps.state.Load()
+	if err != nil {
+		t.Fatalf("loading the state: %v", err)
+	}
+
+	w.CleanupRemovedTargets()
+
+	snap := deps.state.GetSnapshot()
+
+	got := slices.Sorted(maps.Keys(snap.Domains))
+	if want := []string{domain}; !slices.Equal(got, want) {
+		t.Errorf("domain entries %v, want %v", got, want)
+	}
+
+	got = slices.Sorted(maps.Keys(snap.Hostnames))
+	if want := []string{domain, host}; !slices.Equal(got, want) {
+		t.Errorf("hostname entries %v, want %v", got, want)
+	}
+
+	got = slices.Sorted(maps.Keys(snap.Certificates))
+	if want := []string{
+		ip1 + ":443:" + domain, ip1 + ":443:" + host,
+	}; !slices.Equal(got, want) {
+		t.Errorf("certificate entries %v, want %v", got, want)
+	}
+
+	if sent := deps.notifier.getNotifications(); len(sent) != 0 {
+		t.Errorf("sent %v, want nothing", sent)
+	}
+}
+
+// TestRemovedTargetsLeaveThePortEntries loads a state whose port
+// entries name a domain and a hostname now removed from the
+// configuration, and runs the removal that Run does before the first
+// check. The removed names are off each port entry's list of names, the
+// entry only they had is gone, the entry that also names configured
+// names is kept for the port checks, and nothing is notified.
+func TestRemovedTargetsLeaveThePortEntries(t *testing.T) {
+	t.Parallel()
+
+	const (
+		removedDomain = "example.com"
+		removedHost   = "www.example.com"
+	)
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+
+	// The port 443 entries a check of all four names saves: each name
+	// at ip1, except the removed hostname, at ip2.
+	deps.state.SetPortState(ip1+":443", &state.PortState{
+		Open: true, Hostnames: []string{removedDomain, domain, host},
+	})
+	deps.state.SetPortState(ip2+":443", &state.PortState{
+		Open: true, Hostnames: []string{removedHost},
+	})
+
+	err := deps.state.Save()
+	if err != nil {
+		t.Fatalf("saving the state: %v", err)
+	}
+
+	err = deps.state.Load()
+	if err != nil {
+		t.Fatalf("loading the state: %v", err)
+	}
+
+	w.CleanupRemovedTargets()
+
+	if sent := deps.notifier.getNotifications(); len(sent) != 0 {
+		t.Errorf("sent %v, want nothing", sent)
+	}
+
+	snap := deps.state.GetSnapshot()
+
+	got := slices.Sorted(maps.Keys(snap.Ports))
+	if want := []string{ip1 + ":443"}; !slices.Equal(got, want) {
+		t.Fatalf("port entries %v, want %v", got, want)
+	}
+
+	got = snap.Ports[ip1+":443"].Hostnames
+	if want := []string{domain, host}; !slices.Equal(got, want) {
+		t.Errorf("names of port entry %s:443 %v, want %v", ip1, got, want)
+	}
+}
+
+// TestCertificateStateForAnAddressGone runs the port checks on hostname
+// state built here for a configured hostname, with certificate entries
+// saved for it at ip1, ip2 and an IPv6 address. When its nameservers
+// answered with ip1 and the IPv6 address, the entry for ip2 is removed.
+// When none of them answered, its addresses are not known, and every
+// entry is kept. Nothing is notified, and nothing is looked up.
+func TestCertificateStateForAnAddressGone(t *testing.T) {
+	t.Parallel()
+
+	const ip6 = "2001:db8::1"
+
+	tests := []struct {
+		name     string
+		hostname *state.HostnameState
+		want     []string
+	}{
+		{
+			"answered without ip2",
+			saved(map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{
+					"A": {ip1}, "AAAA": {ip6},
+				}),
+			}),
+			[]string{ip1 + ":443:" + host, ip6 + ":443:" + host},
+		},
+		{
+			"no nameserver answered",
+			saved(map[string]*state.NameserverRecordState{
+				nsA: failed(), nsB: failed(),
+			}),
+			[]string{
+				ip1 + ":443:" + host,
+				ip2 + ":443:" + host,
+				ip6 + ":443:" + host,
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			cfg := defaultTestConfig(t)
+			cfg.Hostnames = []string{host}
+
+			deps := newTestDeps(t, cfg)
+			w := watcher.NewForTest(
+				cfg, deps.state, nil,
+				deps.portChecker, deps.tlsChecker, deps.notifier,
+			)
+			w.SetFirstRun(false)
+
+			deps.state.SetHostnameState(host, tt.hostname)
+
+			for _, ip := range []string{ip1, ip2, ip6} {
+				deps.state.SetCertificateState(
+					ip+":443:"+host, &state.CertificateState{Status: "ok"},
+				)
+			}
+
+			w.CheckAllPorts(t.Context())
+
+			got := slices.Sorted(maps.Keys(deps.state.GetSnapshot().Certificates))
+			if !slices.Equal(got, tt.want) {
+				t.Errorf("certificate entries %v, want %v", got, tt.want)
+			}
+
+			if sent := deps.notifier.getNotifications(); len(sent) != 0 {
+				t.Errorf("sent %v, want nothing", sent)
+			}
+		})
+	}
+}

internal/watcher/watcher.go

@@ -131,6 +131,7 @@ func (w *Watcher) Run(ctx context.Context) {
 		"tlsInterval", w.config.TLSInterval.String(),
 	)
 
+	w.cleanupRemovedTargets()
 	w.RunOnce(ctx)
 	w.maybeSendTestNotification(ctx)
 
@@ -201,6 +202,59 @@ func (w *Watcher) detectFirstRun() {
 	}
 }
 
+// cleanupRemovedTargets removes from the loaded state the domain,
+// hostname and certificate entries of names no longer in the
+// configuration, which changes only at a restart, and takes those names
+// off each port entry's list of names, removing a port entry left with
+// none. Nothing is notified. A configured domain's own records are
+// saved as a hostname entry under its name, which is kept.
+func (w *Watcher) cleanupRemovedTargets() {
+	for _, name := range w.state.GetAllDomainNames() {
+		if !w.isDomain(name) {
+			w.state.DeleteDomainState(name)
+		}
+	}
+
+	for _, name := range w.state.GetAllHostnames() {
+		if !w.isConfigured(name) {
+			w.state.DeleteHostnameState(name)
+		}
+	}
+
+	for _, key := range w.state.GetAllCertificateKeys() {
+		if _, hostname := parseCertKey(key); !w.isConfigured(hostname) {
+			w.state.DeleteCertificateState(key)
+		}
+	}
+
+	for _, key := range w.state.GetAllPortKeys() {
+		ps, ok := w.state.GetPortState(key)
+		if !ok {
+			continue
+		}
+
+		var names []string
+
+		for _, name := range ps.Hostnames {
+			if w.isConfigured(name) {
+				names = append(names, name)
+			}
+		}
+
+		if len(names) == 0 {
+			w.state.DeletePortState(key)
+
+			continue
+		}
+
+		w.state.SetPortState(key, &state.PortState{
+			Open:        ps.Open,
+			Hostnames:   names,
+			LastChecked: ps.LastChecked,
+		})
+	}
+}
+
 // runDNSChecks performs DNS resolution for all configured domains
 // and hostnames, updating state with freshly resolved records.
 // This must complete before port or TLS checks run so those
@@ -861,8 +915,10 @@ func (w *Watcher) checkAllPorts(ctx context.Context) {
 	}
 
 	// Phase 3: Remove port state entries that no longer have
-	// any hostname referencing them.
+	// any hostname referencing them, and certificate entries for
+	// an address their name no longer has.
 	w.cleanupStalePorts(associations)
+	w.cleanupStaleCertificates()
 }
 
 // buildPortAssociations constructs a map from IP:port keys to
@@ -946,11 +1002,45 @@ func (w *Watcher) cleanupStalePorts(
 	}
 }
 
+// cleanupStaleCertificates removes the certificate entries for an
+// address their name no longer resolves to. An entry saved for a name
+// none of whose nameservers answered is kept: that name's addresses are
+// not known, not gone.
+func (w *Watcher) cleanupStaleCertificates() {
+	for _, key := range w.state.GetAllCertificateKeys() {
+		ip, hostname := parseCertKey(key)
+
+		if slices.Contains(w.collectIPs(hostname), ip) ||
+			w.noNameserverAnswered(hostname) {
+			continue
+		}
+
+		w.state.DeleteCertificateState(key)
+	}
+}
+
+// parseCertKey splits an "ip:port:hostname" certificate key into its
+// address and hostname.
+func parseCertKey(key string) (string, string) {
+	lastColon := strings.LastIndex(key, ":")
+	if lastColon < 0 {
+		return "", key
+	}
+
+	ip, _ := parsePortKey(key[:lastColon])
+
+	return ip, key[lastColon+1:]
+}
+
+// isConfigured reports whether name is a configured domain or hostname.
+func (w *Watcher) isConfigured(name string) bool {
+	return w.isDomain(name) || slices.Contains(w.config.Hostnames, name)
+}
+
 // noNameserverAnswered reports whether name is a configured domain or
 // hostname and none of its nameservers answered on its last check.
 func (w *Watcher) noNameserverAnswered(name string) bool {
-	if !slices.Contains(w.config.Hostnames, name) &&
-		!slices.Contains(w.config.Domains, name) {
+	if !w.isConfigured(name) {
 		return false
 	}