automc: pg LISTEN/NOTIFY route source + HTTP waker

Adds opt-in extension package internal/automc/ that: - Subscribes to Postgres notifications on a 'servers' table and pushes route changes into server.Routes (no file I/O, no fsnotify). - Provides a WakerFunc that POSTs to a configurable HTTP control plane (server-manager) and polls until state=running. When AUTOMC_DSN is unset, Wire() is a no-op and the binary behaves exactly like upstream itzg/mc-router. Single patch site in main.go (import + 4-line call) keeps upstream rebases trivial. See docs/AUTOMC.md for env vars and the expected DB schema/trigger.
2026-05-27 11:10:02 +02:00
parent 74d0c40022
commit 657fca325e
10 changed files with 608 additions and 0 deletions
@@ -0,0 +1,133 @@
+package automc
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"time"
+
+	"github.com/itzg/mc-router/server"
+	"github.com/jackc/pgx/v5"
+	"github.com/sirupsen/logrus"
+)
+
+const (
+	notifyChannel = "automc_routes_changed"
+	reconnectMin  = 1 * time.Second
+	reconnectMax  = 30 * time.Second
+)
+
+type route struct {
+	name    string
+	domain  string
+	address string
+}
+
+type syncer struct {
+	dsn         string
+	waker       *wakerConfig
+	current     map[string]route
+}
+
+func newSyncer(dsn string, w *wakerConfig) *syncer {
+	return &syncer{dsn: dsn, waker: w, current: map[string]route{}}
+}
+
+func (s *syncer) run(ctx context.Context) {
+	backoff := reconnectMin
+	for {
+		if ctx.Err() != nil {
+			return
+		}
+		err := s.connectAndLoop(ctx)
+		if err != nil && !errors.Is(err, context.Canceled) {
+			logrus.WithError(err).Warnf("automc pgsync disconnected; reconnecting in %s", backoff)
+		}
+		select {
+		case <-ctx.Done():
+			return
+		case <-time.After(backoff):
+		}
+		backoff *= 2
+		if backoff > reconnectMax {
+			backoff = reconnectMax
+		}
+	}
+}
+
+func (s *syncer) connectAndLoop(ctx context.Context) error {
+	conn, err := pgx.Connect(ctx, s.dsn)
+	if err != nil {
+		return fmt.Errorf("pgx connect: %w", err)
+	}
+	defer conn.Close(context.Background())
+
+	if _, err := conn.Exec(ctx, "LISTEN "+notifyChannel); err != nil {
+		return fmt.Errorf("LISTEN: %w", err)
+	}
+	logrus.Infof("automc pgsync connected; LISTEN %s", notifyChannel)
+
+	if err := s.refresh(ctx, conn); err != nil {
+		return fmt.Errorf("initial refresh: %w", err)
+	}
+
+	for {
+		if _, err := conn.WaitForNotification(ctx); err != nil {
+			return fmt.Errorf("wait notification: %w", err)
+		}
+		if err := s.refresh(ctx, conn); err != nil {
+			return fmt.Errorf("refresh: %w", err)
+		}
+	}
+}
+
+func (s *syncer) refresh(ctx context.Context, conn *pgx.Conn) error {
+	rows, err := conn.Query(ctx, `SELECT name, domain, address FROM servers WHERE domain != '' AND address != ''`)
+	if err != nil {
+		return err
+	}
+	defer rows.Close()
+
+	desired := map[string]route{}
+	for rows.Next() {
+		var r route
+		if err := rows.Scan(&r.name, &r.domain, &r.address); err != nil {
+			return err
+		}
+		desired[r.domain] = r
+	}
+	if err := rows.Err(); err != nil {
+		return err
+	}
+
+	add, del := diff(s.current, desired)
+	s.apply(add, del)
+	s.current = desired
+	return nil
+}
+
+func (s *syncer) apply(add []route, del []string) {
+	for _, host := range del {
+		if server.Routes.DeleteMapping(host) {
+			logrus.Infof("automc route -: %s", host)
+		}
+	}
+	for _, r := range add {
+		server.Routes.CreateMapping(r.domain, r.address, "", s.waker.wakerFor(r.name), nil, "", "")
+		logrus.Infof("automc route +: %s → %s (%s)", r.domain, r.address, r.name)
+	}
+}
+
+func diff(prev, next map[string]route) (add []route, del []string) {
+	for host, r := range next {
+		if p, ok := prev[host]; !ok || p.address != r.address || p.name != r.name {
+			add = append(add, r)
+		}
+	}
+	for host := range prev {
+		if _, ok := next[host]; !ok {
+			del = append(del, host)
+		}
+	}
+	return add, del
+}
@@ -0,0 +1,103 @@
+package automc
+
+import (
+	"sort"
+	"testing"
+)
+
+func TestDiff(t *testing.T) {
+	cases := []struct {
+		name        string
+		prev, next  map[string]route
+		wantAddHost []string
+		wantDel     []string
+	}{
+		{
+			name:        "empty to empty",
+			prev:        map[string]route{},
+			next:        map[string]route{},
+			wantAddHost: nil,
+			wantDel:     nil,
+		},
+		{
+			name:        "add one",
+			prev:        map[string]route{},
+			next:        map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			wantAddHost: []string{"a.example.com"},
+			wantDel:     nil,
+		},
+		{
+			name:        "delete one",
+			prev:        map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			next:        map[string]route{},
+			wantAddHost: nil,
+			wantDel:     []string{"a.example.com"},
+		},
+		{
+			name: "address change",
+			prev: map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			next: map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.2:25565"}},
+			wantAddHost: []string{"a.example.com"},
+			wantDel:     nil,
+		},
+		{
+			name: "name change with same address triggers re-register (waker rebind)",
+			prev: map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			next: map[string]route{"a.example.com": {name: "b", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			wantAddHost: []string{"a.example.com"},
+			wantDel:     nil,
+		},
+		{
+			name: "no change",
+			prev: map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			next: map[string]route{"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"}},
+			wantAddHost: nil,
+			wantDel:     nil,
+		},
+		{
+			name: "mixed add + delete",
+			prev: map[string]route{
+				"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"},
+				"b.example.com": {name: "b", domain: "b.example.com", address: "10.0.0.2:25565"},
+			},
+			next: map[string]route{
+				"a.example.com": {name: "a", domain: "a.example.com", address: "10.0.0.1:25565"},
+				"c.example.com": {name: "c", domain: "c.example.com", address: "10.0.0.3:25565"},
+			},
+			wantAddHost: []string{"c.example.com"},
+			wantDel:     []string{"b.example.com"},
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			add, del := diff(tc.prev, tc.next)
+			gotAdd := make([]string, 0, len(add))
+			for _, r := range add {
+				gotAdd = append(gotAdd, r.domain)
+			}
+			sort.Strings(gotAdd)
+			sort.Strings(del)
+			sort.Strings(tc.wantAddHost)
+			sort.Strings(tc.wantDel)
+			if !equalSlice(gotAdd, tc.wantAddHost) {
+				t.Errorf("add: got %v want %v", gotAdd, tc.wantAddHost)
+			}
+			if !equalSlice(del, tc.wantDel) {
+				t.Errorf("del: got %v want %v", del, tc.wantDel)
+			}
+		})
+	}
+}
+
+func equalSlice(a, b []string) bool {
+	if len(a) != len(b) {
+		return false
+	}
+	for i := range a {
+		if a[i] != b[i] {
+			return false
+		}
+	}
+	return true
+}
@@ -0,0 +1,122 @@
+package automc
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"strings"
+	"time"
+
+	"github.com/itzg/mc-router/server"
+)
+
+const (
+	wakerPollInterval = 2 * time.Second
+	wakerPollTimeout  = 90 * time.Second
+)
+
+var wakerPollIntervalForTest = wakerPollInterval
+
+type wakerConfig struct {
+	baseURL string
+	token   string
+	client  *http.Client
+}
+
+func newWakerConfig(baseURL, token string) *wakerConfig {
+	if baseURL == "" {
+		return nil
+	}
+	return &wakerConfig{
+		baseURL: strings.TrimRight(baseURL, "/"),
+		token:   token,
+		client:  &http.Client{Timeout: 10 * time.Second},
+	}
+}
+
+func (w *wakerConfig) wakerFor(serverName string) server.WakerFunc {
+	if w == nil {
+		return nil
+	}
+	return func(ctx context.Context) (string, error) {
+		if err := w.start(ctx, serverName); err != nil {
+			return "", err
+		}
+		return w.pollUntilRunning(ctx, serverName)
+	}
+}
+
+func (w *wakerConfig) start(ctx context.Context, name string) error {
+	u := fmt.Sprintf("%s/servers/%s/start", w.baseURL, url.PathEscape(name))
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, u, nil)
+	if err != nil {
+		return err
+	}
+	w.setAuth(req)
+	resp, err := w.client.Do(req)
+	if err != nil {
+		return fmt.Errorf("waker start %s: %w", name, err)
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode >= 400 && resp.StatusCode != http.StatusConflict {
+		// 409 = already starting/running, treat as success
+		body, _ := io.ReadAll(io.LimitReader(resp.Body, 512))
+		return fmt.Errorf("waker start %s: %s — %s", name, resp.Status, strings.TrimSpace(string(body)))
+	}
+	return nil
+}
+
+func (w *wakerConfig) pollUntilRunning(ctx context.Context, name string) (string, error) {
+	deadline := time.Now().Add(wakerPollTimeout)
+	ticker := time.NewTicker(wakerPollIntervalForTest)
+	defer ticker.Stop()
+
+	for {
+		state, addr, err := w.queryState(ctx, name)
+		if err == nil && state == "running" && addr != "" {
+			return addr, nil
+		}
+		if time.Now().After(deadline) {
+			return "", fmt.Errorf("waker timeout for %s after %s (last state=%q err=%v)", name, wakerPollTimeout, state, err)
+		}
+		select {
+		case <-ctx.Done():
+			return "", ctx.Err()
+		case <-ticker.C:
+		}
+	}
+}
+
+func (w *wakerConfig) queryState(ctx context.Context, name string) (string, string, error) {
+	u := fmt.Sprintf("%s/servers/%s", w.baseURL, url.PathEscape(name))
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, u, nil)
+	if err != nil {
+		return "", "", err
+	}
+	w.setAuth(req)
+	resp, err := w.client.Do(req)
+	if err != nil {
+		return "", "", err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		return "", "", fmt.Errorf("query state %s: %s", name, resp.Status)
+	}
+	var body struct {
+		State   string `json:"state"`
+		Address string `json:"address"`
+	}
+	if err := json.NewDecoder(resp.Body).Decode(&body); err != nil {
+		return "", "", err
+	}
+	return body.State, body.Address, nil
+}
+
+func (w *wakerConfig) setAuth(req *http.Request) {
+	if w.token != "" {
+		req.Header.Set("X-API-Key", w.token)
+	}
+}
@@ -0,0 +1,110 @@
+package automc
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+)
+
+func TestWakerNilWhenURLEmpty(t *testing.T) {
+	w := newWakerConfig("", "")
+	if w != nil {
+		t.Fatalf("expected nil waker config when URL empty, got %+v", w)
+	}
+	if w.wakerFor("foo") != nil {
+		t.Fatalf("expected nil WakerFunc from nil config")
+	}
+}
+
+func TestWakerStartThenPoll(t *testing.T) {
+	var startCalls int32
+	var pollCalls int32
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Header.Get("X-API-Key") != "secret" {
+			http.Error(w, "no auth", http.StatusUnauthorized)
+			return
+		}
+		switch {
+		case r.Method == http.MethodPost && strings.HasSuffix(r.URL.Path, "/start"):
+			atomic.AddInt32(&startCalls, 1)
+			w.WriteHeader(http.StatusAccepted)
+		case r.Method == http.MethodGet:
+			n := atomic.AddInt32(&pollCalls, 1)
+			state := "starting"
+			addr := ""
+			if n >= 2 {
+				state = "running"
+				addr = "10.0.0.5:25565"
+			}
+			_ = json.NewEncoder(w).Encode(map[string]string{
+				"state":   state,
+				"address": addr,
+			})
+		default:
+			http.NotFound(w, r)
+		}
+	}))
+	defer srv.Close()
+
+	wc := newWakerConfig(srv.URL, "secret")
+	wc.client.Timeout = 2 * time.Second
+
+	// Tighten poll interval for the test only.
+	saved := wakerPollIntervalForTest
+	wakerPollIntervalForTest = 10 * time.Millisecond
+	t.Cleanup(func() { wakerPollIntervalForTest = saved })
+
+	fn := wc.wakerFor("test1")
+	if fn == nil {
+		t.Fatal("expected non-nil WakerFunc")
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	addr, err := fn(ctx)
+	if err != nil {
+		t.Fatalf("wake failed: %v", err)
+	}
+	if addr != "10.0.0.5:25565" {
+		t.Errorf("addr: got %q want 10.0.0.5:25565", addr)
+	}
+	if atomic.LoadInt32(&startCalls) != 1 {
+		t.Errorf("expected 1 start call, got %d", startCalls)
+	}
+	if got := atomic.LoadInt32(&pollCalls); got < 2 {
+		t.Errorf("expected >=2 polls, got %d", got)
+	}
+}
+
+func TestWakerStartHandles409(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Method == http.MethodPost && strings.HasSuffix(r.URL.Path, "/start") {
+			http.Error(w, "already running", http.StatusConflict)
+			return
+		}
+		_ = json.NewEncoder(w).Encode(map[string]string{
+			"state":   "running",
+			"address": "10.0.0.6:25565",
+		})
+	}))
+	defer srv.Close()
+
+	wc := newWakerConfig(srv.URL, "")
+	wakerPollIntervalForTest = 10 * time.Millisecond
+
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+	defer cancel()
+	addr, err := wc.wakerFor("x")(ctx)
+	if err != nil {
+		t.Fatalf("expected 409 to be treated as success, got err: %v", err)
+	}
+	if addr != "10.0.0.6:25565" {
+		t.Errorf("addr: got %q", addr)
+	}
+}
@@ -0,0 +1,24 @@
+// Package automc wires automc-specific extensions onto upstream mc-router.
+//
+// All behavior is opt-in via env vars; when AUTOMC_DSN is unset, Wire is a no-op
+// and the binary behaves exactly like upstream itzg/mc-router.
+package automc
+
+import (
+	"context"
+	"os"
+
+	"github.com/sirupsen/logrus"
+)
+
+func Wire(ctx context.Context) error {
+	dsn := os.Getenv("AUTOMC_DSN")
+	if dsn == "" {
+		return nil
+	}
+	waker := newWakerConfig(os.Getenv("AUTOMC_WAKER_URL"), os.Getenv("AUTOMC_WAKER_TOKEN"))
+	s := newSyncer(dsn, waker)
+	go s.run(ctx)
+	logrus.Info("automc: pg route sync started")
+	return nil
+}