ci: check readiness and the release version after deploy

The ops port serves GET /version, so the smoke test can confirm that the
upgraded build is the one answering.

.github/workflows/cd.yml

@@ -59,6 +59,7 @@ jobs:
         shell: bash
         env:
           KUBECONFIG: ${{ runner.temp }}/k3s-kubeconfig
+          HEAD_SHA: ${{ github.event.workflow_run.head_sha }}
         run: |
           set -euo pipefail
           kubectl port-forward --namespace alisten service/alisten-serve 18090:9090 &
@@ -69,4 +70,11 @@ jobs:
             --retry-delay 2 \
             --retry-connrefused \
             --max-time 10 \
-            http://127.0.0.1:18090/healthz
+            http://127.0.0.1:18090/readyz
+          # CI 构建发布镜像时把 tag+commit 写进版本号。
+          expected="${TAG}+${HEAD_SHA}"
+          actual="$(curl --silent --show-error --fail --max-time 10 http://127.0.0.1:18090/version)"
+          if [[ "$actual" != "$expected" ]]; then
+            printf '线上版本是 %s，期望 %s\n' "$actual" "$expected" >&2
+            exit 1
+          fi

internal/adapter/in/ops/ops.go

@@ -8,7 +8,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus/promhttp"
 )
 
-func NewHandler(ready func() bool, metrics prometheus.Gatherer) http.Handler {
+func NewHandler(ready func() bool, metrics prometheus.Gatherer, version string) http.Handler {
 	mux := http.NewServeMux()
 	mux.HandleFunc("GET /healthz", func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, "ok\n")
@@ -20,6 +20,9 @@ func NewHandler(ready func() bool, metrics prometheus.Gatherer) http.Handler {
 		}
 		_, _ = io.WriteString(w, "ok\n")
 	})
+	mux.HandleFunc("GET /version", func(w http.ResponseWriter, _ *http.Request) {
+		_, _ = io.WriteString(w, version+"\n")
+	})
 	mux.Handle("GET /metrics", promhttp.HandlerFor(metrics, promhttp.HandlerOpts{}))
 	return mux
 }

internal/adapter/in/ops/ops_test.go

@@ -15,7 +15,7 @@ func TestHandler(t *testing.T) {
 	registry := prometheus.NewRegistry()
 	registry.MustRegister(prometheus.NewGauge(prometheus.GaugeOpts{Name: "test_gauge", Help: "测试用指标。"}))
 	var ready atomic.Bool
-	handler := NewHandler(ready.Load, registry)
+	handler := NewHandler(ready.Load, registry, "v1.2.3+abc")
 
 	get := func(path string) *httptest.ResponseRecorder {
 		rec := httptest.NewRecorder()
@@ -28,6 +28,12 @@ func TestHandler(t *testing.T) {
 			t.Errorf("/healthz 状态码 = %d，期望 200", rec.Code)
 		}
 	})
+	t.Run("version 返回构建版本", func(t *testing.T) {
+		rec := get("/version")
+		if body := rec.Body.String(); rec.Code != http.StatusOK || body != "v1.2.3+abc\n" {
+			t.Errorf("/version 状态码 = %d，响应 = %q，期望 200 和 \"v1.2.3+abc\\n\"", rec.Code, body)
+		}
+	})
 	t.Run("metrics 输出注册的指标", func(t *testing.T) {
 		rec := get("/metrics")
 		body, _ := io.ReadAll(rec.Body)

internal/bootstrap/healthcheck_test.go

@@ -12,7 +12,7 @@ import (
 
 func TestHealthcheckFollowsReadiness(t *testing.T) {
 	var ready atomic.Bool
-	server := httptest.NewServer(ops.NewHandler(ready.Load, prometheus.NewRegistry()))
+	server := httptest.NewServer(ops.NewHandler(ready.Load, prometheus.NewRegistry(), version))
 	defer server.Close()
 	t.Setenv("ALISTEN_OPS_LISTEN", server.Listener.Addr().String())
 

internal/bootstrap/serve.go

@@ -182,7 +182,7 @@ func runServe(base *slog.Logger) (err error) {
 
 	opsLogger := base.With("component", "ops")
 	publicServer := newHTTPServer(p.handler, base.With("component", "http"))
-	opsServer := newHTTPServer(recoverPanics(ops.NewHandler(ready, metrics), opsLogger), opsLogger)
+	opsServer := newHTTPServer(recoverPanics(ops.NewHandler(ready, metrics, version), opsLogger), opsLogger)
 	go func() { fatal <- fmt.Errorf("HTTP 服务异常退出：%w", publicServer.Serve(publicListener)) }()
 	go func() { fatal <- fmt.Errorf("HTTP 服务异常退出：%w", opsServer.Serve(opsListener)) }()
 	logger.Info("服务已启动", "listen", publicListener.Addr().String(), "ops_listen", opsListener.Addr().String(), "instance", instance)

internal/bootstrap/worker.go

@@ -275,7 +275,7 @@ func (w mediaWorker) run(ctx context.Context, listener net.Listener) error {
 		}),
 	)
 	opsLogger := w.base.With("component", "ops")
-	opsServer := newHTTPServer(recoverPanics(ops.NewHandler(ready, w.metrics), opsLogger), opsLogger)
+	opsServer := newHTTPServer(recoverPanics(ops.NewHandler(ready, w.metrics, version), opsLogger), opsLogger)
 	go func() { fatal <- fmt.Errorf("运维端口异常退出：%w", opsServer.Serve(listener)) }()
 	w.logger.Info("worker 已启动", "ops_listen", listener.Addr().String(), "instance", w.instance)