Phase 5 Batch 2: 4 RM Tool + rm 客户端 + 日志降级链

- internal/rm/client.go: YARN RM 客户端 - ListApps / GetApp / KillApp / GetLogs - GetLogs 降级链数据驱动 (amContainerLogs → aggregated-logs → logs) - 跨主机 redirect 保留 Authorization (URLAllowlist 兜底) - 4 Tool: list_applications / get_application_status / get_application_logs / kill_application - mcp.WithEnum(state) 约束 YARN app 状态 - tool handler 永远 (result, nil), 业务错误用 NewToolResultError - deps.go: +HTTPClient + MaxResponseBytes - main.go: 注入 httpclient.Client - 端到端实测: mock RM 验证 4 Tool + 降级链 + 错误路径 Co-Authored-By: Claude <noreply@anthropic.com>
2026-07-10 16:50:27 +08:00
parent 705665745e
commit c4ac3cc354
9 changed files with 885 additions and 0 deletions
@@ -0,0 +1,131 @@
+package rm
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"strings"
+
+	"spark-mcp-go/internal/cluster"
+	"spark-mcp-go/internal/httpclient"
+)
+
+// Client is a YARN ResourceManager REST API client bound to a single cluster.
+type Client struct {
+	hc      *httpclient.Client
+	cluster *cluster.Cluster
+}
+
+// New creates a ResourceManager client bound to c.
+func New(hc *httpclient.Client, c *cluster.Cluster) *Client {
+	return &Client{hc: hc, cluster: c}
+}
+
+// ListApps returns raw JSON from GET /ws/v1/cluster/apps[?state=&user=].
+func (r *Client) ListApps(ctx context.Context, state, user string) (json.RawMessage, error) {
+	base := r.cluster.RMURL + "/ws/v1/cluster/apps"
+	u, err := url.Parse(base)
+	if err != nil {
+		return nil, err
+	}
+	q := u.Query()
+	if state != "" {
+		q.Set("state", state)
+	}
+	if user != "" {
+		q.Set("user", user)
+	}
+	u.RawQuery = q.Encode()
+	return r.do(ctx, http.MethodGet, u.String(), nil)
+}
+
+// GetApp returns raw JSON from GET /ws/v1/cluster/apps/{id}.
+func (r *Client) GetApp(ctx context.Context, appID string) (json.RawMessage, error) {
+	return r.do(ctx, http.MethodGet, r.cluster.RMURL+"/ws/v1/cluster/apps/"+appID, nil)
+}
+
+// KillApp PUT {"state":"KILLED"} to /ws/v1/cluster/apps/{id}/state.
+func (r *Client) KillApp(ctx context.Context, appID string) (json.RawMessage, error) {
+	body := `{"state":"KILLED"}`
+	return r.do(ctx, http.MethodPut, r.cluster.RMURL+"/ws/v1/cluster/apps/"+appID+"/state",
+		strings.NewReader(body))
+}
+
+// GetLogs walks a fallback chain to retrieve YARN application logs:
+//  1. GET /ws/v1/cluster/apps/{id}/amContainerLogs (RM often 307 to a NodeManager)
+//  2. GET /ws/v1/cluster/apps/{id}/aggregated-logs?container=...
+//  3. GET /ws/v1/cluster/apps/{id}/logs
+//
+// The first successful endpoint wins.  The returned source names the endpoint
+// that produced the body.
+func (r *Client) GetLogs(ctx context.Context, appID, container string) (body []byte, source string, err error) {
+	endpoints := []struct{ path, name string }{
+		{fmt.Sprintf("/ws/v1/cluster/apps/%s/amContainerLogs", appID), "amContainerLogs"},
+		{fmt.Sprintf("/ws/v1/cluster/apps/%s/aggregated-logs?container=%s", appID, url.QueryEscape(container)), "aggregated-logs"},
+		{fmt.Sprintf("/ws/v1/cluster/apps/%s/logs", appID), "logs"},
+	}
+	var lastErr error
+	for _, ep := range endpoints {
+		b, e := r.doBytes(ctx, http.MethodGet, r.cluster.RMURL+ep.path, nil)
+		if e == nil {
+			return b, ep.name, nil
+		}
+		lastErr = e
+	}
+	return nil, "", fmt.Errorf("rm: all log endpoints failed: %w", lastErr)
+}
+
+// do executes a single HTTP request with auth and allowed-host exemptions,
+// returning the validated JSON body.
+func (r *Client) do(ctx context.Context, method, rawURL string, body io.Reader) (json.RawMessage, error) {
+	b, err := r.doBytes(ctx, method, rawURL, body)
+	if err != nil {
+		return nil, err
+	}
+	raw := json.RawMessage(b)
+	if !json.Valid(raw) {
+		return nil, fmt.Errorf("rm: invalid JSON response from %s", rawURL)
+	}
+	return raw, nil
+}
+
+func (r *Client) doBytes(ctx context.Context, method, rawURL string, body io.Reader) ([]byte, error) {
+	req, err := http.NewRequestWithContext(ctx, method, rawURL, body)
+	if err != nil {
+		return nil, err
+	}
+	if err := httpclient.ApplyAuth(req, "", r.cluster); err != nil {
+		return nil, err
+	}
+	resp, err := r.hc.DoWithRedirect(ctx, req, 5, r.allowedHosts()...)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode >= 400 {
+		return nil, fmt.Errorf("rm: %s %s returned %d", method, rawURL, resp.StatusCode)
+	}
+	return io.ReadAll(resp.Body)
+}
+
+func (r *Client) allowedHosts() []string {
+	hosts := []string{extractHost(r.cluster.RMURL)}
+	for _, pat := range r.cluster.URLAllowlist {
+		if pat == "" {
+			continue
+		}
+		hosts = append(hosts, pat)
+	}
+	return hosts
+}
+
+func extractHost(u string) string {
+	parsed, err := url.Parse(u)
+	if err != nil {
+		return ""
+	}
+	return parsed.Hostname()
+}
@@ -0,0 +1,238 @@
+package rm
+
+import (
+	"context"
+	"encoding/base64"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	"spark-mcp-go/internal/cluster"
+	"spark-mcp-go/internal/httpclient"
+)
+
+func testClient(srv *httptest.Server, c *cluster.Cluster) *Client {
+	hc := httpclient.New(httpclient.Config{Timeout: 5 * time.Second, MaxResponseBytes: 1 << 20})
+	return New(hc, c)
+}
+
+func TestListApps(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path != "/ws/v1/cluster/apps" {
+			t.Errorf("path=%q, want /ws/v1/cluster/apps", r.URL.Path)
+		}
+		w.Header().Set("Content-Type", "application/json")
+		w.Write([]byte(`{"apps":{"app":[]}}`))
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL}
+	raw, err := testClient(srv, cl).ListApps(context.Background(), "", "")
+	if err != nil {
+		t.Fatalf("ListApps: %v", err)
+	}
+	if string(raw) != `{"apps":{"app":[]}}` {
+		t.Errorf("body=%q", string(raw))
+	}
+}
+
+func TestListApps_WithQuery(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		q := r.URL.Query()
+		if q.Get("state") != "RUNNING" {
+			t.Errorf("state=%q, want RUNNING", q.Get("state"))
+		}
+		if q.Get("user") != "yarn" {
+			t.Errorf("user=%q, want yarn", q.Get("user"))
+		}
+		w.Header().Set("Content-Type", "application/json")
+		w.Write([]byte(`{"apps":{"app":[{"id":"app_1"}]}}`))
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL}
+	raw, err := testClient(srv, cl).ListApps(context.Background(), "RUNNING", "yarn")
+	if err != nil {
+		t.Fatalf("ListApps: %v", err)
+	}
+	if !strings.Contains(string(raw), "app_1") {
+		t.Errorf("body=%q", string(raw))
+	}
+}
+
+func TestGetApp(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path != "/ws/v1/cluster/apps/app_123" {
+			t.Errorf("path=%q", r.URL.Path)
+		}
+		w.Header().Set("Content-Type", "application/json")
+		w.Write([]byte(`{"app":{"id":"app_123","state":"RUNNING"}}`))
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL}
+	raw, err := testClient(srv, cl).GetApp(context.Background(), "app_123")
+	if err != nil {
+		t.Fatalf("GetApp: %v", err)
+	}
+	if !strings.Contains(string(raw), "RUNNING") {
+		t.Errorf("body=%q", string(raw))
+	}
+}
+
+func TestKillApp(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path != "/ws/v1/cluster/apps/app_123/state" {
+			t.Errorf("path=%q", r.URL.Path)
+		}
+		if r.Method != http.MethodPut {
+			t.Errorf("method=%q, want PUT", r.Method)
+		}
+		body, _ := io.ReadAll(r.Body)
+		if string(body) != `{"state":"KILLED"}` {
+			t.Errorf("body=%q", string(body))
+		}
+		w.Header().Set("Content-Type", "application/json")
+		w.Write([]byte(`{"state":"KILLED"}`))
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL}
+	raw, err := testClient(srv, cl).KillApp(context.Background(), "app_123")
+	if err != nil {
+		t.Fatalf("KillApp: %v", err)
+	}
+	if string(raw) != `{"state":"KILLED"}` {
+		t.Errorf("body=%q", string(raw))
+	}
+}
+
+func TestGetLogs_PrimarySuccess(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/ws/v1/cluster/apps/app_123/amContainerLogs" {
+			w.Write([]byte("am logs"))
+			return
+		}
+		t.Errorf("unexpected path %q", r.URL.Path)
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL}
+	body, source, err := testClient(srv, cl).GetLogs(context.Background(), "app_123", "container_1")
+	if err != nil {
+		t.Fatalf("GetLogs: %v", err)
+	}
+	if source != "amContainerLogs" {
+		t.Errorf("source=%q, want amContainerLogs", source)
+	}
+	if string(body) != "am logs" {
+		t.Errorf("body=%q", string(body))
+	}
+}
+
+func TestGetLogs_PrimaryRedirects(t *testing.T) {
+	srvNM := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		auth := r.Header.Get("Authorization")
+		if auth == "" {
+			t.Error("Authorization header missing after redirect")
+		}
+		w.Write([]byte("nm logs"))
+	}))
+	defer srvNM.Close()
+
+	srvRM := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/ws/v1/cluster/apps/app_123/amContainerLogs" {
+			w.Header().Set("Location", srvNM.URL+"/node/containerlogs/container_1/root")
+			w.WriteHeader(http.StatusTemporaryRedirect)
+			return
+		}
+		t.Errorf("unexpected RM path %q", r.URL.Path)
+	}))
+	defer srvRM.Close()
+
+	nmHost := srvNM.Listener.Addr().String()
+	cl := &cluster.Cluster{
+		RMURL:        srvRM.URL,
+		AuthType:     cluster.AuthBasic,
+		AuthUsername: "u",
+		AuthPassword: "p",
+		URLAllowlist: []string{nmHost},
+	}
+	body, source, err := testClient(srvRM, cl).GetLogs(context.Background(), "app_123", "container_1")
+	if err != nil {
+		t.Fatalf("GetLogs: %v", err)
+	}
+	if source != "amContainerLogs" {
+		t.Errorf("source=%q, want amContainerLogs", source)
+	}
+	if string(body) != "nm logs" {
+		t.Errorf("body=%q", string(body))
+	}
+}
+
+func TestGetLogs_FallbackToAggregated(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		switch r.URL.Path {
+		case "/ws/v1/cluster/apps/app_123/amContainerLogs":
+			w.WriteHeader(http.StatusInternalServerError)
+		case "/ws/v1/cluster/apps/app_123/aggregated-logs":
+			if r.URL.Query().Get("container") != "container_1" {
+				t.Errorf("container=%q, want container_1", r.URL.Query().Get("container"))
+			}
+			w.Write([]byte("aggregated logs"))
+		default:
+			t.Errorf("unexpected path %q", r.URL.Path)
+		}
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL}
+	body, source, err := testClient(srv, cl).GetLogs(context.Background(), "app_123", "container_1")
+	if err != nil {
+		t.Fatalf("GetLogs: %v", err)
+	}
+	if source != "aggregated-logs" {
+		t.Errorf("source=%q, want aggregated-logs", source)
+	}
+	if string(body) != "aggregated logs" {
+		t.Errorf("body=%q", string(body))
+	}
+}
+
+func TestApplyAuth_SimpleUserName(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if got := r.URL.Query().Get("user.name"); got != "alice" {
+			t.Errorf("user.name=%q, want alice", got)
+		}
+		w.Write([]byte(`{"apps":{"app":[]}}`))
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL, AuthType: cluster.AuthSimple, AuthUsername: "alice"}
+	_, err := testClient(srv, cl).ListApps(context.Background(), "", "")
+	if err != nil {
+		t.Fatalf("ListApps: %v", err)
+	}
+}
+
+func TestApplyAuth_BasicHeader(t *testing.T) {
+	captured := ""
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		captured = r.Header.Get("Authorization")
+		w.Write([]byte(`{"apps":{"app":[]}}`))
+	}))
+	defer srv.Close()
+
+	cl := &cluster.Cluster{RMURL: srv.URL, AuthType: cluster.AuthBasic, AuthUsername: "u", AuthPassword: "p"}
+	_, err := testClient(srv, cl).ListApps(context.Background(), "", "")
+	if err != nil {
+		t.Fatalf("ListApps: %v", err)
+	}
+	want := "Basic " + base64.StdEncoding.EncodeToString([]byte("u:p"))
+	if captured != want {
+		t.Errorf("Authorization=%q, want %q", captured, want)
+	}
+}