Misc v4 fixes and improved logging (#1831)

* logs for optional services

* print env vars on startup in debug mode

* routes need to explicitly ask to keep connection alive

* log indicators for now

* make workload api listen host configurable

* expose supervisor metrics and make more configurable

* configurable pull secrets, no defaults

* remove restore route

* run controller to handle queued executing

* fix v3 deploys in v4 project

* update admin worker route

* only start pod cleaner et al in k8s mode

* set new worker group as default if none yet

Author: nicktrn

apps/supervisor/README.md

@@ -8,20 +8,15 @@
 api_url=http://localhost:3030
 wg_name=my-worker
 
-# edit these
+# edit this
 admin_pat=tr_pat_...
-project_id=clsw6q8wz...
 
 curl -sS \
     -X POST \
     "$api_url/admin/api/v1/workers" \
     -H "Authorization: Bearer $admin_pat" \
     -H "Content-Type: application/json" \
-    -d "{
-        \"name\": \"$wg_name\",
-        \"makeDefault\": true,
-        \"projectId\": \"$project_id\"
-    }"
+    -d "{\"name\": \"$wg_name\"}"
 ```
 
 2. Create `.env` and set the worker token
@@ -47,3 +42,26 @@ pnpm exec trigger deploy --self-hosted
 # The additional network flag is required on linux
 pnpm exec trigger deploy --self-hosted --network host
 ```
+
+## Additional worker groups
+
+When adding more worker groups you might also want to make them the default for a specific project. This will allow you to test it without having to change the global default:
+
+```sh
+api_url=http://localhost:3030
+wg_name=my-worker
+
+# edit these
+admin_pat=tr_pat_...
+project_id=clsw6q8wz...
+
+curl -sS \
+    -X POST \
+    "$api_url/admin/api/v1/workers" \
+    -H "Authorization: Bearer $admin_pat" \
+    -H "Content-Type: application/json" \
+    -d "{
+        \"name\": \"$wg_name\",
+        \"makeDefaultForProjectId\": \"$project_id\"
+    }"
+```

apps/supervisor/src/env.ts

@@ -26,6 +26,7 @@ const Env = z.object({
     .transform((s) => z.enum(["http", "https"]).parse(s.toLowerCase()))
     .default("http"),
   TRIGGER_WORKLOAD_API_DOMAIN: z.string().optional(), // If unset, will use orchestrator-specific default
+  TRIGGER_WORKLOAD_API_HOST_INTERNAL: z.string().default("0.0.0.0"),
   TRIGGER_WORKLOAD_API_PORT_INTERNAL: z.coerce.number().default(8020), // This is the port the workload API listens on
   TRIGGER_WORKLOAD_API_PORT_EXTERNAL: z.coerce.number().default(8020), // This is the exposed port passed to the run controller
 
@@ -41,6 +42,7 @@ const Env = z.object({
   DOCKER_NETWORK: z.string().default("host"),
   OTEL_EXPORTER_OTLP_ENDPOINT: z.string().url(),
   ENFORCE_MACHINE_PRESETS: z.coerce.boolean().default(false),
+  KUBERNETES_IMAGE_PULL_SECRETS: z.string().optional(), // csv
 
   // Used by the resource monitor
   OVERRIDE_CPU_TOTAL: z.coerce.number().optional(),
@@ -53,7 +55,10 @@ const Env = z.object({
   EPHEMERAL_STORAGE_SIZE_REQUEST: z.string().default("2Gi"),
 
   // Metrics
+  METRICS_ENABLED: BoolEnv.default(true),
   METRICS_COLLECT_DEFAULTS: BoolEnv.default(true),
+  METRICS_HOST: z.string().default("127.0.0.1"),
+  METRICS_PORT: z.coerce.number().int().default(9090),
 
   // Pod cleaner
   POD_CLEANER_ENABLED: BoolEnv.default(true),
@@ -63,6 +68,9 @@ const Env = z.object({
   // Failed pod handler
   FAILED_POD_HANDLER_ENABLED: BoolEnv.default(true),
   FAILED_POD_HANDLER_RECONNECT_INTERVAL_MS: z.coerce.number().int().default(1000),
+
+  // Debug
+  DEBUG: BoolEnv.default(false),
 });
 
 export const env = Env.parse(stdEnv);

apps/supervisor/src/index.ts

@@ -30,7 +30,7 @@ if (env.METRICS_COLLECT_DEFAULTS) {
 
 class ManagedSupervisor {
   private readonly workerSession: SupervisorSession;
-  private readonly httpServer: HttpServer;
+  private readonly metricsServer?: HttpServer;
   private readonly workloadServer: WorkloadServer;
   private readonly workloadManager: WorkloadManager;
   private readonly logger = new SimpleStructuredLogger("managed-worker");
@@ -44,24 +44,15 @@ class ManagedSupervisor {
   private readonly warmStartUrl = env.TRIGGER_WARM_START_URL;
 
   constructor() {
-    const workloadApiProtocol = env.TRIGGER_WORKLOAD_API_PROTOCOL;
-    const workloadApiDomain = env.TRIGGER_WORKLOAD_API_DOMAIN;
-    const workloadApiPortExternal = env.TRIGGER_WORKLOAD_API_PORT_EXTERNAL;
+    const { TRIGGER_WORKER_TOKEN, MANAGED_WORKER_SECRET, ...envWithoutSecrets } = env;
 
-    if (env.POD_CLEANER_ENABLED) {
-      this.podCleaner = new PodCleaner({
-        namespace: env.KUBERNETES_NAMESPACE,
-        batchSize: env.POD_CLEANER_BATCH_SIZE,
-        intervalMs: env.POD_CLEANER_INTERVAL_MS,
-      });
+    if (env.DEBUG) {
+      console.debug("[ManagedSupervisor] Starting up", { envWithoutSecrets });
     }
 
-    if (env.FAILED_POD_HANDLER_ENABLED) {
-      this.failedPodHandler = new FailedPodHandler({
-        namespace: env.KUBERNETES_NAMESPACE,
-        reconnectIntervalMs: env.FAILED_POD_HANDLER_RECONNECT_INTERVAL_MS,
-      });
-    }
+    const workloadApiProtocol = env.TRIGGER_WORKLOAD_API_PROTOCOL;
+    const workloadApiDomain = env.TRIGGER_WORKLOAD_API_DOMAIN;
+    const workloadApiPortExternal = env.TRIGGER_WORKLOAD_API_PORT_EXTERNAL;
 
     if (this.warmStartUrl) {
       this.logger.log("[ManagedWorker] 🔥 Warm starts enabled", {
@@ -70,12 +61,43 @@ class ManagedSupervisor {
     }
 
     if (this.isKubernetes) {
+      if (env.POD_CLEANER_ENABLED) {
+        this.logger.log("[ManagedWorker] 🧹 Pod cleaner enabled", {
+          namespace: env.KUBERNETES_NAMESPACE,
+          batchSize: env.POD_CLEANER_BATCH_SIZE,
+          intervalMs: env.POD_CLEANER_INTERVAL_MS,
+        });
+        this.podCleaner = new PodCleaner({
+          register,
+          namespace: env.KUBERNETES_NAMESPACE,
+          batchSize: env.POD_CLEANER_BATCH_SIZE,
+          intervalMs: env.POD_CLEANER_INTERVAL_MS,
+        });
+      } else {
+        this.logger.warn("[ManagedWorker] Pod cleaner disabled");
+      }
+
+      if (env.FAILED_POD_HANDLER_ENABLED) {
+        this.logger.log("[ManagedWorker] 🔁 Failed pod handler enabled", {
+          namespace: env.KUBERNETES_NAMESPACE,
+          reconnectIntervalMs: env.FAILED_POD_HANDLER_RECONNECT_INTERVAL_MS,
+        });
+        this.failedPodHandler = new FailedPodHandler({
+          register,
+          namespace: env.KUBERNETES_NAMESPACE,
+          reconnectIntervalMs: env.FAILED_POD_HANDLER_RECONNECT_INTERVAL_MS,
+        });
+      } else {
+        this.logger.warn("[ManagedWorker] Failed pod handler disabled");
+      }
+
       this.resourceMonitor = new KubernetesResourceMonitor(createK8sApi(), "");
       this.workloadManager = new KubernetesWorkloadManager({
         workloadApiProtocol,
         workloadApiDomain,
         workloadApiPort: workloadApiPortExternal,
         warmStartUrl: this.warmStartUrl,
+        imagePullSecrets: env.KUBERNETES_IMAGE_PULL_SECRETS?.split(","),
       });
     } else {
       this.resourceMonitor = new DockerResourceMonitor(new Docker());
@@ -224,16 +246,21 @@ class ManagedSupervisor {
       }
     });
 
-    // Used for health checks and metrics
-    this.httpServer = new HttpServer({ port: 8080, host: "0.0.0.0" }).route("/health", "GET", {
-      handler: async ({ reply }) => {
-        reply.text("OK");
-      },
-    });
+    if (env.METRICS_ENABLED) {
+      this.metricsServer = new HttpServer({
+        port: env.METRICS_PORT,
+        host: env.METRICS_HOST,
+        metrics: {
+          register,
+          expose: true,
+        },
+      });
+    }
 
     // Responds to workload requests only
     this.workloadServer = new WorkloadServer({
       port: env.TRIGGER_WORKLOAD_API_PORT_INTERNAL,
+      host: env.TRIGGER_WORKLOAD_API_HOST_INTERNAL,
       workerClient: this.workerSession.httpClient,
       checkpointClient: this.checkpointClient,
     });
@@ -299,13 +326,10 @@ class ManagedSupervisor {
   async start() {
     this.logger.log("[ManagedWorker] Starting up");
 
-    if (this.podCleaner) {
-      await this.podCleaner.start();
-    }
-
-    if (this.failedPodHandler) {
-      await this.failedPodHandler.start();
-    }
+    // Optional services
+    await this.podCleaner?.start();
+    await this.failedPodHandler?.start();
+    await this.metricsServer?.start();
 
     if (env.TRIGGER_WORKLOAD_API_ENABLED) {
       this.logger.log("[ManagedWorker] Workload API enabled", {
@@ -319,21 +343,16 @@ class ManagedSupervisor {
     }
 
     await this.workerSession.start();
-
-    await this.httpServer.start();
   }
 
   async stop() {
     this.logger.log("[ManagedWorker] Shutting down");
-    await this.httpServer.stop();
+    await this.workerSession.stop();
 
-    if (this.podCleaner) {
-      await this.podCleaner.stop();
-    }
-
-    if (this.failedPodHandler) {
-      await this.failedPodHandler.stop();
-    }
+    // Optional services
+    await this.podCleaner?.stop();
+    await this.failedPodHandler?.stop();
+    await this.metricsServer?.stop();
   }
 }
 

apps/supervisor/src/workloadManager/kubernetes.ts

@@ -182,18 +182,15 @@ export class KubernetesWorkloadManager implements WorkloadManager {
     }
   }
 
+  private getImagePullSecrets(): k8s.V1LocalObjectReference[] | undefined {
+    return this.opts.imagePullSecrets?.map((name) => ({ name }));
+  }
+
   get #defaultPodSpec(): Omit<k8s.V1PodSpec, "containers"> {
     return {
       restartPolicy: "Never",
       automountServiceAccountToken: false,
-      imagePullSecrets: [
-        {
-          name: "registry-trigger",
-        },
-        {
-          name: "registry-trigger-failover",
-        },
-      ],
+      imagePullSecrets: this.getImagePullSecrets(),
       nodeSelector: {
         nodetype: "worker-re2",
       },

apps/supervisor/src/workloadManager/types.ts

@@ -5,6 +5,7 @@ export interface WorkloadManagerOptions {
   workloadApiDomain?: string; // If unset, will use orchestrator-specific default
   workloadApiPort: number;
   warmStartUrl?: string;
+  imagePullSecrets?: string[];
 }
 
 export interface WorkloadManager {

apps/supervisor/src/workloadServer/index.ts

@@ -22,6 +22,7 @@ import {
 } from "@trigger.dev/core/v3/workers";
 import { HttpServer, type CheckpointClient } from "@trigger.dev/core/v3/serverOnly";
 import { type IncomingMessage } from "node:http";
+import { register } from "../metrics.js";
 
 // Use the official export when upgrading to [email protected]
 interface DefaultEventsMap {
@@ -121,7 +122,19 @@ export class WorkloadServer extends EventEmitter<WorkloadServerEvents> {
   }
 
   private createHttpServer({ host, port }: { host: string; port: number }) {
-    return new HttpServer({ port, host })
+    return new HttpServer({
+      port,
+      host,
+      metrics: {
+        register,
+        expose: false,
+      },
+    })
+      .route("/health", "GET", {
+        handler: async ({ reply }) => {
+          reply.text("OK");
+        },
+      })
       .route(
         "/api/v1/workload-actions/runs/:runFriendlyId/snapshots/:snapshotFriendlyId/attempts/start",
         "POST",

apps/webapp/app/routes/admin.api.v1.workers.ts

@@ -7,8 +7,7 @@ import { WorkerGroupService } from "~/v3/services/worker/workerGroupService.serv
 const RequestBodySchema = z.object({
   name: z.string().optional(),
   description: z.string().optional(),
-  projectId: z.string().optional(),
-  makeDefault: z.boolean().optional(),
+  makeDefaultForProjectId: z.string().optional(),
 });
 
 export async function action({ request }: ActionFunctionArgs) {
@@ -35,22 +34,21 @@ export async function action({ request }: ActionFunctionArgs) {
 
   try {
     const rawBody = await request.json();
-    const { name, description, projectId, makeDefault } = RequestBodySchema.parse(rawBody ?? {});
+    const { name, description, makeDefaultForProjectId } = RequestBodySchema.parse(rawBody ?? {});
 
     const service = new WorkerGroupService();
     const { workerGroup, token } = await service.createWorkerGroup({
       name,
       description,
     });
 
-    if (makeDefault && projectId) {
+    if (makeDefaultForProjectId) {
       await prisma.project.update({
         where: {
-          id: projectId,
+          id: makeDefaultForProjectId,
         },
         data: {
           defaultWorkerGroupId: workerGroup.id,
-          engine: "V2",
         },
       });
     }