fix: 修复 ai 模块的 CI 构建错误 (#1500)

Author: ordYang668

.changeset/fluffy-crabs-decide.md

@@ -0,0 +1,5 @@
+---
+"@scow/ai": patch
+---
+
+Fix: resolve CI build error in @scow/ai module

apps/ai/src/server/setup/jobShell.ts

@@ -12,6 +12,7 @@
 
 import { asyncClientCall } from "@ddadaal/tsgrpc-client";
 import * as k8sClient from "@kubernetes/client-node";
+import { JobInfo_PodStatus } from "@scow/ai-scheduler-adapter-protos/build/protos/job";
 import { normalizePathnameWithQuery } from "@scow/utils";
 import { IncomingMessage } from "http";
 import { NextApiRequest } from "next";
@@ -189,8 +190,16 @@ wss.on("connection", async (ws: AliveCheckedWebSocket, req) => {
     return;
   }
 
-  const namespace = job.containerJobInfo?.namespace;
-  const podName = job.containerJobInfo?.podName;
+  // 获取运行中pod的namespace，podName
+  const runningPod = job.pods.find((pod) => pod.podStatus === JobInfo_PodStatus.RUNNING);
+
+  if (!runningPod) {
+    log("[shell] No running pod found for this job.");
+    ws.close(1008, "No running pod found for this job.");
+    return;
+  }
+
+  const { namespace, podName } = runningPod;
 
   if (!namespace || !podName) {
     log("[shell] Namespace or pod not obtained, please check the adapter version");

apps/ai/src/server/trpc/route/jobs/apps.ts

@@ -12,7 +12,7 @@
 
 import { asyncClientCall } from "@ddadaal/tsgrpc-client";
 import { ServiceError } from "@ddadaal/tsgrpc-common";
-import { JobInfo } from "@scow/ai-scheduler-adapter-protos/build/protos/job";
+import { JobInfo, JobInfo_PodStatus } from "@scow/ai-scheduler-adapter-protos/build/protos/job";
 import { AppType } from "@scow/config/build/appForAi";
 import { getPlaceholderKeys } from "@scow/lib-config/build/parse";
 import { OperationResult, OperationType } from "@scow/lib-operation-log";
@@ -502,6 +502,7 @@ export const createAppSession = procedure
         // 用户指定应用工作目录，如果不存在，则默认为用户的appJobsDirectory
         workingDirectory: workingDirectory ?? join(homeDir, appJobsDirectory),
         script: remoteEntryPath,
+        envVariables: [],
         // 对于AI模块，需要传递的额外参数
         // 第一个参数确定是创建应用or训练任务，
         // 第二个参数为创建应用时的appId
@@ -686,8 +687,18 @@ export const saveImage =
           });
         }
 
-        const nodeName = job.containerJobInfo?.nodeName;
-        const containerId = job.containerJobInfo?.containerId;
+        // 查找运行中的Pod
+        const runningPod = job.pods.find((pod) => pod.podStatus === JobInfo_PodStatus.RUNNING);
+
+        if (!runningPod) {
+          throw new TRPCError({
+            code: "NOT_FOUND",
+            message: "No running pod found for this job",
+          });
+        }
+
+        const nodeName = runningPod.nodeName;
+        const containerId = runningPod.containerId;
 
         if (!nodeName || !containerId) {
           throw new TRPCError({

apps/ai/src/server/trpc/route/jobs/jobs.ts

@@ -256,6 +256,7 @@ procedure
             // 如果nodeCount不为1但同时选定镜像又没有框架标签，该接口会报错
             (nodeCount === 1 && !gpuType?.startsWith("huawei.com")) ? "" : framework || "",
           ],
+          envVariables:[],
           psNodeCount:psNodes,
           workerNodeCount:workerNodes,
         }).catch((e) => {