Merge pull request #8 from a645162/dev

fix(GpuTask): Fix single GPU bug
a645162 · Nov 7, 2024 · b0acb0b · b0acb0b
2 parents 955d52c + 0929a8e
commit b0acb0b
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 5 deletions.
diff --git a/src/main/kotlin/com/khm/group/center/controller/api/client/task/GpuTaskController.kt b/src/main/kotlin/com/khm/group/center/controller/api/client/task/GpuTaskController.kt
@@ -48,16 +48,23 @@ class GpuTaskController {
         // Notify
         val machineConfig = MachineConfig.getMachineByNameEng(gpuTaskInfo.serverNameEng)
 
+        println(
+            "Receive task from nvi-notify" +
+                    " ${gpuTaskInfo.taskType}" +
+                    " Project:${gpuTaskInfo.projectName}" +
+                    " User:${gpuTaskInfo.taskUser}"
+        )
+
         val gpuTaskNotify = GpuTaskNotify(
             gpuTaskInfo = gpuTaskInfo,
             machineConfig = machineConfig
         )
 
+        val isMultiCard = gpuTaskInfo.multiDeviceWorldSize > 1
+
         if (
-            !gpuTaskInfo.isDebugMode && (
-                    gpuTaskInfo.multiDeviceWorldSize > 1 &&
-                            gpuTaskInfo.multiDeviceLocalRank == 0
-                    )
+            !gpuTaskInfo.isDebugMode &&
+            (!isMultiCard || gpuTaskInfo.multiDeviceLocalRank == 0)
         ) {
             gpuTaskNotify.sendTaskMessage()
         }

diff --git a/src/main/resources/application.yml b/src/main/resources/application.yml
@@ -13,7 +13,7 @@ spring:
     username: root
     password: 123456
     driver-class-name: com.mysql.cj.jdbc.Driver
-    url: jdbc:mysql://localhost:3306/group_center?useUnicode=true&characterEncoding=utf-8&useSSL=false&allowPublicKeyRetrieval=true&serverTimezone=Asia/Shanghai
+    url: jdbc:mysql://localhost:15096/group_center?useUnicode=true&characterEncoding=utf-8&useSSL=false&allowPublicKeyRetrieval=true&serverTimezone=Asia/Shanghai
     type: com.alibaba.druid.pool.DruidDataSource
 
   sql: