You need to enable JavaScript to run this app.
优惠活动
大模型
产品
解决方案
定价
更多

在Kubernetes中运行的Spark集群应使用何种Master URL?

问题:Minikube中Spark集群的Master URL配置问题

我在minikube中运行了一个Spark集群,Pod运行状态如下:

$ k get pods
NAME             READY   STATUS    RESTARTS   AGE
spark-master-0   1/1     Running   0          7d
spark-worker-0   1/1     Running   0          7d
spark-worker-1   1/1     Running   0          7d

我已执行端口转发到master节点:

$ k port-forward spark-master-0 8080:8080
Forwarding from 127.0.0.1:8080 -> 8080
Forwarding from [::1]:8080 -> 8080

现在我尝试创建JavaSparkSession,但不知道该用什么Master URL:

SparkConf sparkConf = new SparkConf()
    .setMaster("???")
    .setAppName("test-spark");
SparkSession.builder()
    .config(sparkConf)
    .getOrCreate();

我试过以下几种URL,全都无效:

  • local
  • localhost:8080
  • http://localhost:8080
  • spark://localhost:8080
  • k8s://http://192.168.1.2:8080(参考某回答)
  • k8s://https://kubernetes:443(参考某回答)

所有URL都会抛出这个错误:

Caused by: org.apache.spark.SparkException: Could not parse Master URL: '<MASTER URL FROM LIST ABOVE>'

请问正确的Master URL应该是什么?谢谢解答。


编辑补充
使用URL spark://192.168.58.2:7077时能部分运行,但会出现以下错误:

2023-04-12T14:53:50.220-07:00  INFO 40229 --- [           main] o.s.jetty.server.handler.ContextHandler  : Started o.s.j.s.ServletContextHandler@49c1e294{/,null,AVAILABLE,@Spark}
2023-04-12T14:53:50.255-07:00  INFO 40229 --- [er-threadpool-0] s.d.c.StandaloneAppClient$ClientEndpoint : Connecting to master spark://192.168.58.2:7077...
2023-04-12T14:54:10.265-07:00  INFO 40229 --- [er-threadpool-0] s.d.c.StandaloneAppClient$ClientEndpoint : Connecting to master spark://192.168.58.2:7077...
2023-04-12T14:54:30.264-07:00  INFO 40229 --- [er-threadpool-0] s.d.c.StandaloneAppClient$ClientEndpoint : Connecting to master spark://192.168.58.2:7077...
2023-04-12T14:54:50.272-07:00  WARN 40229 --- [           main] o.a.s.s.c.StandaloneSchedulerBackend     : Application ID is not initialized yet.
2023-04-12T14:54:50.272-07:00 ERROR 40229 --- [on-retry-thread] o.a.s.s.c.StandaloneSchedulerBackend     : Application has been killed. Reason: All masters are unresponsive! Giving up.
2023-04-12T14:54:50.284-07:00  INFO 40229 --- [p-spark-context] o.s.jetty.server.AbstractConnector       : Stopped Spark@7b6141f0{HTTP/1.1, (http/1.1)}{0.0.0.0:4040}
2023-04-12T14:54:50.286-07:00  INFO 40229 --- [p-spark-context] org.apache.spark.ui.SparkUI              : Stopped Spark web UI at http://mgc14dlpj2f2.lan:4040
2023-04-12T14:54:50.289-07:00  INFO 40229 --- [p-spark-context] o.a.s.s.c.StandaloneSchedulerBackend     : Shutting down all executors
...
java.lang.NullPointerException: Cannot invoke "org.apache.spark.rpc.RpcEndpointRef.askSync(Object, scala.reflect.ClassTag)" because the return value of "org.apache.spark.storage.BlockManagerMaster.driverEndpoint()" is null
...

内容的提问来源于stack exchange,提问作者ritratt

相关产品推荐
方舟 Agent Plan

超全模态模型 × Harness 升级,最新支持 Deepseek-V4.1-Flash、GLM-5.3 系列、Doubao-Seedream-5.0-pro、Kimi-K3 (部分), 限时 9.9 元起

最近更新时间:2026.07.25 08:42:39