Flink集群所需的任务槽与作业中使用的最高并行度恰好一样。
但是当我运行JobCounter = 4和2个插槽(2 TM * 1)的WordCount示例作业时,我发现Dispatcher仍然接受该作业并完成了一些任务,但是几分钟后,发生了此错误:
org.apache.flink.runtime.jobmanager.scheduler.NoResourceAvailableException: Could not allocate enough slots within timeout of 300000 ms to run the job. Please make sure that the cluster has enough resources.
如果并行度大于插槽总数,是否可以将我的作业配置为快速失败?
答案 0 :(得分:2)