[ https://issues.apache.org/jira/browse/FLINK-10205?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16591251#comment-16591251 ]
Fabian Hueske commented on FLINK-10205: --------------------------------------- Most jobs are implemented in a way that the split assignment does not affect the semantics of the job. I don't think we should restrict input split assignment to make the small set of jobs with non-deterministic logic behave deterministic. Also, there are a few more sources of non-determinism (order, partitioning) that would need to be removed. The orchestration overhead to make all operations behave determistically is too high. You can implement a custom [InputSplitAssigner|https://github.com/apache/flink/blob/master/flink-core/src/main/java/org/apache/flink/core/io/InputSplitAssigner.java] if you want to have deterministic input split assignment. I would close this issue as "Won't Fix" > Batch Job: InputSplit Fault tolerant for DataSourceTask > ------------------------------------------------------- > > Key: FLINK-10205 > URL: https://issues.apache.org/jira/browse/FLINK-10205 > Project: Flink > Issue Type: Sub-task > Components: JobManager > Reporter: JIN SUN > Priority: Major > Original Estimate: 168h > Remaining Estimate: 168h > > Today DataSource Task pull InputSplits from JobManager to achieve better > performance, however, when a DataSourceTask failed and rerun, it will not get > the same splits as its previous version. this will introduce inconsistent > result or even data corruption. > Furthermore, if there are two executions run at the same time (in batch > scenario), this two executions should process same splits. > we need to fix the issue to make the inputs of a DataSourceTask > deterministic. The propose is save all splits into ExecutionVertex and > DataSourceTask will pull split from there. > -- This message was sent by Atlassian JIRA (v7.6.3#76005)