[
https://issues.apache.org/jira/browse/FLINK-10205?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16658517#comment-16658517
]
ASF GitHub Bot commented on FLINK-10205:
----------------------------------------
isunjin commented on issue #6684: [FLINK-10205] Batch Job: InputSplit Fault
tolerant for DataSource…
URL: https://github.com/apache/flink/pull/6684#issuecomment-431734273
Great discussion, thanks everybody.
@wenlong88, the scenario you mention is what i try to fix.
[here](https://github.com/isunjin/flink/commit/b61b58d963ea11d34e2eb7ec6f4fe4bfed4dca4a)
is a concrete example, a simple word count job will have data inconsistent
while failover, the job should fail but success with zero output.
@tillrohrmann, **_InputSplitAssigner_** generate a list of _**InputSplit**_,
the order might not matter, but every input should be proceed exactly once, if
a task fail while process a _**InputSplit**_, this _**InputSplit**_ should be
processed again, however, in batch scenario, it might not true,
[this](https://github.com/isunjin/flink/commit/b61b58d963ea11d34e2eb7ec6f4fe4bfed4dca4a)
repro shows that the current codebase doesn't has this logic and thus it has
data inconsistent issue.
Its not a problem in Streaming scenario, as the _**InputSplit**_ will be
treat as a record, eg: in _**ContinuousFileMonitoringFunction**_, it will
collect _**InputSplit**_ and every _**InputSplit**_ will be guaranteed
process exactly once by FLINK, @wenlong88 will this work in your scenario?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> 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
> Affects Versions: 1.6.1, 1.6.2, 1.7.0
> Reporter: JIN SUN
> Assignee: JIN SUN
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.7.0
>
> 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.
> document:
> [https://docs.google.com/document/d/1FdZdcA63tPUEewcCimTFy9Iz2jlVlMRANZkO4RngIuk/edit?usp=sharing]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
