[
https://issues.apache.org/jira/browse/FLINK-3257?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15744752#comment-15744752
]
ASF GitHub Bot commented on FLINK-3257:
---------------------------------------
Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/1668
@senorcarbone I agree, let's fix the multiple checkpoints issue and do the
rest in FLIP-15
The other operators have a pretty simply way of doing this:
- for synchronous checkpointed operators, no need to do anything, the
synchronous part of one checkpoint is over when the next starts (because it is
synchronous ;-))
- for asynchronously checkpointed state, the state backend needs to be
able to hold multiple snapshots, which are saved by multiple background threads
- none of the operators deal with in-flight data, which makes their job
easy
Dealing with in-flight data probably means that you need to open a
ListState for each checkpoint that arrives and add the feed back values to each
state, until that particular checkpoints barrier comes back through the
feedback channel. I think that should be sufficient.
> Add Exactly-Once Processing Guarantees in Iterative DataStream Jobs
> -------------------------------------------------------------------
>
> Key: FLINK-3257
> URL: https://issues.apache.org/jira/browse/FLINK-3257
> Project: Flink
> Issue Type: Improvement
> Reporter: Paris Carbone
> Assignee: Paris Carbone
>
> The current snapshotting algorithm cannot support cycles in the execution
> graph. An alternative scheme can potentially include records in-transit
> through the back-edges of a cyclic execution graph (ABS [1]) to achieve the
> same guarantees.
> One straightforward implementation of ABS for cyclic graphs can work as
> follows along the lines:
> 1) Upon triggering a barrier in an IterationHead from the TaskManager start
> block output and start upstream backup of all records forwarded from the
> respective IterationSink.
> 2) The IterationSink should eventually forward the current snapshotting epoch
> barrier to the IterationSource.
> 3) Upon receiving a barrier from the IterationSink, the IterationSource
> should finalize the snapshot, unblock its output and emit all records
> in-transit in FIFO order and continue the usual execution.
> --
> Upon restart the IterationSource should emit all records from the injected
> snapshot first and then continue its usual execution.
> Several optimisations and slight variations can be potentially achieved but
> this can be the initial implementation take.
> [1] http://arxiv.org/abs/1506.08603
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
