[jira] [Commented] (FLINK-4809) Operators should tolerate checkpoint failures

["ASF GitHub Bot (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-4809?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16214831#comment-16214831 ] 

ASF GitHub Bot commented on FLINK-4809:
---------------------------------------

GitHub user StefanRRichter opened a pull request:

    https://github.com/apache/flink/pull/4883

    [FLINK-4809] Operators should tolerate checkpoint failures.

    ## What is the purpose of the change
    
    This PR implements FLINK-4809 and allows tasks to tolerate errors that happen in their checkpointing procedure. This tolerance is optional and can activated through `CheckpointOptions`,  the default behaviour is to fail the task, as before this PR.
    
    When fault tolerance is activated, an error will not fail the task. Instead, the task will decline the checkpoint to the checkpoint coordinator.
    
    ## Brief change log
    
      - *Feature configuration through `CheckpointOptions` and `ExecutionConfig`.*
      - *Introduced `CheckpointExceptionHandler` and integrated with StreamTask. This handler will either transform the exception to a `declineCheckpoint` or rethrow to fail the task. *
      - *Tests for configuration forwarding and functionality.*
    
    ## Verifying this change
    
    This change added tests and can be verified as follows:
    
      - *`CheckpointExceptionHandlerTest`: Unit test for the implementations of `CheckpointExceptionHandlerTest` and their factory.*
      - *`TaskCheckpointingBehaviourTest`: Tests the functionality to either fail tasks or only decline checkpoints in case of failure.*
      - *`CheckpointExceptionHandlerConfigurationTest`: Checks everything about configuration and configuration forwarding for the feature.*
    
    ## Does this pull request potentially affect one of the following parts:
    
      - Dependencies (does it add or upgrade a dependency): (no)
      - The public API, i.e., is any changed class annotated with `@Public(Evolving)`: (yes)
      - The serializers: (no)
      - The runtime per-record code paths (performance sensitive): (no)
      - Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes)
    
    ## Documentation
    
      - Does this pull request introduce a new feature? (yes)
      - If yes, how is the feature documented? (docs and JavaDocs)

You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/StefanRRichter/flink decline-on-checkpoint-exception

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/flink/pull/4883.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #4883
    
----
commit 9156a70319f4227b2a07edd9b07dae5fcfbaa01c
Author: Stefan Richter <s....@data-artisans.com>
Date:   2017-10-20T08:59:45Z

    [FLINK-4809] Operators should tolerate checkpoint failures.

----


> Operators should tolerate checkpoint failures
> ---------------------------------------------
>
>                 Key: FLINK-4809
>                 URL: https://issues.apache.org/jira/browse/FLINK-4809
>             Project: Flink
>          Issue Type: Sub-task
>          Components: State Backends, Checkpointing
>            Reporter: Stephan Ewen
>            Assignee: Stefan Richter
>             Fix For: 1.4.0
>
>
> Operators should try/catch exceptions in the synchronous and asynchronous part of the checkpoint and send a {{DeclineCheckpoint}} message as a result.
> The decline message should have the failure cause attached to it.
> The checkpoint barrier should be sent anyways as a first step before attempting to make a state checkpoint, to make sure that downstream operators do not block in alignment.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)