[
https://issues.apache.org/jira/browse/FLINK-20846?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Till Rohrmann closed FLINK-20846.
---------------------------------
Resolution: Fixed
Fixed via
1bbb8b6141635106cb137820b819fedaa8fdfce6
fa49e8458390e8bdb0c179b4c8d8e0e9b55e8596
01e4745283f1031e02aba21b2f91ebcbb87d3acd
224822e8f7197d55d0e711f5efbf3cc43750e31e
1538eb23930fedc5e7e0a1334a15e7ec86c79a06
> Decouple checkpoint services from CheckpointCoordinator
> -------------------------------------------------------
>
> Key: FLINK-20846
> URL: https://issues.apache.org/jira/browse/FLINK-20846
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Coordination
> Affects Versions: 1.13.0
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.13.0
>
>
> In order to reuse checkpoints across different {{ExecutionGraph}} instances
> (w/o HA), we need to decouple the checkpoint services
> {{CompletedCheckpointStore}} and {{CheckpointIDCounter}} from the
> {{CheckpointCoordinator}}. The reason is that the {{CheckpointCoordinator}}
> is bound to the lifetime of an {{ExecutionGraph}} and shuts the services down
> once the {{ExecutionGraph}} reaches a terminal state. Once this is no longer
> the case, it would be possible to create a new {{ExecutionGraph}} which can
> resume from a previous checkpoint if the appropriate
> {{CompletedCheckpointStore}} is provided.
> For future scheduler implementations which want to reuse the
> {{ExecutionGraph}} but need to create different instances this is a
> requirement to support state recovery.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
