[ https://issues.apache.org/jira/browse/FLINK-3756?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15257815#comment-15257815 ]
ASF GitHub Bot commented on FLINK-3756: --------------------------------------- Github user aljoscha commented on the pull request: https://github.com/apache/flink/pull/1883#issuecomment-214683500 Also, I think we should rename the task business classes to accurately reflect what they are for. But this can be done once the rest of the state reworking is in. > Introduce state hierarchy in CheckpointCoordinator > -------------------------------------------------- > > Key: FLINK-3756 > URL: https://issues.apache.org/jira/browse/FLINK-3756 > Project: Flink > Issue Type: Sub-task > Components: JobManager > Affects Versions: 1.1.0 > Reporter: Till Rohrmann > Assignee: Till Rohrmann > Fix For: 1.1.0 > > > Currently, the {{CheckpointCoordinator}} stores the state handles for all > tasks in a list. This lack of structure does not allow to easily group all > tasks belonging to the same {{ExecutionJobVertex}} together. With the > introduction of key groups, it will be necessary to store the key-group state > handles of the different tasks grouped together to make them easily > accessible for redistribution. > Therefore, I propose to introduce the notion of a state for task groups which > groups the state for all tasks belonging to the same {{ExecutionJobVertex}} > together. This task group can then also be used to accumulate the key-group > states from the different tasks. -- This message was sent by Atlassian JIRA (v6.3.4#6332)