[
https://issues.apache.org/jira/browse/FLINK-3761?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15288838#comment-15288838
]
ASF GitHub Bot commented on FLINK-3761:
---------------------------------------
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1988#issuecomment-220002453
@gyfora Yes, this is correct. We'll have to put the timers into partitioned
state and upon restarting iterate over the saved timers to re-set any callbacks
at the `StreamTask`.
Doing this naively would slow things down. What I had in mind is to still
keep the timers in an in-memory `Map` but before snapshotting put them into a
partitioned state. This way, we don't have the overhead every time that we deal
with timers but just when checkpointing.
> Introduce key group state backend
> ---------------------------------
>
> Key: FLINK-3761
> URL: https://issues.apache.org/jira/browse/FLINK-3761
> Project: Flink
> Issue Type: Sub-task
> Components: state backends
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
>
> After an off-line discussion with [~aljoscha], we came to the conclusion that
> it would be beneficial to reflect the differences between a keyed and a
> non-keyed stream also in the state backends. A state backend which is used
> for a keyed stream offers a value, list, folding and value state and has to
> group its keys into key groups.
> A state backend for non-keyed streams can only offer a union state to make it
> work with dynamic scaling. A union state is a state which is broadcasted to
> all tasks in case of a recovery. The state backends can then select what
> information they need to recover from the whole state (formerly distributed).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
