GitHub user StephanEwen opened a pull request:
https://github.com/apache/flink/pull/643
[streaming] New Source and state checkpointing interfaces
These interfaces allow streaming sources/operators to interact with the
state checkpointing in a more precise manner.
The interfaces are so far intended mainly for internal state handling
(sources, window operators).
An open question what state interface we want to expose to the user code
such that, on scale-in and scale-out, the state repartitions consistently with
the data partitions.
- For groupByKey operations, we could scope the state by key.
- For all other operations, state may be repartitioned in a user-defined
manner. This would give no guarantees about alignment with partitions. Think of
a pipeline `partionBy -> map()`. The state accumulated in the `map()` function
is partitioned by a key, but the state entries may not be able to be linked to
a certain key or key-hash value.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/StephanEwen/incubator-flink cp_interfaces
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/643.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 #643
----
commit 76d63d5b03643d3c6e51a882eb9427e421806342
Author: Stephan Ewen <se...@apache.org>
Date: 2015-04-30T20:05:27Z
[streaming] New Source and state checkpointing interfaces that allow
operations to interact with the state checkpointing in a more precise manner.
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
