Stefan Richter created FLINK-6480:
-------------------------------------
Summary: Add executor service parameter Snapshotable::snapshot(...)
Key: FLINK-6480
URL: https://issues.apache.org/jira/browse/FLINK-6480
Project: Flink
Issue Type: Improvement
Components: State Backends, Checkpointing
Reporter: Stefan Richter
We could change the way how async snapshots are executed. Currently, the
snapshot method looks like:
{code:java}
RunnableFuture<S> snapshot(
long checkpointId,
long timestamp,
CheckpointStreamFactory streamFactory,
CheckpointOptions checkpointOptions) throws Exception;
{code}
and I would suggest to change this to:
{code:java}
Future<S> snapshot(
long checkpointId,
long timestamp,
CheckpointStreamFactory streamFactory,
CheckpointOptions checkpointOptions,
ExecutorService executor) throws Exception;
{code}
The implication of this change is, that the FutureRunnable can immediately be
scheduled to the provided executor service. This makes it easier to reason
about the Future's lifecycle and to parallelize multiple snapshot tasks in an
operator instance's checkpointing.
The Future's ownership is only transferred once, from the backend to the
executor service, so cancelation on shutdown only either happens already in the
backend or when the executor service is shut down.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
