[
https://issues.apache.org/jira/browse/FLINK-8360?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16332075#comment-16332075
]
ASF GitHub Bot commented on FLINK-8360:
---------------------------------------
Github user StefanRRichter commented on a diff in the pull request:
https://github.com/apache/flink/pull/5239#discussion_r162584893
--- Diff:
flink-streaming-java/src/test/java/org/apache/flink/streaming/api/operators/StreamOperatorSnapshotRestoreTest.java
---
@@ -98,7 +101,26 @@ public Integer getKey(Integer value) throws Exception {
TypeInformation.of(Integer.class),
MAX_PARALLELISM,
1 /* num subtasks */,
- 0 /* subtask index */);
+ 0 /* subtask index */) {
+
+ @Override
+ protected StreamTaskStateManager
createStreamTaskStateManager(
+ Environment env,
+ StateBackend stateBackend,
+ ProcessingTimeService processingTimeService) {
+
+ return new StreamTaskStateManagerImpl(env,
stateBackend, processingTimeService) {
+ @Override
+ protected <K>
InternalTimeServiceManager<?, K> internalTimeServiceManager(
+ AbstractKeyedStateBackend<K>
keyedStatedBackend,
+ KeyContext keyContext,
+
Iterable<KeyGroupStatePartitionStreamProvider> rawKeyedStates) throws Exception
{
+
+ return null;
--- End diff --
Yes, this is a workaround, because this is a dedicated test for raw keyed
states - which is otherwise consumed immediately to restore a timer service.
This will fail because we wrote only test data to the stream, not the data for
a timer service.
> Implement task-local state recovery
> -----------------------------------
>
> Key: FLINK-8360
> URL: https://issues.apache.org/jira/browse/FLINK-8360
> Project: Flink
> Issue Type: New Feature
> Components: State Backends, Checkpointing
> Reporter: Stefan Richter
> Assignee: Stefan Richter
> Priority: Major
> Fix For: 1.5.0
>
>
> This issue tracks the development of recovery from task-local state. The main
> idea is to have a secondary, local copy of the checkpointed state, while
> there is still a primary copy in DFS that we report to the checkpoint
> coordinator.
> Recovery can attempt to restore from the secondary local copy, if available,
> to save network bandwidth. This requires that the assignment from tasks to
> slots is as sticky is possible.
> For starters, we will implement this feature for all managed keyed states and
> can easily enhance it to all other state types (e.g. operator state) later.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
