[
https://issues.apache.org/jira/browse/FLINK-8679?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16372613#comment-16372613
]
ASF GitHub Bot commented on FLINK-8679:
---------------------------------------
Github user StefanRRichter commented on a diff in the pull request:
https://github.com/apache/flink/pull/5518#discussion_r169905515
--- Diff:
flink-state-backends/flink-statebackend-rocksdb/src/main/java/org/apache/flink/contrib/streaming/state/RocksDBKeyedStateBackend.java
---
@@ -207,6 +208,9 @@
/** Unique ID of this backend. */
private UUID backendUID;
+ /** The byte array for namespace serialization in getKeys(). */
+ private final ByteArrayOutputStreamWithPos namespaceOutputStream;
--- End diff --
It feels like this member is in a too broad scope. While this maximizes
caching, I wonder if creating one of there per `getKeys(...)` all is not
cleaner and still efficient enough. What do you think?
> RocksDBKeyedBackend.getKeys(stateName, namespace) doesn't filter data with
> namespace
> ------------------------------------------------------------------------------------
>
> Key: FLINK-8679
> URL: https://issues.apache.org/jira/browse/FLINK-8679
> Project: Flink
> Issue Type: Bug
> Components: State Backends, Checkpointing
> Affects Versions: 1.5.0
> Reporter: Sihua Zhou
> Assignee: Sihua Zhou
> Priority: Blocker
> Fix For: 1.5.0
>
>
> Currently, `RocksDBKeyedBackend.getKeys(stateName, namespace)` is odds. It
> doesn't use the namespace to filter data. And
> `HeapKeyedBackend.getKeys(stateName, namespace)` has done that, I think they
> should be consistent at least.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
