[
https://issues.apache.org/jira/browse/FLINK-33317?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Rui Fan updated FLINK-33317:
----------------------------
Description:
See FLINK-33315 for details.
This Jira focus on avoid unnecessary memory usage, it can optimize the memory
cost of Replica_3 in FLINK-33315.
Solution:
* After getStreamOperatorFactory, the StreamConfig#SERIALIZEDUDF can be
cleaned
In general, we don't clean any configuration. However, the SERIALIZED_UDF may
be very large when operator includes some large objects.
{@link #getStreamOperatorFactory}
is used to create a StreamOperator and usually only needs to be called once.
Callers can clean it to reduce the memory after calling the \{@link
#getStreamOperatorFactory}.
was:
See FLINK-33315 for details.
This Jira focus on avoid unnecessary memory usage, it can optimize the memory
cost of Replica_3 in FLINK-33315.
Solution:
* Define a threshold for cleaning mechanism of StreamConfig#SERIALIZEDUDF
* After getStreamOperatorFactory, the StreamConfig#SERIALIZEDUDF can be
cleaned
In general, we don't clean any configuration. However, the SERIALIZED_UDF may
be very large when operator includes some large objects.
{@link #getStreamOperatorFactory} is used to create a StreamOperator and
usually only needs to be called once.
Callers can clean it to reduce the memory after calling the \{@link
#getStreamOperatorFactory}.
> Add cleaning mechanism for StreamConfig#SERIALIZEDUDF
> -----------------------------------------------------
>
> Key: FLINK-33317
> URL: https://issues.apache.org/jira/browse/FLINK-33317
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Configuration
> Affects Versions: 1.17.0, 1.18.0
> Reporter: Rui Fan
> Assignee: Rui Fan
> Priority: Major
>
> See FLINK-33315 for details.
> This Jira focus on avoid unnecessary memory usage, it can optimize the memory
> cost of Replica_3 in FLINK-33315.
>
> Solution:
> * After getStreamOperatorFactory, the StreamConfig#SERIALIZEDUDF can be
> cleaned
> In general, we don't clean any configuration. However, the SERIALIZED_UDF may
> be very large when operator includes some large objects.
> {@link #getStreamOperatorFactory}
> is used to create a StreamOperator and usually only needs to be called once.
> Callers can clean it to reduce the memory after calling the \{@link
> #getStreamOperatorFactory}.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
