[
https://issues.apache.org/jira/browse/FLINK-17970?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17118223#comment-17118223
]
Yang Wang commented on FLINK-17970:
-----------------------------------
Sorry for jumping here.
But i think it may have some problems when we use the following code to create
a cached thread pool. If the {{maxPoolSize}} is not the {{Integer.MAX_VALUE}}
and use {{SynchronousQueue}}, then we have a risk that the execution is blocked
and {{RejectedExecutionException}} will be thrown. So we need to add a
{{RejectedExecutionHandler}}.
{code:java}
new ThreadPoolExecutor(0, maxPoolSize, 60L, TimeUnit.SECONDS, new
SynchronousQueue<>(), threadFactory);
{code}
> Increase default value of IO pool executor to 4 * #cores
> --------------------------------------------------------
>
> Key: FLINK-17970
> URL: https://issues.apache.org/jira/browse/FLINK-17970
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Coordination
> Affects Versions: 1.11.0
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.11.0
>
>
> Currently, the default value of
> {{ClusterOptions.CLUSTER_IO_EXECUTOR_POOL_SIZE}} is #cores. I propose to
> increase it to 4 * #cores to support a higher load of blocking IO operations.
> Moreover, I propose to use a cached thread pool instead of a fixed thread
> pool. That way, only those use cases which have high IO load will actually
> occupy the required resources to start more threads.
> Last but not least, I propose to change the config option name from
> {{cluster.io-executor.pool-size}} to {{cluster.io-pool.size}} which is a bit
> shorter.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
