[
https://issues.apache.org/jira/browse/FLINK-8125?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16264144#comment-16264144
]
ASF GitHub Bot commented on FLINK-8125:
---------------------------------------
GitHub user StephanEwen opened a pull request:
https://github.com/apache/flink/pull/5059
[FLINK-8125] [core] Introduce limiting of file system connections
## What is the purpose of the change
This change introduces a way to limit the number of streams that Flink
FileSystems concurrently open.
For example, for very small HDFS clusters with few RPC handlers, a large
Flink job trying to build up many connections during a checkpoint causes
failures due to rejected connections.
For example, by adding setting like `fs.hdfs.limit.total: 128`, this would
limit the total number of connections per TaskManager to the filesystem with
scheme "hdfs" to 128.
## Brief change log
- Adds the `LimitedConnectionsFileSystem` which can wrap existing file
systems and track and limit the number of input- and output streams
- Adds code to `FileSystem` to wrap the file system factories in
`ConnectionLimitingFactory` which wraps the file system in a
`LimitedConnectionsFileSystem` if connection limiting is configured.
- Adds documentation for Flink's file system support in general,
including the here added configuration settings
## Verifying this change
- This change is covered by various unit tests:
- `LimitedConnectionsFileSystemTest`
- `LimitedConnectionsFileSystemDelegationTest`
- `LimitedConnectionsConfigurationTest`
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / **no)**
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / **no)**
- The serializers: (yes / **no** / don't know)
- The runtime per-record code paths (performance sensitive): (yes /
**no** / don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / **no** / don't know)
- The S3 file system connector: (yes / **no** / don't know)
## Documentation
- Does this pull request introduce a new feature? (**yes** / no)
- If yes, how is the feature documented? (not applicable / **docs** /
JavaDocs / not documented)
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/StephanEwen/incubator-flink fs_connections
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/5059.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #5059
----
commit 81b1df9fd82b4e5d7feeed257afac72525b89aa9
Author: Stephan Ewen <se...@apache.org>
Date: 2017-11-08T19:14:34Z
[hotfix] [core] Fix lots of checkstyle errors in core.fs
commit c17d00654fcfeb3ed5729810bb2a9b3ba127ddb9
Author: Stephan Ewen <se...@apache.org>
Date: 2017-11-08T22:57:04Z
[FLINK-8125] [core] Introduce limiting of outgoing file system connections
----
> Support limiting the number of open FileSystem connections
> ----------------------------------------------------------
>
> Key: FLINK-8125
> URL: https://issues.apache.org/jira/browse/FLINK-8125
> Project: Flink
> Issue Type: Improvement
> Components: Core
> Reporter: Stephan Ewen
> Assignee: Stephan Ewen
> Fix For: 1.5.0, 1.4.1
>
>
> We need a way to limit the number of streams that Flink FileSystems
> concurrently open.
> For example, for very small HDFS clusters with few RPC handlers, a large
> Flink job trying to build up many connections during a checkpoint causes
> failures due to rejected connections.
> I propose to add a file system that can wrap another existing file system The
> file system may track the progress of streams and close streams that have
> been inactive for too long, to avoid locked streams of taking up the complete
> pool.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
