[
https://issues.apache.org/jira/browse/FLINK-18424?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17245194#comment-17245194
]
Arvid Heise commented on FLINK-18424:
-------------------------------------
Such questions are usually better asked on
[https://flink.apache.org/community.html#mailing-lists|http://example.com] as
the intent and feature request is less than clear.
Using a NFS is supported right now and works okayish for smaller setups. I
can't see a FTP integration happening out of the box because of its many quirks
and I don't see how setting up an FTP server is any easier than setting up an
S3-compliant server.
If you are still interested in the topic, please visit the mailing list, so
that we can discuss the motivation and get to alternatives.
> How does standalone mode not rely on distributed state storage
> --------------------------------------------------------------
>
> Key: FLINK-18424
> URL: https://issues.apache.org/jira/browse/FLINK-18424
> Project: Flink
> Issue Type: Wish
> Components: Runtime / Checkpointing
> Affects Versions: 1.10.0
> Reporter: Echo Lee
> Priority: Major
>
> Currently, Flink provides three types of checkpoints, memory, filesystem and
> rocksdb. Both filesystem and rocksdb need to set the checkpoint directory,
> directory can be local or remote. But if we set the checkpoint of the local
> directory in standalone mode, restore state may not be able to find the
> directory.
> This means that the checkpoint in standalone mode must rely on distributed
> storage. If we don’t want to rely on distributed storage, is there a better
> way to achieve it?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
