[
https://issues.apache.org/jira/browse/FLINK-10954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16698635#comment-16698635
]
Stefan Richter commented on FLINK-10954:
----------------------------------------
I agree with [~StephanEwen] that we should work with hardlink for the RocksDB
local state. We were choosing the temp dir mostly to support eventual cleanup
in non-container environments and the local state dir structur in general has a
layout to support cleanups. I see that this case is a bit tricky and I think it
can become more tricky if we consider that we support multiple instance
directories for RocksDB. Each instance directory could be on a different
physical device and that means we would also require to have a local state
directory co-located per physical device to work with the hardlinks. While we
could enhance the local state store to supply particular local state
directories on request, I wonder what the best way could be to ensure i) that
there is a corresponding local dir for each physical device that holds instance
dirs and ii) the best way to establish a connection between the two so that we
can request the right directories.
> Hardlink from files of previous local stored state might cross devices
> ----------------------------------------------------------------------
>
> Key: FLINK-10954
> URL: https://issues.apache.org/jira/browse/FLINK-10954
> Project: Flink
> Issue Type: Bug
> Components: State Backends, Checkpointing
> Affects Versions: 1.6.2
> Reporter: Yun Tang
> Assignee: Yun Tang
> Priority: Major
> Fix For: 1.6.3, 1.8.0, 1.7.1
>
>
> Currently, local recovery's base directories is initialized from
> '{{io.tmp.dirs}}' if parameter '{{taskmanager.state.local.root-dirs}}' is not
> set. For Yarn environment, the tmp dirs is replaced by its '{{LOCAL_DIRS}}',
> which might consist of directories from different devices, such as
> /dump/1/nm-local-dir, /dump/2/nm-local-dir. The local directory for RocksDB
> is initialized from IOManager's spillingDirectories, which might located in
> different device from local recovery's folder. However, hard-link between
> different devices is not allowed, it will throw exception below:
> {code:java}
> java.nio.file.FileSystemException: target -> souce: Invalid cross-device link
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
