[
https://issues.apache.org/jira/browse/FLINK-14159?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16939204#comment-16939204
]
jackylau commented on FLINK-14159:
----------------------------------
[~klion26], It will be at cost of cpu when the state is large and It needn't to
compress twice. so i think it may need some conments on the flink compression
code and ref docs when using rocksdb backend or a extra compress flag to choose
which compression you want (flink or rocksdb)
> flink rocksdb StreamCompressionDecorator not right
> --------------------------------------------------
>
> Key: FLINK-14159
> URL: https://issues.apache.org/jira/browse/FLINK-14159
> Project: Flink
> Issue Type: Bug
> Components: Runtime / State Backends
> Affects Versions: 1.9.0
> Reporter: jackylau
> Priority: Major
> Fix For: 1.10.0
>
>
> I think the current flink rocksdb StreamCompressionDecorator is not right
> when calling method
> getCompressionDecorator(executionConfig) ,which defalut value is false.That
> is to say, current compression is none.But I find rocksdb using
> {{options.compression}} to specify the compression to use. By default it is
> Snappy, which you can see here
> [https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide]. And I use
> rocksdb tool sstdump to find it is indeed snappy compression.
> So I think it should be return SnappyStreamCompressionDecorator.INSTANCE
> rather than getCompressionDecorator( executionConfig)
> Coud i commit a PR?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
