你好,

看到你的compaction.file-size配置成了2kb,这个是希望合并以后的文件的target size只有2kb么

On Wed, Aug 4, 2021 at 5:39 PM [email protected] <[email protected]>
wrote:

> 你好,
>     在使用filesystem
> connector过程中,开启了compaction,使用parquet列式文件,指定3个并行度,但发现无论如何也触发不了合并,因为列式文件是checkpoint触发时才会滚动,这样同一checkpoint内会产生与并行度相同的文件,按说此时文件数已经大于1了,为什么不合并呢?
>
> create table fs_parquet_compact
> (userid bigint, name string, part string)
> PARTITIONED BY (part)
> with(
> 'connector' = 'filesystem',
> 'path' = 'hdfs:///data/fs_parquet_compact',
> 'format' = 'parquet',
> 'auto-compaction' = 'true',
> 'compaction.file-size' = '2kb',
> 'sink.rolling-policy.file-size' = '500b',
> 'sink.rolling-policy.rollover-interval' = '800s',
> 'sink.rolling-policy.check-interval' = '60s'
> );
>
>
>
> [email protected]
>


-- 
Best regards!
Rui Li

回复