BucketingSink ????????????
??flink on yarn??1??TaskManager??4??slot??TaskManager4G??JobManager1G??BucketingSinkhdfs??3??checkpoint??1003300427427*300=128100=125KB80??80*300=24000=23KB??Flink ?? TaskManager?? jvm ??1G?? JVM (Heap/Non-Heap) Type Committed Used Maximum Heap 2.68 GB 863 MB 2.68 GB Non-Heap 84.3 MB 82.8 MB -1 B Total2.76 GB 946 MB 2.68 GB 3s chepoint hdfs3s??
Re: env.readFile只读取新文件
恐怕没有现成的,自己写一个,继承 SourceFunction Thanks, Biao /'bɪ.aʊ/ On Wed, Jul 31, 2019 at 4:49 PM 王佩 wrote: > 如下代码: > > DataStreamSource source = env.readFile( > textInputFormat, > "/data/appData/streamingWatchFile/source", > FileProcessingMode.PROCESS_CONTINUOUSLY, > 10 * 1000 > ); > > 当被监控目录下的某个文件被修改,如touch了一下,整个文件会重复处理一遍。 > > 有没有什么方法,可以做到只读取新文件。想实现只读取新的Parquet文件的效果。 >