spark.streaming.backpressure.enabled:false 开启流控
WriteAheadLogUtils.scala:
spark.streaming.receiver.writeAheadLog.enable:false 控制是否开户write ahead log
spark.streaming.receiver.writeAheadLog.class>>FileBasedWriteAheadLog OR BatchedWriteAheadLog
spark.streaming.receiver.writeAheadLog.rollingIntervalSecs:60S
spark.streaming.receiver.writeAheadLog.maxFailures:3
spark.streaming.receiver.writeAheadLog.closeFileAfterWrite:false,为True时,并行
FileBasedWriteAheadLog 每隔60S,往$checkpointdir/receivedData/$streamid目录写
文件名格式为log-starttime-stoptime
FileBasedWriteAheadLog启动时会从该目录文件恢复,内存结构为pastLogs。
另外receiveblock write ahead log地址:
$checkpointdir/receivedBlockMetadata/$streamid
Spark streaming block生成间隔 BlockGenerator.scala
spark.streaming.blockInterval:2000(ms)
spark.streaming.blockQueueSize:10(内存buf)
网友评论