Flume composition, put transaction, take transaction

Taildir source
a. Breakpoint continuation （ Network failure can also read data ）
b. It can monitor the changes of data generated by files in real time
c. Multi directory
d. Taildir What to do if I hang up ？—— You won't lose the number , Breakpoint continuation
e. How to deal with duplicate data ？—— Don't deal with , It is not normally handled in production environments , Because it will affect the transmission efficiency

Kafka channel（ The next level is Kafka）
The data is stored in Kafka in , Disk based ;
advantage ： High reliability
Less Flume Of Sink Stage , Improved transmission efficiency .

File channel（ Finance 、 Be accurate about money ）
Data is stored on disk ;
advantage ： High reliability
shortcoming ： Slow transmission efficiency

Memory channel（ Ordinary log ）
Data stored in memory ;
advantage ： Fast transmission efficiency
shortcoming ： Poor reliability

Source To Channel yes Put Business
Channel To Sink yes Take Business