Use flink sql for real-time calculation (deployment mode: on yarn). To use
the checkpoint, you need to configure the following in the flink-conf.yaml:


state.backend: filesystem

state.checkpoints.dir: hdfs:///flink/flink-checkpoints

state.savepoints.dir: hdfs:///flink/flink-savepoints

state.checkpoints.num-retained: 10


There are two problems:

1.checkpoint generates a folder at the initial stage in the HDFS.

2. The savepoint cannot record the data status.

For example: The simple wordcount cannot record the accumulated value.

回复