Apache Flink

Hello Team,

I’ve a question around savepoint vs checkpoint for graceful shut down of flink streaming job.
I’ve a flink streaming job(version 1.16).
In my job I’ve enabled
• checkpointing(via rocksDB)
• incremental checkpointing
• checkpoint external storage on hdfs
• ExternalizedCheckpointCleanup.RETAIN_ON_CANCELLATION
if job fails, it can automatically resume from latest retained checkpoint using --savepointPath option

But in case, if I want to gracefully shutdown the job using stop method

```./bin/flink stop \
      --savepointPath /tmp/flink-savepoints \
      $JOB_ID```
Since I already have checkpoint dir in my job do I need to specify a separate dir for --savepointPath for graceful shutdown ?
If yes,
For restarting the job, which dir should I use
the one specified via --savepointPath  or retained checkpointDir?