By breakpoints follow-up data are found each topic can go to, but in the end will be blocked on the floor of the operations performed DataFrame;
Such as:
df.write.format("parquet").mode("append").save(path)
Careful observation can discover the type of log: INFO scheduler.JobScheduler: Added Jobs for Time ××××× log;
The reason: Receiver running thread is not enough
Solution: increase the available thread
loca mode, consider increasing the local [n], n increasing size;
reference: