Spark GC time is very high causing task execution slow

Viewed 6642

I am running spark application on 5 node cluster. Each node have configuration as 16 core and 64 GB ram.

There are 4 stages in my application. 3 stages executed fast for 4th stage ie mapToPair is it taking too much time(almost 4 hours).

Below is the job statistics for jobExecutors GC Time

I have checked executors logs as well, didn't see any errors/exception. Could you please suggest me if any configuration changes needs to be done to improve the performance?

2 Answers

You can try to repartition the dataframe inside the application and increase cores, with partitons = cores and check again.

Related