I read the 2 GB limits in spark due to bytebuffer limitation of lnteger.MAX_INT. I understood that memory block size can't exceed due to this, so any record in an RDD partition should be less than 2gb.
But I read here(Why does Spark RDD partition has 2GB limit for HDFS?) an RDD partition itself can't exceed 2gb. And repartition into smaller partitions worked. Is it true? Please help me understand it..