Github user ravipesala commented on the issue:
https://github.com/apache/incubator-carbondata/pull/369
@allwefantasy Thanks for your suggestion. Yes, we can use the Spark's
memory manager instead of our own. I will open the interface and provide
implementation to use
Github user allwefantasy commented on the issue:
https://github.com/apache/incubator-carbondata/pull/369
Does this PR have considered allocating memory from TaskMemoryManagerï¼
Many Spark application runs on Yarn,if you use off-heap,it's easy to trigger
behavior of yarn's killing