Spark: shuffle operation leading to long GC pause

Question

I'm running Spark 2 and am trying to shuffle around 5 terabytes of json. I'm running into very long garbage collection pauses during shuffling of a Dataset:

val operations = spark.read.json(inPath).as[MyClass]
operations.repartition(partitions, operations("id")).write.parquet("s3a://foo")

Are there any obvious configuration tweaks to deal with this issue? My configuration is as follows:

spark.driver.maxResultSize 6G
spark.driver.memory 10G
spark.executor.extraJavaOptions -XX:+UseG1GC -XX:MaxPermSize=1G -XX:+HeapDumpOnOutOfMemoryError
spark.executor.memory   32G
spark.hadoop.fs.s3a.buffer.dir  /raid0/spark
spark.hadoop.fs.s3n.buffer.dir  /raid0/spark
spark.hadoop.fs.s3n.multipart.uploads.enabled   true
spark.hadoop.parquet.block.size 2147483648
spark.hadoop.parquet.enable.summary-metadata    false
spark.local.dir /raid0/spark
spark.memory.fraction 0.8
spark.mesos.coarse  true
spark.mesos.constraints  priority:1
spark.mesos.executor.memoryOverhead 16000
spark.network.timeout   600
spark.rpc.message.maxSize    1000
spark.speculation   false
spark.sql.parquet.mergeSchema   false
spark.sql.planner.externalSort  true
spark.submit.deployMode client
spark.task.cpus 1

Asberry

Are you sure you need 1G MaxPermSize? Also probably it worth using Java8 (so in case of OpenJDK/OracleJDK that option is ignored). – Cornel

Answer 1

Adding the following flags got rid of the GC pauses.

spark.executor.extraJavaOptions -XX:+UseG1GC -XX:InitiatingHeapOccupancyPercent=35 -XX:ConcGCThreads=12

I think it does take a fair amount of tweaking though. This databricks post was very very helpful.

Asberry