pandasUDF and pyarrow 0.15.0

Question

I have recently started getting a bunch of errors on a number of pyspark jobs running on EMR clusters. The erros are

java.lang.IllegalArgumentException
    at java.nio.ByteBuffer.allocate(ByteBuffer.java:334)
    at org.apache.arrow.vector.ipc.message.MessageSerializer.readMessage(MessageSerializer.java:543)
    at org.apache.arrow.vector.ipc.message.MessageChannelReader.readNext(MessageChannelReader.java:58)
    at org.apache.arrow.vector.ipc.ArrowStreamReader.readSchema(ArrowStreamReader.java:132)
    at org.apache.arrow.vector.ipc.ArrowReader.initialize(ArrowReader.java:181)
    at org.apache.arrow.vector.ipc.ArrowReader.ensureInitialized(ArrowReader.java:172)
    at org.apache.arrow.vector.ipc.ArrowReader.getVectorSchemaRoot(ArrowReader.java:65)
    at org.apache.spark.sql.execution.python.ArrowPythonRunner$$anon$1.read(ArrowPythonRunner.scala:162)
    at org.apache.spark.sql.execution.python.ArrowPythonRunner$$anon$1.read(ArrowPythonRunner.scala:122)
    at org.apache.spark.api.python.BasePythonRunner$ReaderIterator.hasNext(PythonRunner.scala:406)
    at org.apache.spark.InterruptibleIterator.hasNext(InterruptibleIterator.scala:37)
    at org.apache.spark.sql.execution.python.ArrowEvalPythonExec$$anon$2.<init>(ArrowEvalPythonExec.scala:98)
    at org.apache.spark.sql.execution.python.ArrowEvalPythonExec.evaluate(ArrowEvalPythonExec.scala:96)
    at org.apache.spark.sql.execution.python.EvalPythonExec$$anonfun$doExecute$1.apply(EvalPythonExec.scala:127)...

They all seem to happen in apply functions of a pandas series. The only change I found is that pyarrow has been updated on Saturday (05/10/2019). Tests seem to work with 0.14.1

So my question is if anyone know if this is a bug in the new updated pyarrow or is there some significant change that will make pandasUDF hard to use in the future?

Answer 1

It's not a bug. We made an important protocol change in 0.15.0 that makes the default behavior of pyarrow incompatible with older versions of Arrow in Java -- your Spark environment seems to be using an older version.

Your options are

• Set the environment variable ARROW_PRE_0_15_IPC_FORMAT=1 from where you are using Python
• Downgrade to pyarrow < 0.15.0 for now.

Hopefully the Spark community will be able to upgrade to 0.15.0 in Java soon so this issue goes away.

This is discussed in http://arrow.apache.org/blog/2019/10/06/0.15.0-release/