[SYSTEMDS-1370] Py4JError: An error occurred while calling z:org.apache.sysml.runtime.instructions.spark.utils.RDDConverterUtilsExt.convertPy4JArrayToMB.

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Not Applicable
• Fix Version/s: SystemML 0.14
• Component/s: APIs
• Environment: pyspark with local Spark 2.1
• Assignee: Niketan Pansare
• Reporter: Berthold Reinwald
• Created: 03/Mar/17 08:55
• Updated: 02/May/17 18:33
• Resolved: 17/Mar/17 19:23

Description

Do we have undocumented limits for RDDConverterUtilsExt.convertPy4JArrayToMB?

Below simple script works for 23100 rows, while 46900 fails. This is how to easily and consistently reproduce.

START:
$pyspark --master local --jars $SYSTEMML_HOME/SystemML.jar --driver-memory 8G --executor-memory 2G

PYTHON SCRIPT:
from systemml import MLContext, dml
import pandas as pd

sc.version
ml = MLContext(sc)
print "Spark Version:", sc.version
print "SystemML Version:", ml.version()
print "SystemML Built-Time:", ml.buildTime()

1. !! number of rows 23100 works, while 46900 fails
   nr = 46900

X_pd = pd.DataFrame(range(1, (nr*784)+1,1),dtype=float).values.reshape(nr,784)

script ="""
write(X, $Xfile, format="csv")
"""
prog = dml(script).input(X=X_pd).input(**

{"$Xfile":"/tmp/X_pd.csv"}

)
ml.execute(prog)

OUTPUT:
Spark Version: 2.1.0
SystemML Version: 0.14.0-incubating-SNAPSHOT
SystemML Built-Time: 2017-03-03 07:33:40 UTC
---------------------------------------------------------------------------
Py4JError Traceback (most recent call last)
.......

Py4JError: An error occurred while calling z:org.apache.sysml.runtime.instructions.spark.utils.RDDConverterUtilsExt.convertPy4JArrayToMB. Trace:
java.lang.NegativeArraySizeException
at py4j.Base64.decode(Base64.java:321)
at py4j.Protocol.getBytes(Protocol.java:173)
at py4j.Protocol.getObject(Protocol.java:294)
at py4j.commands.AbstractCommand.getArguments(AbstractCommand.java:82)
at py4j.commands.CallCommand.execute(CallCommand.java:77)
at py4j.GatewayConnection.run(GatewayConnection.java:214)
at java.lang.Thread.run(Thread.java:745)