6

I would like to read Parquet data stored on S3 from PySpark.

I've downloaded spark from here:

http://www.apache.org/dist/spark/spark-2.1.0/spark-2.1.0-bin-hadoop2.7.tgz

And installed it to Python naively

cd python
python setup.py install

This seems to function fine and I can import pyspark, make a SparkContext, etc.. However when I go to read some publicly accessible parquet data I get the following:

import pyspark
sc = pyspark.SparkContext('local[4]')
sql = pyspark.SQLContext(sc)
df = sql.read.parquet('s3://bucket-name/mydata.parquet')

And I receive the following exception

Py4JJavaError: An error occurred while calling o55.parquet.
: java.io.IOException: No FileSystem for scheme: s3
    at org.apache.hadoop.fs.FileSystem.getFileSystemClass(FileSystem.java:2660)
    at org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:2667)
    at org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:94)
    at org.apache.hadoop.fs.FileSystem$Cache.getInternal(FileSystem.java:2703)
    at org.apache.hadoop.fs.FileSystem$Cache.get(FileSystem.java:2685)
    at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:373)
    at org.apache.hadoop.fs.Path.getFileSystem(Path.java:295)
    at org.apache.spark.sql.execution.datasources.DataSource$$anonfun$14.apply(DataSource.scala:372)
    at org.apache.spark.sql.execution.datasources.DataSource$$anonfun$14.apply(DataSource.scala:370)
    at scala.collection.TraversableLike$$anonfun$flatMap$1.apply(TraversableLike.scala:241)
    at scala.collection.TraversableLike$$anonfun$flatMap$1.apply(TraversableLike.scala:241)
    at scala.collection.immutable.List.foreach(List.scala:381)
    at scala.collection.TraversableLike$class.flatMap(TraversableLike.scala:241)
    at scala.collection.immutable.List.flatMap(List.scala:344)
    at org.apache.spark.sql.execution.datasources.DataSource.resolveRelation(DataSource.scala:370)
    at org.apache.spark.sql.DataFrameReader.load(DataFrameReader.scala:152)
    at org.apache.spark.sql.DataFrameReader.parquet(DataFrameReader.scala:441)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)
    at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)
    at py4j.Gateway.invoke(Gateway.java:280)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:214)
    at java.lang.Thread.run(Thread.java:745)

This error pops up a bit from google searches. So far none of the solutions provided have been helpful.

I'm on Linux (Ubuntu 16.04) on a personal computer without much else installed (everything is pretty stock).

Update

I downgraded to http://www.apache.org/dist/spark/spark-2.1.0/spark-2.1.0-bin-hadoop2.4.tgz to have AWS included by default.

Now unfortunately my AWS credentials aren't being picked up. I've tried a few things:

  1. Including them as SparkConf parameters

    conf = (pyspark.SparkConf()
               .set('fs.s3.awsAccessKeyId', ...')
               .set('fs.s3.awsSecretAccessKey', '...'))
sc = pyspark.SparkContext('local[4]', conf=conf)
  2. Including them in my local .aws/credentials file
  3. Including them in the URL (doesn't work because my access key has a forward slash)

Unfortunately in all cases I receive a traceback like the following

IllegalArgumentException: 'AWS Access Key ID and Secret Access Key must be specified as the username or password (respectively) of a s3 URL, or by setting the fs.s3.awsAccessKeyId or fs.s3.awsSecretAccessKey properties (respectively).'
Arafat Nalkhande
  • 8,663
  • 7
  • 35
  • 57
MRocklin
  • 48,441
  • 20
  • 124
  • 196
  • You've tried the `s3n` and `s3a` prefixes, correct? – quasiben Mar 13 '17 at 03:17
  • I think the hadoop-aws.jar is not bundled with spark. According to this other [SO post](http://stackoverflow.com/questions/30851244/spark-read-file-from-s3-using-sc-textfile-s3n) and [SPARK-7481](https://issues.apache.org/jira/browse/SPARK-7481) the issue of bundling aws dependencies is still being discussed. You can grab the latest [hadoop-aws jar](https://repo1.maven.org/maven2/org/apache/hadoop/hadoop-aws/2.7.3/hadoop-aws-2.7.3.jar) – quasiben Mar 13 '17 at 03:23
  • Updated original question with current issues since reading these excellent comments – MRocklin Mar 13 '17 at 14:24

1 Answers1

2

Using the Hadoop-2.4 build of the pre-built spark 2.X binary (which I believe ships with s3 functionality) you can programmatically configure spark to pull s3 data in the following manner:

import pyspark
conf = pyspark.SparkConf()

sc = pyspark.SparkContext('local[4]', conf=conf)
sc._jsc.hadoopConfiguration().set("fs.s3n.awsAccessKeyId", "")
sc._jsc.hadoopConfiguration().set("fs.s3n.awsSecretAccessKey", "")

sql = pyspark.SQLContext(sc)
df = sql.read.parquet('s3n://bucket-name/mydata.parquet')

A critical thing to note is the prefix s3n in both the URI for the bucket and the configuration name

quasiben
  • 1,279
  • 9
  • 17