[SPARK-26631] Issue while reading Parquet data from Hadoop Archive files (.har)

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Incomplete
• Affects Version/s: 2.2.0
• Component/s: SQL
• Labels: bulk-closed
• Assignee: Unassigned
• Reporter: Sathish
• Created: 16/Jan/19 04:53
• Updated: 28/Jul/22 12:57
• Resolved: 25/May/21 01:40

Description

While reading Parquet file from Hadoop Archive file Spark is failing with below exception

 

scala> val hardf = sqlContext.read.parquet("har:///tmp/testarchive.har/userdata1.parquet") org.apache.spark.sql.AnalysisException: Unable to infer schema for Parquet. It must be specified manually.;   at org.apache.spark.sql.execution.datasources.DataSource$$anonfun$9.apply(DataSource.scala:208)   at org.apache.spark.sql.execution.datasources.DataSource$$anonfun$9.apply(DataSource.scala:208)   at scala.Option.getOrElse(Option.scala:121)   at org.apache.spark.sql.execution.datasources.DataSource.getOrInferFileFormatSchema(DataSource.scala:207)   at org.apache.spark.sql.execution.datasources.DataSource.resolveRelation(DataSource.scala:393)   at org.apache.spark.sql.DataFrameReader.loadV1Source(DataFrameReader.scala:239)   at org.apache.spark.sql.DataFrameReader.load(DataFrameReader.scala:227)   at org.apache.spark.sql.DataFrameReader.parquet(DataFrameReader.scala:622)   at org.apache.spark.sql.DataFrameReader.parquet(DataFrameReader.scala:606)   ... 49 elided

 

Whereas the same parquet file can be read normally without any issues

scala> val df = sqlContext.read.parquet("hdfs:///tmp/testparquet/userdata1.parquet")

df: org.apache.spark.sql.DataFrame = [registration_dttm: timestamp, id: int ... 11 more fields]

 

Here are the steps to reproduce the issue

 

a) hadoop fs -mkdir /tmp/testparquet

b) Get sample parquet data and rename the file to userdata1.parquet

wget https://github.com/Teradata/kylo/blob/master/samples/sample-data/parquet/userdata1.parquet?raw=true

c) hadoop fs -put userdata.parquet /tmp/testparquet

d) hadoop archive -archiveName testarchive.har -p /tmp/testparquet /tmp

e) We should be able to see the file under har file

hadoop fs -ls har:///tmp/testarchive.har

f) Launch spark2 / spark shell

g)

val sqlContext = new org.apache.spark.sql.SQLContext(sc)
    val df = sqlContext.read.parquet("har:///tmp/testarchive.har/userdata1.parquet")

is there anything which I am missing here.

 

Issue Links

• Is contained by: SPARK-39910 DataFrameReader API cannot read files from hadoop archives (.har)