问题
I am running CDH 4.4 with Spark 0.9.0 from a Cloudera parcel.
I have a bunch of Avro files that were created via Pig's AvroStorage UDF. I want to load these files in Spark, using a generic record or the schema onboard the Avro files. So far I've tried this:
import org.apache.avro.mapred.AvroKey
import org.apache.avro.mapreduce.AvroKeyInputFormat
import org.apache.hadoop.io.NullWritable
import org.apache.commons.lang.StringEscapeUtils.escapeCsv
import org.apache.hadoop.fs.Path
import org.apache.hadoop.fs.FileSystem
import org.apache.hadoop.conf.Configuration
import java.net.URI
import java.io.BufferedInputStream
import java.io.File
import org.apache.avro.generic.{GenericDatumReader, GenericRecord}
import org.apache.avro.specific.SpecificDatumReader
import org.apache.avro.file.DataFileStream
import org.apache.avro.io.DatumReader
import org.apache.avro.file.DataFileReader
import org.apache.avro.mapred.FsInput
val input = "hdfs://hivecluster2/securityx/web_proxy_mef/2014/05/29/22/part-m-00016.avro"
val inURI = new URI(input)
val inPath = new Path(inURI)
val fsInput = new FsInput(inPath, sc.hadoopConfiguration)
val reader = new GenericDatumReader[GenericRecord]
val dataFileReader = DataFileReader.openReader(fsInput, reader)
val schemaString = dataFileReader.getSchema
val buf = scala.collection.mutable.ListBuffer.empty[GenericRecord]
while(dataFileReader.hasNext) {
buf += dataFileReader.next
}
sc.parallelize(buf)
This works for one file, but it can't scale - I am loading all the data into local RAM and then distributing it across the spark nodes from there.
回答1:
To answer my own question:
import org.apache.spark.SparkContext
import org.apache.spark.SparkContext._
import org.apache.avro.generic.GenericRecord
import org.apache.avro.mapred.AvroKey
import org.apache.avro.mapred.AvroInputFormat
import org.apache.avro.mapreduce.AvroKeyInputFormat
import org.apache.hadoop.io.NullWritable
import org.apache.commons.lang.StringEscapeUtils.escapeCsv
import org.apache.hadoop.fs.FileSystem
import org.apache.hadoop.fs.Path
import org.apache.hadoop.conf.Configuration
import java.io.BufferedInputStream
import org.apache.avro.file.DataFileStream
import org.apache.avro.io.DatumReader
import org.apache.avro.file.DataFileReader
import org.apache.avro.file.DataFileReader
import org.apache.avro.generic.{GenericDatumReader, GenericRecord}
import org.apache.avro.mapred.FsInput
import org.apache.avro.Schema
import org.apache.avro.Schema.Parser
import org.apache.hadoop.mapred.JobConf
import java.io.File
import java.net.URI
// spark-shell -usejavacp -classpath "*.jar"
val input = "hdfs://hivecluster2/securityx/web_proxy_mef/2014/05/29/22/part-m-00016.avro"
val jobConf= new JobConf(sc.hadoopConfiguration)
val rdd = sc.hadoopFile(
input,
classOf[org.apache.avro.mapred.AvroInputFormat[GenericRecord]],
classOf[org.apache.avro.mapred.AvroWrapper[GenericRecord]],
classOf[org.apache.hadoop.io.NullWritable],
10)
val f1 = rdd.first
val a = f1._1.datum
a.get("rawLog") // Access avro fields
回答2:
This works for me:
import org.apache.avro.generic.GenericRecord
import org.apache.avro.mapred.{AvroInputFormat, AvroWrapper}
import org.apache.hadoop.io.NullWritable
...
val path = "hdfs:///path/to/your/avro/folder"
val avroRDD = sc.hadoopFile[AvroWrapper[GenericRecord], NullWritable, AvroInputFormat[GenericRecord]](path)
来源：https://stackoverflow.com/questions/23944615/how-can-i-load-avros-in-spark-using-the-schema-on-board-the-avro-files