Spark Streaming接收Kafka数据存储到Hbase

Spark Streaming接收Kafka数据存储到Hbase

主要参考了这篇文章https://yq.aliyun.com/articles/60712([点我])(https://yq.aliyun.com/articles/60712), 不过这篇文章使用的spark貌似是spark1.x的。我这里主要是改为了spark2.x的方式

kafka生产数据

闲话少叙,直接上代码:

  1. import java.util.{Properties, UUID
  2.  
  3. import org.apache.kafka.clients.producer.{KafkaProducer, ProducerRecord
  4. import org.apache.kafka.common.serialization.StringSerializer 
  5.  
  6. import scala.util.Random 
  7.  
  8.  
  9. object KafkaProducerTest
  10. def main(args: Array[String]): Unit = { 
  11. val rnd = new Random() 
  12. // val topics = "world" 
  13. val topics = "test" 
  14. val brokers = "localhost:9092" 
  15. val props = new Properties() 
  16. props.put("delete.topic.enable", "true"
  17. props.put("key.serializer", classOf[StringSerializer]) 
  18. // props.put("value.serializer", "org.apache.kafka.common.serialization.StringDeserializer") 
  19. props.put("value.serializer", classOf[StringSerializer]) 
  20. props.put("bootstrap.servers", brokers) 
  21. //props.put("batch.num.messages","10");//props.put("batch.num.messages","10"); 
  22.  
  23. //props.put("queue.buffering.max.messages", "20"); 
  24. //linger.ms should be 0~100 ms 
  25. props.put("linger.ms", "50"
  26. //props.put("block.on.buffer.full", "true"); 
  27. //props.put("max.block.ms", "100000"); 
  28. //batch.size and buffer.memory should be changed with "the kafka message size and message sending speed" 
  29. props.put("batch.size", "16384"
  30. props.put("buffer.memory", "1638400"
  31.  
  32. props.put("queue.buffering.max.messages", "1000000"
  33. props.put("queue.enqueue.timeout.ms", "20000000"
  34. props.put("producer.type", "sync"
  35.  
  36. val producer = new KafkaProducer[String,String](props) 
  37. for(i <- 1001 to 2000){ 
  38. val key = UUID.randomUUID().toString.substring(0,5
  39. val value = "fly_" + i + "_" + key 
  40. producer.send(new ProducerRecord[String, String](topics,key, value))//.get() 
  41.  

  42.  
  43. producer.flush() 


  44.  

生产的数据格式为(key,value) = (uuid, fly_i_key) 的形式

spark streaming 读取kafka并保存到hbase

当kafka里面有数据后,使用spark streaming 读取,并存。直接上代码:

  1. import java.util.UUID 
  2.  
  3. import org.apache.hadoop.hbase.HBaseConfiguration 
  4. import org.apache.hadoop.hbase.client.{Mutation, Put
  5. import org.apache.hadoop.hbase.io.ImmutableBytesWritable 
  6. import org.apache.hadoop.hbase.mapreduce.TableOutputFormat 
  7. import org.apache.hadoop.hbase.util.Bytes 
  8. import org.apache.hadoop.mapred.JobConf 
  9. import org.apache.hadoop.mapreduce.OutputFormat 
  10. import org.apache.kafka.clients.consumer.ConsumerRecord 
  11. import org.apache.kafka.common.serialization.StringDeserializer 
  12. import org.apache.spark.rdd.RDD 
  13. import org.apache.spark.sql.SparkSession 
  14. import org.apache.spark.streaming.kafka010.ConsumerStrategies.Subscribe 
  15. import org.apache.spark.streaming.kafka010.KafkaUtils 
  16. import org.apache.spark.streaming.kafka010.LocationStrategies.PreferConsistent 
  17. import org.apache.spark.streaming.{Seconds, StreamingContext
  18.  
  19. /** 
  20. * spark streaming 写入到hbase 
  21. * Sparkstreaming读取Kafka消息再结合SparkSQL,将结果保存到HBase 
  22. */ 
  23.  
  24.  
  25. object OBDSQL
  26. case class Person(name: String, age: Int, key: String) 
  27.  
  28. def main(args: Array[String]): Unit = { 
  29. val spark = SparkSession 
  30. .builder() 
  31. .appName("sparkSql"
  32. .master("local[4]"
  33. .getOrCreate() 
  34.  
  35. val sc = spark.sparkContext 
  36.  
  37. val ssc = new StreamingContext(sc, Seconds(5)) 
  38.  
  39. val topics = Array("test"
  40. val kafkaParams = Map
  41. "bootstrap.servers" -> "localhost:9092,anotherhost:9092"
  42. "key.deserializer" -> classOf[StringDeserializer], 
  43. "value.deserializer" -> classOf[StringDeserializer], 
  44. // "group.id" -> "use_a_separate_group_id_for_each_stream", 
  45. "group.id" -> "use_a_separate_group_id_for_each_stream_fly"
  46. // "auto.offset.reset" -> "latest", 
  47. "auto.offset.reset" -> "earliest"
  48. // "auto.offset.reset" -> "none", 
  49. "enable.auto.commit" -> (false: java.lang.Boolean

  50.  
  51. val lines = KafkaUtils.createDirectStream[String, String]( 
  52. ssc, 
  53. PreferConsistent
  54. Subscribe[String, String](topics, kafkaParams) 

  55.  
  56. // lines.map(record => (record.key, record.value)).print() 
  57. // lines.map(record => record.value.split("_")).map(x=> (x(0),x(1), x(2))).print() 
  58.  
  59. lines.foreachRDD((rdd: RDD[ConsumerRecord[String, String]]) => { 
  60. import spark.implicits._ 
  61. if (!rdd.isEmpty()) { 
  62.  
  63. // temp table 
  64. rdd.map(_.value.split("_")).map(p => Person(p(0), p(1).trim.toInt, p(2))).toDF.createOrReplaceTempView("temp"
  65.  
  66. // use spark sql 
  67. val rs = spark.sql("select * from temp"
  68.  
  69. // create hbase conf 
  70. val hconf = HBaseConfiguration.create 
  71. hconf.set("hbase.zookeeper.quorum", "localhost"); //ZKFC 
  72. hconf.set("hbase.zookeeper.property.clientPort", "2181"
  73. hconf.set("hbase.defaults.for.version.skip", "true"
  74. hconf.set(TableOutputFormat.OUTPUT_TABLE, "t1") // t1是表名, 表里面有一个列簇 cf1 
  75. hconf.setClass("mapreduce.job.outputformat.class", classOf[TableOutputFormat[String]], classOf[OutputFormat[String, Mutation]]) 
  76. val jobConf = new JobConf(hconf) 
  77.  
  78. // convert every line to hbase lines 
  79. rs.rdd.map(line => { 
  80. val put = new Put(Bytes.toBytes(UUID.randomUUID().toString.substring(0, 9))) 
  81. put.addColumn(Bytes.toBytes("cf1"
  82. , Bytes.toBytes("name"
  83. , Bytes.toBytes(line.get(0).toString) 

  84. put.addColumn(Bytes.toBytes("cf1"
  85. , Bytes.toBytes("age"
  86. , Bytes.toBytes(line.get(1).toString) 

  87. put.addColumn(Bytes.toBytes("cf1"
  88. , Bytes.toBytes("key"
  89. , Bytes.toBytes(line.get(2).toString) 

  90. (new ImmutableBytesWritable, put) 
  91. }).saveAsNewAPIHadoopDataset(jobConf) 

  92. }) 
  93.  
  94. lines.map(record => record.value.split("_")).map(x=> (x(0),x(1), x(2))).print() 
  95.  
  96. ssc start() 
  97. ssc awaitTermination() 
  98.  


  99.  

转载于:https://www.cnblogs.com/flyu6/p/10191488.html

你可能感兴趣的:(大数据,java,scala)