大叔经验分享（65）spark读取不到hive表

spark 2.4.3

spark读取hive表，步骤：

1）hive-site.xml

hive-site.xml放到$SPARK_HOME/conf下

2）enableHiveSupport

SparkSession.builder.enableHiveSupport().getOrCreate()

3) 测试代码

    val sparkConf = new SparkConf().setAppName(getName)
    val sc = new SparkContext(sparkConf)
    val spark = SparkSession.builder.config(sparkConf).enableHiveSupport().getOrCreate()
    spark.sql("show databases").rdd.foreach(println)

使用$SPARK_HOME/bin/spark-submit提交任务后发现并不能读取到hive的数据库，相关日志如下

19/05/31 13:11:31 WARN SparkContext: Using an existing SparkContext; some configuration may not take effect.
19/05/31 13:11:31 INFO SharedState: loading hive config file: file:/export/spark-2.4.3-bin-hadoop2.6/conf/hive-site.xml
19/05/31 13:11:31 INFO SharedState: spark.sql.warehouse.dir is not set, but hive.metastore.warehouse.dir is set. Setting spark.sql.warehouse.dir to the value of hive.metastore.warehouse.dir ('/user/hive/warehouse').
19/05/31 13:11:31 INFO SharedState: Warehouse path is '/user/hive/warehouse'.
19/05/31 13:11:31 INFO StateStoreCoordinatorRef: Registered StateStoreCoordinator endpoin

说明已经读到hive-site.xml；

进一步测试，使用$SPARK_HOME/bin/spark-sql或者$SPARK_HOME/bin/spark-shell发现都可以读到hive数据库，很神奇有没有，

$SPARK_HOME/bin/spark-shell启动的类为org.apache.spark.repl.Main

"${SPARK_HOME}"/bin/spark-submit --class org.apache.spark.repl.Main --name "Spark shell" "$@"

跟进org.apache.spark.repl.Main代码

...
      val builder = SparkSession.builder.config(conf)
      if (conf.get(CATALOG_IMPLEMENTATION.key, "hive").toLowerCase(Locale.ROOT) == "hive") {
        if (SparkSession.hiveClassesArePresent) {
          // In the case that the property is not set at all, builder's config
          // does not have this value set to 'hive' yet. The original default
          // behavior is that when there are hive classes, we use hive catalog.
          sparkSession = builder.enableHiveSupport().getOrCreate()
          logInfo("Created Spark session with Hive support")
        } else {
          // Need to change it back to 'in-memory' if no hive classes are found
          // in the case that the property is set to hive in spark-defaults.conf
          builder.config(CATALOG_IMPLEMENTATION.key, "in-memory")
          sparkSession = builder.getOrCreate()
          logInfo("Created Spark session")
        }
      } else {
        // In the case that the property is set but not to 'hive', the internal
        // default is 'in-memory'. So the sparkSession will use in-memory catalog.
        sparkSession = builder.getOrCreate()
        logInfo("Created Spark session")
      }
      sparkContext = sparkSession.sparkContext
      sparkSession
...

发现和测试代码有些差异，关键是在倒数第二行，这里是先创建SparkSession，再从SparkSession中获取SparkContext，另外注意到之前有个WARN级别的日志

19/05/31 13:11:31 WARN SparkContext: Using an existing SparkContext; some configuration may not take effect.

修改测试代码

    val sparkConf = new SparkConf().setAppName(getName)
    //val sc = new SparkContext(sparkConf)
    val spark = SparkSession.builder.config(sparkConf).enableHiveSupport().getOrCreate()
    val sc = spark.sparkContext
    spark.sql("show databases").rdd.foreach(println)

这次果然ok了，详细原因有空再看，未完待续；