是否为不同查询创建新的SparkSession？

Question

我想从elasticsearch获取两个数据

一个是用查询过滤的，另一个没有过滤器。

 // with query
 session = get_spark_session(query=query)

 df = session.read.option(
     "es.resource", "analytics-prod-2019.08.02"
 ).format("org.elasticsearch.spark.sql").load()


 df.show() // empty result

 // without query
 session = get_spark_session()

 df = session.read.option(
     "es.resource", "analytics-prod-2019.08.02"
 ).format("org.elasticsearch.spark.sql").load()

 df.show() // empty result


 def get_spark_session(query=None, excludes=[]):

     conf = pyspark.SparkConf()
     conf.set("spark.driver.allowMultipleContexts", "true")
     conf.set("es.index.auto.create", "true")
     conf.set("es.nodes.discovery", "true")
     conf.set("es.scroll.size", 10000)
     conf.set("es.read.field.exclude", excludes)
     conf.set("spark.driver.extraClassPath", "/usr/local/elasticsearch-hadoop/dist/elasticsearch-spark-20_2.11-6.6.2.jar")
     if query:
         conf.set("es.query", query)


     sc = SparkSession.builder.config(conf=conf).getOrCreate()

     return sc 

问题是会话是否被重用。

当我第一次运行filtered查询，第二次运行non-filtered查询时，两者都给出了空结果

但是当我第一次运行non-filtered查询时，它显示了一些结果，随后的filtered查询显示空结果。

 // below, I reverse the order
 // without query
 session = get_spark_session()

 df = session.read.option(
     "es.resource", "analytics-prod-2019.08.02"
 ).format("org.elasticsearch.spark.sql").load()

 df.show() // some result

 // with query
 session = get_spark_session(query=query)

 df = session.read.option(
     "es.resource", "analytics-prod-2019.08.02"
 ).format("org.elasticsearch.spark.sql").load()


 df.show() // empty result

**编辑

这样我就可以用下面的代码得到想要的结果：

def get_spark_session(query=None, excludes=[]):

    conf = pyspark.SparkConf()
    conf.set("spark.driver.allowMultipleContexts", "true")
    conf.set("es.index.auto.create", "true")
    conf.set("es.nodes.discovery", "true")
    conf.set("es.scroll.size", 10000)
    conf.set("es.read.field.exclude", excludes)
    conf.set("spark.driver.extraClassPath", "/usr/local/elasticsearch-hadoop/dist/elasticsearch-spark-20_2.11-6.6.2.jar")
    if query:
        conf.set("es.query", query)
    else:
        conf.set("es.query", "") # unset the query 

Answer 1

SparkSession.builder获取现有的SparkSession，或者，如果没有现有的，则基于此构建器中设置的选项创建一个新的构建器。在您的例子中，spark配置正在被重用。从配置中删除"es.query“应该可以解决这个问题：

def get_spark_session(query=None, excludes=[]):
     conf = pyspark.SparkConf()
     conf.unset("es.query")
     conf.set("spark.driver.allowMultipleContexts", "true")
     conf.set("es.index.auto.create", "true")
     conf.set("es.nodes.discovery", "true")
     conf.set("es.scroll.size", 10000)
     conf.set("es.read.field.exclude", excludes)
     conf.set("spark.driver.extraClassPath", "/usr/local/elasticsearch-hadoop/dist/elasticsearch-spark-20_2.11-6.6.2.jar")
     if query:
         conf.set("es.query", query)    

     sc = SparkSession.builder.config(conf=conf).getOrCreate()    
     return sc