spark写相同的HIVE表或者相同的HDFS路径时抛出异常

1. 写相同的hive表

代码示例

import os
import time
import logging
from pyspark.sql import SparkSession

spark = SparkSession \
            .builder \
            .appName("mytest") \
            .enableHiveSupport() \
            .getOrCreate()

df = spark.sql("select * from xxx.xxx")
print(df.show())
df.write.mode("overwrite").saveAsTable("xxx.xxx")

抛出异常

---------------------------------------------------------------------------
AnalysisException                         Traceback (most recent call last)
/tmp/ipykernel_36930/1113601968.py in <module>
     12 df = spark.sql("select * from dev_sztoc_crm.leon_test")
     13 print(df.show())
---> 14 df.write.mode("overwrite").saveAsTable("dev_sztoc_crm.leon_test")

/usr/share/spark3/python/pyspark/sql/readwriter.py in saveAsTable(self, name, format, mode, partitionBy, **options)
   1158         if format is not None:
   1159             self.format(format)
-> 1160         self._jwrite.saveAsTable(name)
   1161 
   1162     def json(self, path, mode=None, compression=None, dateFormat=None, timestampFormat=None,

/usr/share/spark3/python/lib/py4j-0.10.9-src.zip/py4j/java_gateway.py in __call__(self, *args)
   1302 
   1303         answer = self.gateway_client.send_command(command)
-> 1304         return_value = get_return_value(
   1305             answer, self.gateway_client, self.target_id, self.name)
   1306 

/usr/share/spark3/python/pyspark/sql/utils.py in deco(*a, **kw)
    115                 # Hide where the exception came from that shows a non-Pythonic
    116                 # JVM exception message.
--> 117                 raise converted from None
    118             else:
    119                 raise

AnalysisException: Cannot overwrite table xxx.xxx that is also being read from

 查看hive表，数据还在！

正确代码示例

import os
import time
import logging
from pyspark.sql import SparkSession

spark = SparkSession \
            .builder \
            .appName("pyspark-test") \
            .enableHiveSupport() \
            .getOrCreate()

df = spark.sql("select * from dev_sztoc_crm.leon_test")

# 将结果保存在临时表中
df.write.mode("Overwrite").saveAsTable("dev_sztoc_crm.leon_test_tmp")
# 将临时表覆盖结果表
spark.table("dev_sztoc_crm.leon_test_tmp").write.mode("overwrite").saveAsTable("dev_sztoc_crm.leon_test")
# 删除临时表
spark.sql("DROP TABLE IF EXISTS dev_sztoc_crm.leon_test_tmp")

2. spark写相同的hdfs路径 

import os
import time
import logging
from pyspark.sql import SparkSession

spark = SparkSession \
            .builder \
            .appName("mytest")
            .getOrCreate()

df = spark.read.csv("xxx")
df.write.mode("overwrite").save("xxx")

抛出异常

Caused by: java.io.FileNotFoundException: File does not exist: /user/leon/overwrite/xxx.csv

……

It is possible the underlying files have been updated. You can explicitly invalidate the cache in Spark by running 'REFRESH TABLE tableName' command in SQL or by recreating the Dataset/DataFrame involved.

 查看原来的hdfs路径，发现文件已经被清空！

正确代码示例

import os
import time
import logging
from pyspark.sql import SparkSession

spark = SparkSession \
            .builder \
            .appName("mytest") \
            .getOrCreate()

jvm = spark._jvm
jsc = spark._jsc
fs = jvm.org.apache.hadoop.fs.FileSystem.get(jsc.hadoopConfiguration())

ori_path = "/user/leon/overwrite/"
tmp_path = "/user/leon/overwrite_tmp/"

df = spark.read.csv(ori_path)
df.write.mode("overwrite").csv(tmp_path)

df = spark.read.csv(tmp_path)
df.write.mode("overwrite").csv(ori_path)

fs.delete(spark.sparkContext._jvm.org.apache.hadoop.fs.Path(tmp_path), True)

3. 结论

spark读hive表并回写hive表，spark读hdfs路径并回写hdfs路径，均会抛出异常，前者数据还在，后者数据将会丢失。回写hive表，需要借助临时表；回写hdfs路径，需要借助临时目录。