spark on yarn cluster模式，出現異常NoSuchMethodError: com.google.common.base.Stopwatch

異常出現

調用的命令如下：

spark2-submit \
--master yarn \
--deploy-mode cluster \
--class com.bigdata.PreWarningScalaAppV2 \
--jars /var/lib/hadoop-hdfs/converter-moshi-2.1.0.jar,/var/lib/hadoop-hdfs/fastjson-1.2.58.jar,/var/lib/hadoop-hdfs/guava-20.0.jar,/var/lib/hadoop-hdfs/influxdb-java-2.5.jar,file:/var/lib/hadoop-hdfs/kafka-clients-2.0.0.jar,file:/var/lib/hadoop-hdfs/logging-interceptor-3.5.0.jar,file:/var/lib/hadoop-hdfs/moshi-1.2.0.jar,file:/var/lib/hadoop-hdfs/okhttp-3.5.0.jar,file:/var/lib/hadoop-hdfs/okio-1.11.0.jar,file:/var/lib/hadoop-hdfs/retrofit-2.1.0.jar,file:/var/lib/hadoop-hdfs/spark-streaming-kafka-0-10_2.11-2.4.4.jar,file:/var/lib/hadoop-hdfs/mysql-connector-java-5.1.48.jar \
/var/lib/hadoop-hdfs/prewarning-1.0.jar

查看日誌，異常如下：

一般這種情況是版本衝突導致的

版本查看

我們的版本是20.0

而系統環境自帶的guava版本是11.0

說明用的是自帶的11.0版本，低版本里可能沒有這個新的類Stopwatch
所以我們要讓系統用我們的jar包

解決問題

添加如下的配置

--conf "spark.driver.userClassPathFirst=true" \

官網的說明：是否使用用戶添加的jars

最後執行命令

spark2-submit \
--master yarn \
--deploy-mode cluster \
--class com.bigdata.PreWarningScalaAppV2 \
--jars /var/lib/hadoop-hdfs/converter-moshi-2.1.0.jar,/var/lib/hadoop-hdfs/fastjson-1.2.58.jar,/var/lib/hadoop-hdfs/guava-20.0.jar,/var/lib/hadoop-hdfs/influxdb-java-2.5.jar,file:/var/lib/hadoop-hdfs/kafka-clients-2.0.0.jar,file:/var/lib/hadoop-hdfs/logging-interceptor-3.5.0.jar,file:/var/lib/hadoop-hdfs/moshi-1.2.0.jar,file:/var/lib/hadoop-hdfs/okhttp-3.5.0.jar,file:/var/lib/hadoop-hdfs/okio-1.11.0.jar,file:/var/lib/hadoop-hdfs/retrofit-2.1.0.jar,file:/var/lib/hadoop-hdfs/spark-streaming-kafka-0-10_2.11-2.4.4.jar,file:/var/lib/hadoop-hdfs/mysql-connector-java-5.1.48.jar \
--conf "spark.driver.userClassPathFirst=true" \
/var/lib/hadoop-hdfs/prewarning-1.0.jar

效果如下，運行ok：

CDH也顯示一直運行了