赞
踩
spark.default.parallelism=1000
spark.sql.shuffle.partitions=1000
spark.rpc.askTimeout=30000
spark.speculation=true
适当调整推测执行的参数,可以使任务跑的更快
spark.speculation.interval=500
spark.speculation.quantile=0.85
spark.speculation.multiplie=1.6
spark.sql.files.openCostInBytes=33554432 // 32M,默认是4M,用于小文件合并
spark.sql.files.maxPartitionBytes=268435456 // 256M,默认是64M,每个分区最大的文件大小,针对于大文件切分
spark.blacklist.enabled=true
spark.driver.maxResultSize=10g
spark.yarn.maxAppAttempts=1
spark.hadoop.mapreduce.input.fileinputformat.split.minsize=10240000
spark.hadoop.mapreduce.input.combinefileinputformat.split.minsize=10240000
spark.mapreduce.input.fileinputformat.split.maxsize=51200000
spark.mapreduce.input.fileinputformat.split.minsize=51200000
spark.driver.userClassPathFirst=false
spark.executor.userClassPathFirst=false
spark.driver.extraClassPath=__app__.jar
spark.executor.extraClassPath=__app__.jar
spark.dynamicAllocation.enabled=false
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。