spark on yarn日志切割
由于spark on yarn的日志会在stderr里面一直追加，而且streaming又是长应用，那么stderr会一直增长。解决办法就是把stderr按天切割，这样我们就可以删除之前的日志了。
1. 首先，把集群中的hadoop的log4j文件/etc/hadoop/conf/log4j.properties添加如下两行：
log4j.additivity.com.kingsoft = false com.kingsoft是项目的包名
log4j.additivity.org.apache.spark = false
然后把log4j.appender.console.layout.ConversionPattern=%d{yy/MM/dd HH:mm:ss} %p %c{2}: %m%n的%c{2}改成%c{30}
最终的结果如下：
log4j.rootLogger=${root.logger}
root.logger=INFO,console
log4j.appender.console=org.apache.log4j.ConsoleAppender
log4j.appender.console.target=System.err
log4j.appender.console.layout=org.apache.log4j.PatternLayout
log4j.appender.console.layout.ConversionPattern=%d{yy/MM/dd HH:mm:ss} %p %c{30}: %m%n
log4j.additivity.com.kingsoft = false
log4j.additivity.org.apache.spark = false
2. 编写一个 log4j.properties 文件，内容如下：
log4j.rootCategory=INFO, click_all
# Settings to quiet third party logs that are too verbose
log4j.logger.org.apache.spark.repl.SparkIMain$exprTyper=INFO
log4j.logger.org.apache.spark.repl.SparkILoop$SparkILoopInterpreter=INFO
log4j.appender.click_all = org.apache.log4j.DailyRollingFileAppender
log4j.appender.click_all.File = /sparklogs/click_all.log
log4j.appender.click_all.DatePattern = '.'yyyy-MM-dd_HH
log4j.appender.click_all.Append = true
log4j.appender.click_all.Threshold = INFO
log4j.appender.click_all.layout = org.apache.log4j.PatternLayout
log4j.appender.click_all.layout.ConversionPattern = %d{yy/MM/dd HH:mm:ss} %p %c{30}: %m%n
log4j.additivity.com._58city = false
log4j.logger.com._58city = INFO, click
log4j.appender.click = org.apache.log4j.DailyRollingFileAppender
log4j.appender.click.File = /sparklogs/click.log
log4j.appender.click.DatePattern = '.'yyyy-MM-dd_HH
log4j.appender.click.Append = true
log4j.appender.click.Threshold = INFO
log4j.appender.click.layout = org.apache.log4j.PatternLayout
log4j.appender.click.layout.ConversionPattern = %d{yy/MM/dd HH:mm:ss} %p %c{30}: %m%n