yarn中有一個比較重要的配置yarn.nodemanager.local-dirs,若是配置的很差,在飽和狀態運行下集羣會出現不少問題:
1 默認配置${hadoop.tmp.dir}/nm-local-dir,系統盤一般只有幾百G,配置在這裏會常常報磁盤空間不足的錯誤;
2 配置到其中1個數據盤,好比/data0/,這個數據盤會常常沒有響應,或者直接壞掉;
3 正確的配置是配置到全部的數據盤,一般是12個,其實yarn的description寫的很清楚node
<property>app
<description>List of directories to store localized files in. Anoop
application's localized file directory will be found in:this
${yarn.nodemanager.local-dirs}/usercache/${user}/appcache/application_${appid}.blog
Individual containers' work directories, called container_${contid}, willip
be subdirectories of this.hadoop
</description>io
<name>yarn.nodemanager.local-dirs</name>集羣
<value>${hadoop.tmp.dir}/nm-local-dir</value>配置
</property>
<property>
<name>hadoop.tmp.dir</name>
<value>/tmp/hadoop-tmp</value>
<description>A base for other temporary directories.</description>
</property>
CDH建議