YARN IPC Large response size 问题

YARN报错

2017-08-25 03:51:58,815 WARN org.apache.hadoop.ipc.Server: Large response size 4739374 for call org.apache.hadoop.yarn.api.ApplicationClientProtocolPB.getApplications from 10.135.8.101:38352 Call#33361 Retry#0
2017-08-25 03:53:39,255 WARN org.apache.hadoop.ipc.Server: Large response size 4739374 for call org.apache.hadoop.yarn.api.ApplicationClientProtocolPB.getApplications from 10.135.8.101:38456 Call#33364 Retry#0
2017-08-25 03:55:19,700 WARN org.apache.hadoop.ipc.Server: Large response size 4739374 for call org.apache.hadoop.yarn.api.ApplicationClientProtocolPB.getApplications from 10.135.8.101:38556 Call#33367 Retry#0
2017-08-25 03:57:00,262 WARN org.apache.hadoop.ipc.Server: Large response size 4739374 for call org.apache.hadoop.yarn.api.ApplicationClientProtocolPB.getApplications from 10.135.8.101:38674 Call#33370 Retry#0
2017-08-25 03:58:40,687 WARN org.apache.hadoop.ipc.Server: Large response size 4739374 for call org.apache.hadoop.yarn.api.ApplicationClientProtocolPB.getApplications from 10.135.8.101:38804 Call#33373 Retry#0

 

解决办法：

1、在hdfs-site中添加如下参数

<property>
     <name>ipc.server.max.response.size</name>
     <value>5242880</value>
</property>

2、可能造成OOM问题

增大-xmx参数的大小

 

其他问题

正常来说这里的IPC时间返回大概是10s/1min这个级别，如果返回的太频繁就可能会出现RM OOM的问题。

这个问题需要深入源码去分析，待有结论再更新上来。

 

链接

1、https://mapr.com/community/s/question/0D50L00006BIsu9SAD/yarn-crash-max-number-of-completed-apps-kept-in-memory-met

2、https://issues.apache.org/jira/browse/HADOOP-14858

3、https://mapr.com/community/s/question/0D50L00006BIt35SAD/why-yarn-crashes-

4、https://issues.apache.org/jira/browse/YARN-7150

5、https://www.jianshu.com/p/ce998c10b471