1. Service RPC port
NameNode uses port 8020 by default to listen to all RPC requests (HDP version), including client data requests, DataNode heartbeat and block reporting, ZKFC module monitoring and inspection And switch control. As the cluster size increases and the RPC request comes larger, the RPC request response time becomes longer and longer.
NameNode also provides DataNode with a dedicated Service RPC port for reporting data blocks and heartbeat status, which can reduce competition with client applications RPC queue requests. This port is not opened by default, and you need to configure the parameters manually.
Service RPC port is also used for ZKFC health check and auto failover control of NameNode HA module.
2. Parameter configuration and HA cluster operation steps
1. For clusters without NameNode HA, the parameter configuration is as follows,
dfs.namenode.servicerpc-address
namenode-host:8040
2. For the cluster with HA enabled on the NameNode, the configuration is as follows,
dfs.namenode.servicerpc-address.test.nn1
test1st.com:8040
dfs.namenode.servicerpc-address.test.nn2
test2st.com:8040
3. Steps to restart the node to make the parameters take effect
< span style="color: #000000;">The entire operation steps are as follows,
1) Restart standby namenode
2) Restart active namenode
3) rolling restart datanode,
4)stop ZKFC
5) reformat ZKFC and start zkfc
hdfs zkfc -formatZK
Note:< span> Changing the service RPC port settings requires restarting NameNodes, DataNodes and ZooKeeper failover controllers to fully take effect. If you have a NameNode HA setup, you can restart the NameNode one at a time, and then restart the remaining components to avoid cluster downtime.
If it is a generation environment, pay more attention to it. After I added the above configuration to the hdfs-site.xml file in the generation environment, I haven’t restarted it yet. Cluster, it is found that the configuration file has taken effect, causing both namenodes to hang up,,, it is best to modify one namenode first, and then modify the other after successful restart, otherwise it may cause the company’s business to be indescribable problem. 0.0
The original link: Hortonworks community
< p>
dfs.namenode.servicerpc-address
namenode-host:8040
dfs.namenode.servicerpc-address.test.nn1
test1st.com:8040
dfs.namenode.servicerpc-address.test.nn2
test2st.com:8040
The entire operation steps are as follows,
1) Restart standby namenode
2) Restart active namenode
3) rolling restart datanode,
4)stop ZKFC
5) reformat ZKFC and start zkfc
hdfs zkfc -formatZK