Hadoop Resource Monitoring __java multithreading

Within Hadoop is a resource-monitoring module that includes the following two:


Metrics* and Jmx/jmxjsonservlet

In fact, notice that in the Metrics2/util/mbeans, there is a sentence:
Final Mbeanserver mbs = Managementfactory.getplatformmbeanserver ();
It is also observed to comply with the JMX standard.

Simply put, jmx/jmxjsonservlet through the built-in jetty access to JMX Mbeanserver (its init () function proves it)

METRICS2 is also the same when it starts.


Users can do resource monitoring and information collection by accessing Jmx/jmxjsonservlet or by collecting adapter under METRICS2.

In Conf/hadoop-env, there are:
Export hadoop_namenode_opts= "-dcom.sun.management.jmxremote $HADOOP _namenode_opts"
Export hadoop_secondarynamenode_opts= "-dcom.sun.management.jmxremote $HADOOP _secondarynamenode_opts"
Export hadoop_datanode_opts= "-dcom.sun.management.jmxremote $HADOOP _datanode_opts"
Export hadoop_balancer_opts= "-dcom.sun.management.jmxremote $HADOOP _balancer_opts"
Export hadoop_jobtracker_opts= "-dcom.sun.management.jmxremote $HADOOP _jobtracker_opts"
You can modify the remote port opening for classes in Hadoop.

Conf/hadoop-metrics2.properties can be configured for export of resource monitoring information, such as ganglia.