Kamon JMX Collector

JMX metrics collector for Kamon. Compatible with the 1.x branch. It was inspired by the Kamon JMX library that is only compatible with the legacy version of Kamon.

Usage

1. Add the Segence OSS Releases Maven repo to your project, click here for instructions
2. Import the artifact, click here for instructions.
   • SBT example:

     libraryDependencies += "io.kamon" %% "kamon-jmx-collector" % "1.0.0-RC1"

   • Gradle example:

     compile 'io.kamon:kamon-jmx-collector_2.12:1.0.0-RC1'

3. The artifact does not include all dependencies to avoid versioning conflict within your application. The required libraries have to be added.
   • SBT example:

     libraryDependencies += "io.kamon" %% "kamon-core" % "2.0.4"
     libraryDependencies += "com.typesafe" % "config" % "1.3.3"
     libraryDependencies += "com.typesafe.akka" %% "akka-actor" % "2.5.23"

   • Gradle example:

     compile(
     "io.kamon:kamon-core_2.12:2.0.4",
     "com.typesafe:config:1.3.3",
     "com.typesafe.akka:akka-actor_2.12:2.5.23"
     )

     4. Add a configuration to your application.conf file describing the MBeans to query. Example:

        kamon {
        jmx-collector {
        mbeans = [
        {
        "metric-name": "os-mbean",
        "jmxQuery": "java.lang:type=OperatingSystem",
        "attributes": [
          { "name": "AvailableProcessors", "type": "histogram", "keys": [ ] }
        ]
        },
        {
        "metric-name": "os-memory",
        "jmxQuery": "java.lang:type=Memory",
        "attributes": [
          { "name": "HeapMemoryUsage",                "type": "histogram", "keys": [ "committed", "max" ] },
          { "name": "ObjectPendingFinalizationCount", "type": "histogram", "keys": [ ]  }
        ]
        },
        {
        "metric-name": "kafka-consumer",
        "jmxQuery": "kafka.consumer:type=consumer-metrics,client-id=*",
        "attributes": [
          { "name": "connection-count", "type": "histogram", "keys": [ ]  }
        ]
        },
        {
        "metric-name": "kafka-producer-metrics",
        "jmxQuery": "kafka.producer:type=producer-metrics,client-id=*",
        "attributes": [
          { "name": "network-io-rate",    "type": "punctual-gauge", "keys": [ ]  },
          { "name": "outgoing-byte-rate", "type": "punctual-gauge", "keys": [ ]  }
        ]
        },
        {
        "metric-name": "kafka-producer-node-metrics",
        "jmxQuery": "kafka.producer:type=producer-node-metrics,client-id=*,node-id=*",
        "attributes": [
          { "name": "incoming-byte-rate", "type": "punctual-gauge", "keys": [ ]  },
          { "name": "outgoing-byte-rate", "type": "punctual-gauge", "keys": [ ]  }
        ]
        }
        ],
        initial-delay = 1 second,
        value-check-interval = 2 seconds
        }
        }

     5. Launch the metric collector. Example:

        import akka.actor.ActorSystem
        import kamon.Kamon
        ...
        implicit val system = ActorSystem()
        KamonJmxMetricCollector()

Configuration details

The following table explains the different configuration parameters and how to set them:

As JMX metrics are generated in runtime it is advised to set the kamon.jmx-collector.initial-delay parameter to a sensible value, depending on how long it takes for the application to initialise JMX metrics. Failing to define kamon.jmx-collector.mbeans[*].attributes[*].keys for a CompositeData attribute will make it impossible to retrieve metrics.

Kamon metric names are generated in the following format: jmx-${metric-name from configuration}-${attribute name}. Tags are added to the metrics as long as there are wildcard queries in the jmxQuery attribute.

Finding the available JMX queries and attributes

Either a command line tool like jmxterm can be used or a graphical tool like either JConsole or VisualVM. VisualVM needs the MBeans Browser plugin installed to see the available JMX Mbeans like on the image below:

It is visible that there are several attributes available under the java.lang:type=Memory object name. ObjectPendingFinalizationCount is of type int so attribute keys do not needed to be defined in the configuration.

However, HeapMemoryUsage is of type javax.management.openmbean.CompositeData that can not be converted to a numerical data type directly suitable for Kamon metrics; its attribute keys have to be defined.

The field openType contains the following value: javax.management.openmbean.CompositeType(name=java.lang.management.MemoryUsage,items=((itemName=committed,itemType=javax.management.openmbean.SimpleType(name=java.lang.Long)),(itemName=init,itemType=javax.management.openmbean.SimpleType(name=java.lang.Long)),(itemName=max,itemType=javax.management.openmbean.SimpleType(name=java.lang.Long)),(itemName=used,itemType=javax.management.openmbean.SimpleType(name=java.lang.Long))))

This shows us the available attribute keys: committed, init, max and used. Their data type is java.lang.Long that is safe to be used in Kamon metric instruments.

Some of these attribute keys have to be defined in the configuration in order to pass them to Kamon. They have to be added under the parameter kamon.jmx-collector.mbeans[*].attributes[*].keys as a JSON array like ["committed", "init", "max", "used].