5 Discovery of JMX objects

Overview

It is possible to discover all JMX MBeans or MBean attributes or to specify a pattern for the discovery of these objects.

It is mandatory to understand the difference between an Mbean and Mbean attributes for discovery rule configuration. An MBean is an object which can represent a device, an application, or any resource that needs to be managed.

For example, there is an Mbean which represents a web server. Its attributes are connection count, thread count, request timeout, http file cache, memory usage, etc. Expressing this thought in human comprehensive language we can define a coffee machine as an Mbean which has the following attributes to be monitored: water amount per cup, average consumption of water for a certain period of time, number of coffee beans required per cup, coffee beans and water refill time, etc.

Item key

In discovery rule configuration, select JMX agent in the Type field.

Two item keys are supported for JMX object discovery - jmx.discovery[] and jmx.get[]:

If no parameters are passed, all MBean attributes from JMX are requested. Not specifying parameters for JMX discovery or trying to receive all attributes for a wide range like *:type=*,name=* may lead to potential performance problems.

Using jmx.discovery

This item returns a JSON object with low-level discovery macros describing MBean objects or attributes. For example, in the discovery of MBean attributes (reformatted for clarity):

[
    {
        "{#JMXVALUE}":"0",
        "{#JMXTYPE}":"java.lang.Long",
        "{#JMXOBJ}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXDESC}":"java.lang:type=GarbageCollector,name=PS Scavenge,CollectionCount",
        "{#JMXATTR}":"CollectionCount"
    },
    {
        "{#JMXVALUE}":"0",
        "{#JMXTYPE}":"java.lang.Long",
        "{#JMXOBJ}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXDESC}":"java.lang:type=GarbageCollector,name=PS Scavenge,CollectionTime",
        "{#JMXATTR}":"CollectionTime"
    },
    {
        "{#JMXVALUE}":"true",
        "{#JMXTYPE}":"java.lang.Boolean",
        "{#JMXOBJ}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXDESC}":"java.lang:type=GarbageCollector,name=PS Scavenge,Valid",
        "{#JMXATTR}":"Valid"
    },
    {
        "{#JMXVALUE}":"PS Scavenge",
        "{#JMXTYPE}":"java.lang.String",
        "{#JMXOBJ}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXDESC}":"java.lang:type=GarbageCollector,name=PS Scavenge,Name",
        "{#JMXATTR}":"Name"
    },
    {
        "{#JMXVALUE}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXTYPE}":"javax.management.ObjectName",
        "{#JMXOBJ}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXDESC}":"java.lang:type=GarbageCollector,name=PS Scavenge,ObjectName",
        "{#JMXATTR}":"ObjectName"
    }
]

In the discovery of MBeans (reformatted for clarity):

[
    {
        "{#JMXDOMAIN}":"java.lang",
        "{#JMXTYPE}":"GarbageCollector",
        "{#JMXOBJ}":"java.lang:type=GarbageCollector,name=PS Scavenge",
        "{#JMXNAME}":"PS Scavenge"
    }
]

Supported macros

The following macros are supported for use in the discovery rule filter and prototypes of items, triggers and graphs:

Limitations

There are some limitations associated with the algorithm of creating LLD macro names from MBean property names:

• attribute names are changed to uppercase

• attribute names are ignored (no LLD macros are generated) if they consist of unsupported characters for LLD macro names. Supported characters can be described by the following regular expression: A-Z0-9_\.

• if an attribute is called “obj” or “domain” they will be ignored because of the overlap with the values of the reserved Zabbix properties {#JMXOBJ} and {#JMXDOMAIN} (supported since Zabbix 3.4.3.)

Please consider this jmx.discovery (with “beans” mode) example. MBean has the following properties defined:

name=test
тип=Type
attributes []=1,2,3
Name=NameOfTheTest
domAin=some

As a result of JMX discovery, the following LLD macros will be generated:

• {#JMXDOMAIN} - Zabbix internal, describing the domain of MBean

• {#JMXOBJ} - Zabbix internal, describing MBean object

• {#JMXNAME} - created from “name” property

Ignored properties are:

• тип : its name contains unsupported characters (non-ASCII)

• attributes[] : its name contains unsupported characters (square brackets are not supported)

• Name : it’s already defined (name=test)

• domAin : it’s a Zabbix reserved name

Examples

Let’s review two more practical examples of a LLD rule creation with the use of Mbean. To understand the difference between a LLD rule collecting Mbeans and a LLD rule collecting Mbean attributes better please take a look at following table:

Example 1: Discovering Mbeans

This rule will return 3 objects: the top row of the column: MBean1, MBean2, MBean3.

For more information about objects please refer to supported macros table, Discovery of MBeans section.

Discovery rule configuration collecting Mbeans (without the attributes) looks like the following:

Key used:

jmx.discovery[beans,"*:type=GarbageCollector,name=*"]

All the garbage collectors without attributes will be discovered. As Garbage collectors have the same attribute set, we can use desired attributes in item prototypes the following way:

Keys used:

jmx[{#JMXOBJ},CollectionCount] 
jmx[{#JMXOBJ},CollectionTime] 
jmx[{#JMXOBJ},Valid]

LLD discovery rule will result in something close to this (items are discovered for two Garbage collectors):

Example 2: Discovering Mbean attributes

This rule will return 9 objects with the following fields: MBean1Attribute1, MBean2Attribute1, Mbean3Attribute1,MBean1Attribute2,MBean2Attribute2, Mbean3Attribute2, MBean1Attribute3, MBean2Attribute3, Mbean3Attribute3.

For more information about objects please refer to supported macros table, Discovery of MBean attributes section.

Discovery rule configuration collecting Mbean attributes looks like the following:

Key used:

jmx.discovery[attributes,"*:type=GarbageCollector,name=*"]

All the garbage collectors with a single item attribute will be discovered.

In this particular case an item will be created from prototype for every MBean attribute. The main drawback of this configuration is that trigger creation from trigger prototypes is impossible as there is only one item prototype for all attributes. So this setup can be used for data collection, but is not recommended for automatic monitoring.

Using jmx.get

jmx.get[] is similar to the jmx.discovery[] item, but is does not turn Java object properties into low-level discovery macro names and therefore can return values without limitations that are associated with LLD macro name generation such as hyphens or non-ASCII characters.

When using jmx.get[] for discovery, low-level discovery macros can be defined separately in the custom LLD macro tab of the discovery rule configuration, using JSONPath to point to the required values.

Discovering MBeans

Discovery item: jmx.get[beans,"com.example:type=*,*"]

Response:

[
    {
        "object": "com.example:type=Hello,data-src=data-base,ключ=значение",
        "domain": "com.example",
        "properties": {
            "data-src": "data-base",
            "ключ": "значение",
            "type": "Hello"
        }
    },
    {
        "object": "com.example:type=Atomic",
        "domain": "com.example",
        "properties": {
            "type": "Atomic"
        }
    }
]

Discovering MBean attributes

Discovery item: jmx.get[attributes,"com.example:type=*,*"]

Response:

[
    {
        "object": "com.example:type=*",
        "domain": "com.example",
        "properties": {
            "type": "Simple"
        }
    },
    {
        "object": "com.zabbix:type=yes,domain=zabbix.com,data-source=/dev/rand,ключ=значение,obj=true",
        "domain": "com.zabbix",
        "properties": {
            "type": "Hello",
            "domain": "com.example",
            "data-source": "/dev/rand",
            "ключ": "значение",
            "obj": true
        }
    }
]