JMX health rules establish the health status of entities in a monitored Java application based on JMX metrics.
JMX Instance Names
For health rules based on JMX metrics, you can create health rules on an a node or on an entity called a JMX instance name. If you create the health rule on the JMX instance name, you have the option of restricting it to specific nodes. If you create the health rule on the node, you have the option of restricting it to specific JMX instance names.
A JMX metric is identified by its JMX metric path, its JMX instance name, and its metric name. These are determined by the "Metric Path", "Object Name Match Pattern", and "Metric Name" fields specified in the AppDynamics JMX metric MBean configuration screens. The instance name is derived from the value returned by the object match pattern. Identical JMX instance name values are considered distinct when their metric paths differ.
JMX instance names appear in the Metric Browser under the appropriate JMX metric path. Specific JMX metrics are reported under each JMX instance name.
For details about how JMX metrics are configured, see Configure JMX Metrics from MBeans.
Creating JMX Health Rules
To create a health rule on one or more JMX metrics, in the Overview panel of the health rule wizard set the type of the rule to Node Health-JMX (connection pools,thread pools, etc).
Then in the Affects panel, select whether the affected entity should be the JMX instance name or the node. In either case, you can configure the rule to cover all JMX instance names in the application or specific available JMX instance names.
In some IT organizations, different teams are responsible for different MBeans. In other organizations, different teams are responsible for different nodes or tiers.
The way your organization is set up determines how you think about JMX health, especially where it intersects with node health.
- As a team member responsible for a node, do you consider your node unhealthy because one or more of its MBeans is unhealthy? If so, which nodes or how many?
- As a team member responsible for your JMX infrastructure, are you primarily interested in the health of your JMX data regardless of the nodes that use it? Or are you interested in just specific JMX metrics in specific nodes? Which ones?
For the purpose of configuring JMX health rules and using them in policies, the ultimate question is: who will receive the alert when the agent reports unhealthy performance detected by JMX metrics? The flexibility of the AppDynamics healht rules lets you fine-tune your JMX health rules so that the right people are alerted for the code for which they are responsible.
JMX Health Rules Affecting a Node
In one organization, teams are responsible for nodes. Mark is responsible for WEB1_NODE, Tao for WEB2_NODE, and so on.
If an MBean in WEB1_NODE generates JMX metrics that violate a critical condition, and a health rule is configured to evaluate a JMX object in that node, Mark or someone on Mark's team will get the alert. The configuration of the health rule would be:
If different people on Mark's team are responsible for different MBeans used in WEB1_NODE, they could refine the rule further by selecting specific JMX instance names for evaluation. For example, the last decision of this configuration can restrict the rule to evaluate only metrics in the "jdbc/cartDS" JMX instance name:
Mark's team could create a similar rule for the remaining JMX instance names to use in a policy that alerts whoever on Mark's team is responsible for the "jdbc/ECommerceDB" and "jdbc/OracleECommerceDB" JMX instance names.
Mark's team could also create different rules that evaluate a different set of JMX metrics by choosing a different metric path in the Select JMX Objects field. For example, they could select All Web Container Runtimes.
For problems on WEB2_NODE, they would create a different health rule with WEB2_NODE as the affected node, so that the alerts for those problems go to Tao's team.
JMX Health Rules Affecting a JMX Instance Name
In another organization, teams are responsible for various parts of the JMX infrastructure. Mary is responsible for "jdbc/cartDS", Meera for "jdbc/ECommerceDB", and so on, regardless of which nodes use these MBeans.
So if metrics in "jdbc/cartDS" violate a critical condition, they want Mary to get the alert because her team is responsible. The configuration of the health rule would be:
The rule could be refined to evaluate the JMX metrics in the specified JMX instance names in all nodes in the application or only in specific nodes.