Use OpenTelemetry Java Agent to expose JVM monitoring metrics¶
In Opentelemetry Agent v1.20.0 and above, Opentelemetry Agent has added the JMX Metric Insight module.
It collects and exposes metrics by instrumenting the metrics exposed by MBeans locally available in the application.
It also has some built-in monitoring samples for common Java Servers or frameworks, please refer to predefined metrics.
If your application has integrated Opentelemetry Agent to collect application traces, then you no longer need to introduce other Agents to expose JMX metrics for our application.
Following the below example to start collect your application JVM metrics:
apiVersion: apps/v1
kind: Deployment
metadata:
name: my-app
labels:
app: my-app
spec:
selector:
matchLabels:
app: my-app
replicas: 1
template:
metadata:
labels:
app: my-app
annotations:
instrumentation.opentelemetry.io/inject-java: "insight-system/insight-opentelemetry-autoinstrumentation" # marking this pod will automatically integrate Opentelemetry Agent
# the below key-values tell prometheus operator how to scrape metrics
insight.opentelemetry.io/metric-scrape: "true" # whether to collect
insight.opentelemetry.io/metric-path: "/" # path to collect metrics
insight.opentelemetry.io/metric-port: "9464" # port for collecting metrics
spec:
containers:
- name: myapp
image: jaegertracing/vertx-create-span:operator-e2e-tests
ports:
- containerPort: 8080
protocol: TCP
- containerPort: 9464 # the pod need expose 9464 port and the relate SVC also need it
protocol: TCP
If there is a port conflict, you can specify a new metric port for your application:- Dotel.exporter.prometheus.port=8464， After the modification, the port values at other locations should also be modified.
Expose metrics for Java middleware¶
Opentelemetry Agent also has some built-in middleware monitoring samples, please refer to Predefined Metrics.
By default, no type is specified, and it needs to be specified through -Dotel.jmx.target.system JVM Options, such as -Dotel.jmx.target.system=jetty,kafka-broker .