|
This version is still in development and is not considered stable yet. For the latest stable version, please use Micrometer 1.14.6! |
JVM Metrics
Micrometer provides several binders for monitoring the JVM:
new ClassLoaderMetrics().bindTo(registry); (1)
new JvmMemoryMetrics().bindTo(registry); (2)
new JvmGcMetrics().bindTo(registry); (3)
new ProcessorMetrics().bindTo(registry); (4)
new JvmThreadMetrics().bindTo(registry); (5)
new JvmThreadDeadlockMetrics().bindTo(registry); (6)| 1 | Gauges loaded and unloaded classes. |
| 2 | Gauges buffer and memory pool utilization. |
| 3 | Gauges max and live data size, promotion and allocation rates, and the number of times the GC pauses (or concurrent phase time in the case of CMS). |
| 4 | Gauges current CPU total and load average. |
| 5 | Gauges thread peak, the number of daemon threads, and live threads. |
| 6 | Gauges the number of threads that are deadlocked. |
Micrometer also provides a meter binder for ExecutorService. You can instrument your ExecutorService, as follows:
new ExecutorServiceMetrics(executor, executorServiceName, tags).bindTo(registry);Metrics created from the binder vary based on the type of ExecutorService.
For ThreadPoolExecutor, the following metrics are provided:
-
executor.completed(FunctionCounter): The approximate total number of tasks that have completed execution. -
executor.active(Gauge): The approximate number of threads that are actively executing tasks. -
executor.queued(Gauge): The approximate number of tasks that are queued for execution. -
executor.pool.size(Gauge): The current number of threads in the pool.
For ForkJoinPool, the following metrics are provided:
-
executor.steals(FunctionCounter): Estimate of the total number of tasks stolen from one thread’s work queue by another. The reported value underestimates the actual total number of steals when the pool is not quiescent. -
executor.queued(Gauge): An estimate of the total number of tasks currently held in queues by worker threads. -
executor.active(Gauge): An estimate of the number of threads that are currently stealing or running tasks. -
executor.running(Gauge): An estimate of the number of worker threads that are not blocked but are waiting to join tasks or for other managed synchronization threads. -
executor.parallelism(Gauge): The targeted parallelism level of this pool. -
executor.pool.size(Gauge): The current number of threads in the pool.
To use the following ExecutorService instances, --add-opens java.base/java.util.concurrent=ALL-UNNAMED is required:
-
Executors.newSingleThreadScheduledExecutor() -
Executors.newSingleThreadExecutor() -
Executors.newThreadPerTaskExecutor() -
Executors.newVirtualThreadPerTaskExecutor()
Java 21 Metrics
Virtual Threads
Micrometer provides metrics for virtual threads released in Java 21. In order to utilize it, you need to add the io.micrometer:micrometer-java21 dependency to your classpath to use the binder:
new VirtualThreadMetrics().bindTo(registry);The binder measures the duration (and counts the number of events) of virtual threads being pinned; it also counts the number of events when starting or unparking a virtual thread failed.
If you are running your application with Java 24 or later on a JVM that has jdk.management.VirtualThreadSchedulerMXBean provided as a platform MXBean, the following additional virtual thread metrics will be provided.
| Meter name | Type | Tag(s) | Description |
|---|---|---|---|
|
Gauge |
Virtual thread scheduler’s target parallelism |
|
|
Gauge |
Current number of platform threads that the scheduler has started but have not terminated; -1 if not known. |
|
|
Gauge |
|
Approximate current number of virtual threads that are unfinished and mounted to a platform thread by the scheduler |
|
Gauge |
|
Approximate current number of virtual threads that are unfinished and queued waiting to be scheduled |
Note that aggregating the values of jvm.threads.virtual.live across the different tags gives the total number of virtual threads started but not ended.
