Semantic conventions for JVM metrics
Status: Mixed
This document describes semantic conventions for JVM metrics in OpenTelemetry.
- JVM memory
- JVM garbage collection
- JVM threads
- JVM classes
- JVM CPU
- Development
- Release Candidate
JVM memory
Status: Stable
Description: Java Virtual Machine (JVM) metrics captured under the namespace jvm.memory.*
Metric: jvm.memory.used
This metric is recommended.
This metric is obtained from MemoryPoolMXBean#getUsage().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.memory.used | UpDownCounter | By | Measure of memory used. |  |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.memory.pool.name | | Recommended | string | Name of the memory pool. [1] | G1 Old Gen; G1 Eden space; G1 Survivor Space |
jvm.memory.type | | Recommended | string | The type of memory. | heap; non_heap (see more) |
[1] jvm.memory.pool.name: Pool names are generally obtained via MemoryPoolMXBean#getName().
jvm.memory.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
heap | Heap memory. | |
non_heap | Non-heap memory | |
Metric: jvm.memory.committed
This metric is recommended.
This metric is obtained from MemoryPoolMXBean#getUsage().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.memory.committed | UpDownCounter | By | Measure of memory committed. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.memory.pool.name | | Recommended | string | Name of the memory pool. [1] | G1 Old Gen; G1 Eden space; G1 Survivor Space |
jvm.memory.type | | Recommended | string | The type of memory. | heap; non_heap (see more) |
[1] jvm.memory.pool.name: Pool names are generally obtained via MemoryPoolMXBean#getName().
jvm.memory.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
heap | Heap memory. | |
non_heap | Non-heap memory | |
Metric: jvm.memory.limit
This metric is recommended.
This metric is obtained from MemoryPoolMXBean#getUsage().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.memory.limit | UpDownCounter | By | Measure of max obtainable memory. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.memory.pool.name | | Recommended | string | Name of the memory pool. [1] | G1 Old Gen; G1 Eden space; G1 Survivor Space |
jvm.memory.type | | Recommended | string | The type of memory. | heap; non_heap (see more) |
[1] jvm.memory.pool.name: Pool names are generally obtained via MemoryPoolMXBean#getName().
jvm.memory.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
heap | Heap memory. | |
non_heap | Non-heap memory | |
Metric: jvm.memory.used_after_last_gc
This metric is recommended.
This metric is obtained from MemoryPoolMXBean#getCollectionUsage().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.memory.used_after_last_gc | UpDownCounter | By | Measure of memory used, as measured after the most recent garbage collection event on this pool. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.memory.pool.name | | Recommended | string | Name of the memory pool. [1] | G1 Old Gen; G1 Eden space; G1 Survivor Space |
jvm.memory.type | | Recommended | string | The type of memory. | heap; non_heap (see more) |
[1] jvm.memory.pool.name: Pool names are generally obtained via MemoryPoolMXBean#getName().
jvm.memory.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
heap | Heap memory. | |
non_heap | Non-heap memory | |
JVM garbage collection
Status: Stable
Description: Java Virtual Machine (JVM) metrics captured under the namespace jvm.gc.*
Metric: jvm.gc.duration
This metric is recommended.
This metric is obtained by subscribing to
GarbageCollectionNotificationInfo events provided by GarbageCollectorMXBean. The duration value is obtained from GcInfo
This metric SHOULD be specified with ExplicitBucketBoundaries advisory parameter of [ 0.01, 0.1, 1, 10 ].
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.gc.duration | Histogram | s | Duration of JVM garbage collection actions. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.gc.action | | Recommended | string | Name of the garbage collector action. [1] | end of minor GC; end of major GC |
jvm.gc.name | | Recommended | string | Name of the garbage collector. [2] | G1 Young Generation; G1 Old Generation |
jvm.gc.cause |  | Opt-In | string | Name of the garbage collector cause. [3] | System.gc(); Allocation Failure |
[1] jvm.gc.action: Garbage collector action is generally obtained via GarbageCollectionNotificationInfo#getGcAction().
[2] jvm.gc.name: Garbage collector name is generally obtained via GarbageCollectionNotificationInfo#getGcName().
[3] jvm.gc.cause: Garbage collector cause is generally obtained via GarbageCollectionNotificationInfo#getGcCause().
JVM threads
Status: Stable
Description: Java Virtual Machine (JVM) metrics captured under the namespace jvm.thread.*
Metric: jvm.thread.count
This metric is recommended. This metric is obtained from a combination of
ThreadMXBean#getAllThreadIds()ThreadMXBean#getThreadInfo()ThreadInfo#getThreadState()ThreadInfo#isDaemon()(requires Java 9+)
Note that this is the number of platform threads (as opposed to virtual threads).
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.thread.count | UpDownCounter | {thread} | Number of executing platform threads. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.thread.daemon | | Recommended | boolean | Whether the thread is daemon or not. | |
jvm.thread.state | | Recommended | string | State of the thread. | runnable; blocked (see more) |
jvm.thread.state has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
blocked | A thread that is blocked waiting for a monitor lock is in this state. | |
new | A thread that has not yet started is in this state. | |
runnable | A thread executing in the Java virtual machine is in this state. | |
terminated | A thread that has exited is in this state. | |
timed_waiting | A thread that is waiting for another thread to perform an action for up to a specified waiting time is in this state. | |
waiting | A thread that is waiting indefinitely for another thread to perform a particular action is in this state. | |
JVM classes
Status: Stable
Description: Java Virtual Machine (JVM) metrics captured under the namespace jvm.class.*
Metric: jvm.class.loaded
This metric is recommended.
This metric is obtained from ClassLoadingMXBean#getTotalLoadedClassCount().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.class.loaded | Counter | {class} | Number of classes loaded since JVM start. | |
Metric: jvm.class.unloaded
This metric is recommended.
This metric is obtained from ClassLoadingMXBean#getUnloadedClassCount().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.class.unloaded | Counter | {class} | Number of classes unloaded since JVM start. | |
Metric: jvm.class.count
This metric is recommended.
This metric is obtained from ClassLoadingMXBean#getLoadedClassCount().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.class.count | UpDownCounter | {class} | Number of classes currently loaded. | |
JVM CPU
Status: Stable
Description: Java Virtual Machine (JVM) metrics captured under the namespace jvm.cpu.*
Metric: jvm.cpu.time
This metric is recommended.
This metric is obtained from com.sun.management.OperatingSystemMXBean#getProcessCpuTime() on HotSpot
and com.ibm.lang.management.OperatingSystemMXBean#getProcessCpuTime() on OpenJ9.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.cpu.time | Counter | s | CPU time used by the process as reported by the JVM. | |
Metric: jvm.cpu.count
This metric is recommended.
This metric is obtained from Runtime#availableProcessors().
Note that this is always an integer value (i.e. fractional or millicores are not represented).
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.cpu.count | UpDownCounter | {cpu} | Number of processors available to the Java virtual machine. | |
Metric: jvm.cpu.recent_utilization
This metric is recommended.
This metric is obtained from com.sun.management.OperatingSystemMXBean#getProcessCpuLoad() on HotSpot
and com.ibm.lang.management.OperatingSystemMXBean#getProcessCpuLoad() on OpenJ9.
Note that the JVM does not provide a definition of what “recent” means.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.cpu.recent_utilization | Gauge | 1 | Recent CPU utilization for the process as reported by the JVM. [1] | |
[1]: The value range is [0.0,1.0]. This utilization is not defined as being for the specific interval since last measurement (unlike system.cpu.utilization). Reference.
Development
Status: Development
Description: In-development Java Virtual Machine (JVM) metrics captured under jvm.
JVM executors
Status: Development
Description: JVM metrics describing individual executors under the jvm.executor.* namespace.
Instrumentations SHOULD emit each metric only when the executor implementation exposes the corresponding value, and SHOULD NOT emit an estimated or substituted value in its place.
Metric: jvm.executor.thread.count
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.thread.count | UpDownCounter | {thread} | The number of executor threads that are currently in the state described by the jvm.executor.thread.state attribute. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.thread.state | | Required | string | State of an executor thread. | active; idle (see more) |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
jvm.executor.thread.state has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
active | The thread is currently executing a task. | |
idle | The thread currently exists but is not executing a task. | |
Metric: jvm.executor.thread.core
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.thread.core | UpDownCounter | {thread} | The number of core threads configured for the executor. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
Metric: jvm.executor.thread.limit
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.thread.limit | UpDownCounter | {thread} | The maximum number of threads allowed for the executor. [1] | |
[1]: Instrumentations SHOULD emit this metric only when the maximum thread count is finite and represents a meaningful operational limit. This metric SHOULD NOT be emitted for unbounded or effectively unbounded executors. For example, an executor with a maximum thread count of Integer.MAX_VALUE is effectively unbounded.
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
Metric: jvm.executor.queue.size
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.queue.size | UpDownCounter | {task} | The number of tasks currently queued for execution. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
Metric: jvm.executor.queue.capacity
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.queue.capacity | UpDownCounter | {task} | The maximum number of tasks the executor queue can hold. [1] | |
[1]: Instrumentations SHOULD emit this metric only when the queue has a finite capacity that represents a meaningful operational limit. This metric SHOULD NOT be emitted for unbounded or effectively unbounded queues.
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
Metric: jvm.executor.task.completed
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.task.completed | Counter | {task} | The number of tasks completed by the executor. [1] | |
[1]: A task SHOULD be counted as completed when the executor has finished processing it, regardless of whether it returns normally, terminates with an exception, or is canceled. Cancellation during processing does not complete the task until processing ends. A task that never begins processing on the executor SHOULD NOT be counted. A task initially rejected but subsequently resubmitted to and processed by the executor SHOULD be counted when processing ends.
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
Metric: jvm.executor.task.rejected
This metric is recommended.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.executor.task.rejected | Counter | {task} | The number of task submissions rejected by the executor. [1] | |
[1]: A task submission SHOULD be counted as rejected when the executor invokes its rejection handler, regardless of whether the handler subsequently executes the task in the caller (for example, CallerRunsPolicy) or discards or replaces a queued task (for example, DiscardOldestPolicy). This metric counts rejected submissions, not tasks ultimately discarded from the queue.
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.executor.name | | Required | string | Name of the executor. [1] | pool-1-thread-*; ForkJoinPool.commonPool-worker-*; orders-vt-* |
jvm.executor.type | | Required | string | The fully qualified class name of the executor implementation. | java.util.concurrent.ThreadPoolExecutor; java.util.concurrent.ForkJoinPool |
jvm.executor.owner.name | | Recommended | string | Name of the logical component that owns the executor. [2] | tomcat; jetty; netty |
[1] jvm.executor.name: The executor name SHOULD be low-cardinality. If jvm.executor.owner.name is present, the combination of jvm.executor.owner.name and jvm.executor.name SHOULD be unique within the instrumented application; otherwise, jvm.executor.name SHOULD be unique when an executor-specific name is available. If the executor implementation does not provide an explicit name, instrumentation SHOULD derive a name from available implementation-specific metadata, such as a normalized worker thread name. If no executor-specific name or metadata is available, instrumentation MAY use unknown. The unknown value does not identify an executor, and metrics from multiple executors using this value MAY be aggregated. Users requiring per-executor metric series SHOULD configure a distinct name through the executor’s ThreadFactory. Instrumentations SHOULD document how the name is derived and how users can configure it. Once metric series registration for an executor has started, jvm.executor.name and jvm.executor.owner.name MUST remain stable for the lifetime of that registration. Instrumentations SHOULD determine these attributes before the first worker is registered. Instrumentations MUST NOT re-register cumulative metrics under a new attribute set or replay prior cumulative values under new attributes.
[2] jvm.executor.owner.name: The executor owner name SHOULD be a low-cardinality identifier for the component that creates and configures the executor. It is neither the executor implementation type nor an OpenTelemetry instrumentation scope.
Metric: jvm.memory.init
This metric is recommended.
This metric is obtained from MemoryPoolMXBean#getUsage().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.memory.init | UpDownCounter | By | Measure of initial memory requested. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.memory.pool.name | | Recommended | string | Name of the memory pool. [1] | G1 Old Gen; G1 Eden space; G1 Survivor Space |
jvm.memory.type | | Recommended | string | The type of memory. | heap; non_heap (see more) |
[1] jvm.memory.pool.name: Pool names are generally obtained via MemoryPoolMXBean#getName().
jvm.memory.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.
| Value | Description | Stability |
|---|---|---|
heap | Heap memory. | |
non_heap | Non-heap memory | |
Metric: jvm.system.cpu.utilization
This metric is Opt-In.
This metric is obtained from com.sun.management.OperatingSystemMXBean#getSystemCpuLoad() on Java version 8..13, com.sun.management.OperatingSystemMXBean#getCpuLoad() on Java version 14+,
and com.ibm.lang.management.OperatingSystemMXBean#getSystemCpuLoad() on OpenJ9.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.system.cpu.utilization | Gauge | 1 | Recent CPU utilization for the whole system as reported by the JVM. [1] | |
[1]: The value range is [0.0,1.0]. This utilization is not defined as being for the specific interval since last measurement (unlike system.cpu.utilization). Reference.
Metric: jvm.system.cpu.load_1m
This metric is Opt-In.
This metric is obtained from OperatingSystemMXBean#getSystemLoadAverage().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.system.cpu.load_1m | Gauge | {run_queue_item} | Average CPU load of the whole system for the last minute as reported by the JVM. [1] | |
[1]: The value range is [0,n], where n is the number of CPU cores - or a negative number if the value is not available. This utilization is not defined as being for the specific interval since last measurement (unlike system.cpu.utilization). Reference.
Metric: jvm.buffer.memory.used
This metric is recommended.
This metric is obtained from BufferPoolMXBean#getMemoryUsed().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.buffer.memory.used | UpDownCounter | By | Measure of memory used by buffers. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.buffer.pool.name | | Recommended | string | Name of the buffer pool. [1] | mapped; direct |
[1] jvm.buffer.pool.name: Pool names are generally obtained via BufferPoolMXBean#getName().
Metric: jvm.buffer.memory.limit
This metric is recommended.
This metric is obtained from BufferPoolMXBean#getTotalCapacity().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.buffer.memory.limit | UpDownCounter | By | Measure of total memory capacity of buffers. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.buffer.pool.name | | Recommended | string | Name of the buffer pool. [1] | mapped; direct |
[1] jvm.buffer.pool.name: Pool names are generally obtained via BufferPoolMXBean#getName().
Metric: jvm.buffer.count
This metric is recommended.
This metric is obtained from BufferPoolMXBean#getCount().
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.buffer.count | UpDownCounter | {buffer} | Number of buffers in the pool. | |
Attributes:
| Key | Stability | Requirement Level | Value Type | Description | Example Values |
|---|---|---|---|---|---|
jvm.buffer.pool.name | | Recommended | string | Name of the buffer pool. [1] | mapped; direct |
[1] jvm.buffer.pool.name: Pool names are generally obtained via BufferPoolMXBean#getName().
Release Candidate
Status: Release Candidate
Description: Release-candidate Java Virtual Machine (JVM) metrics captured under jvm.
Metric: jvm.file_descriptor.count
This metric is Opt-In.
This metric is obtained from UnixOperatingSystemMXBean#getOpenFileDescriptorCount().
Instrumentations SHOULD emit this metric only when the value is finite and
represents a meaningful operational limit. This metric SHOULD NOT be emitted for non-meaningful values.
For example, a negative value or equal to Long.MAX_VALUE (2^63-1) SHOULD NOT be emitted.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.file_descriptor.count | UpDownCounter | {file_descriptor} | Number of open file descriptors as reported by the JVM. |  |
Metric: jvm.file_descriptor.limit
This metric is Opt-In.
This metric is obtained from UnixOperatingSystemMXBean#getMaxFileDescriptorCount().
Instrumentations SHOULD emit this metric only when the value is finite and
represents a meaningful operational limit. This metric SHOULD NOT be emitted for non-meaningful values.
For example, a negative value or equal to Long.MAX_VALUE (2^63-1) SHOULD NOT be emitted.
| Name | Instrument Type | Unit (UCUM) | Description | Stability | Entity Associations |
|---|---|---|---|---|---|
jvm.file_descriptor.limit | UpDownCounter | {file_descriptor} | Measure of max open file descriptors as reported by the JVM. | |
Feedback
Was this page helpful?
Thank you. Your feedback is appreciated!
Please let us know how we can improve this page. Your feedback is appreciated!