# OBI exported metrics > Learn about the application, runtime, and network metrics OBI can export. --- Site [llms.txt](/ro/llms.txt) --- OBI exports application, runtime, and network metrics in OpenTelemetry and Prometheus formats. For metric definitions, including MCP, GPU, and runtime metrics, see the [OBI telemetry catalog](https://github.com/open-telemetry/opentelemetry-ebpf-instrumentation/blob/v1.0.0/site/docs/metrics.md). ## Stability OBI preserves the names, types, units, and meaning of telemetry marked `stable` in its registry. These guarantees cover both OTLP and built-in Prometheus output. A stable metric can have attributes that are still in development; those attributes may change. Telemetry marked development, alpha, beta, or release candidate can also change. Network flow attribute names may change while their semantic conventions are being aligned. The metrics and attributes you receive depend on enabled features, protocol support, attribute selection, and available runtime metadata. See the [telemetry compatibility contract](https://github.com/open-telemetry/opentelemetry-ebpf-instrumentation/blob/v1.0.0/TELEMETRY.md) for the compatibility guarantees and exclusions. OBI attaches the following schema URL to its telemetry: `https://open-telemetry.github.io/opentelemetry-ebpf-instrumentation/schemas/obi/1.0.0`. Schema transformations cannot express every conditional rename, removal, or Prometheus-specific change. ## Exported metrics | Family | Name (OTel) | Name (Prometheus) | Type | Unit | Description | | ------------ | ------------------------------------- | --------------------------------------------- | ------------- | ----------- | --------------------------------------------------------------------------------------------------------------------- | | Application | `http.client.request.duration` | `http_client_request_duration_seconds` | Histogram | seconds | Duration of HTTP service calls from the client side | | Application | `http.client.request.body.size` | `http_client_request_body_size_bytes` | Histogram | bytes | Size of the HTTP request body as sent by the client | | Application | `http.client.response.body.size` | `http_client_response_body_size_bytes` | Histogram | bytes | Size of the HTTP response body as sent by the client | | Application | `http.server.request.duration` | `http_server_request_duration_seconds` | Histogram | seconds | Duration of HTTP service calls from the server side | | Application | `http.server.request.body.size` | `http_server_request_body_size_bytes` | Histogram | bytes | Size of the HTTP request body as received at the server side | | Application | `http.server.response.body.size` | `http_server_response_body_size_bytes` | Histogram | bytes | Size of the HTTP response body as received at the server side | | Application | `rpc.client.call.duration` | `rpc_client_call_duration_seconds` | Histogram | seconds | Duration of RPC service calls from the client side | | Application | `rpc.server.call.duration` | `rpc_server_call_duration_seconds` | Histogram | seconds | Duration of RPC service calls from the server side | | Application | `db.client.operation.duration` | `db_client_operation_duration_seconds` | Histogram | seconds | Duration of database client operations | | Application | `db.server.operation.duration` | `db_server_operation_duration_seconds` | Histogram | seconds | Duration of server-side Redis, Memcached, and SQL operations (Experimental) | | Application | `messaging.client.operation.duration` | `messaging_client_operation_duration_seconds` | Histogram | seconds | Duration of messaging client operations across supported systems such as Kafka, MQTT, NATS, and AMQP (Experimental) | | Application | `messaging.process.duration` | `messaging_process_duration_seconds` | Histogram | seconds | Duration of messaging process operations across supported systems such as Kafka, MQTT, NATS, and AMQP (Experimental) | | Application | `gen_ai.client.operation.duration` | `gen_ai_client_operation_duration_seconds` | Histogram | seconds | Duration of GenAI client operations (Experimental) | | Application | `gen_ai.client.token.usage` | `gen_ai_client_token_usage` | Histogram | 1 | Number of GenAI input/output tokens consumed, labeled by token type (Experimental) | | Go runtime | `go.memory.limit` | `go_memory_limit_bytes` | Gauge | bytes | Runtime memory limit configured for an instrumented Go service | | Go runtime | `go.memory.gc.goal` | `go_memory_gc_goal_bytes` | Gauge | bytes | Current garbage-collection heap goal | | Go runtime | `go.memory.gc.cycles` | `go_memory_gc_cycles_total` | Counter | cycles | Completed Go garbage-collection cycles | | Go runtime | `go.memory.gc.pause.duration` | `go_memory_gc_pause_duration_seconds` | Histogram | seconds | Cumulative stop-the-world garbage-collection pause durations | | Go runtime | `go.memory.used` | `go_memory_used_bytes` | Gauge | bytes | Memory used by stack and other runtime memory categories | | Go runtime | `go.memory.allocated` | `go_memory_allocated_bytes_total` | Counter | bytes | Cumulative allocated heap bytes | | Go runtime | `go.memory.allocations` | `go_memory_allocations_total` | Counter | allocations | Cumulative heap allocation count | | Go runtime | `go.cpu.time` | `go_cpu_time_seconds_total` | Counter | seconds | Cumulative runtime CPU time by state | | Go runtime | `go.goroutine.count` | `go_goroutine_count` | Gauge | goroutines | Current goroutine count | | Go runtime | `go.processor.limit` | `go_processor_limit` | Gauge | threads | Current `GOMAXPROCS` value | | Go runtime | `go.config.gogc` | `go_config_gogc_percent` | Gauge | percent | Current `GOGC` heap target percentage | | Go runtime | `go.schedule.duration` | `go_schedule_duration_seconds` | Histogram | seconds | Cumulative runnable-to-running goroutine latency | | JVM runtime | `jvm.memory.used` | `jvm_memory_used_bytes` | Gauge | bytes | Current JVM memory used, labeled by memory type and pool | | JVM runtime | `jvm.memory.committed` | `jvm_memory_committed_bytes` | Gauge | bytes | Current JVM memory committed, labeled by memory type and pool | | JVM runtime | `jvm.memory.limit` | `jvm_memory_limit_bytes` | Gauge | bytes | Current JVM memory limit, labeled by memory type and pool | | JVM runtime | `jvm.memory.used_after_last_gc` | `jvm_memory_used_after_last_gc_bytes` | Gauge | bytes | JVM memory used after the last garbage collection | | .NET runtime | `dotnet.process.cpu.count` | `dotnet_process_cpu_count` | UpDownCounter | CPUs | Runtime processor count, including processor-count overrides, container quota, and affinity | | .NET runtime | `dotnet.process.cpu.time` | `dotnet_process_cpu_time_seconds_total` | Counter | seconds | Process CPU time, labeled by `cpu.mode` as `user` or `system` | | .NET runtime | `dotnet.gc.last_collection.heap.size` | `dotnet_gc_last_collection_heap_size_bytes` | UpDownCounter | bytes | Heap size after the latest collection, including fragmentation, labeled by heap generation | | Node runtime | `nodejs.eventloop.time` | `nodejs_eventloop_time_seconds_total` | Counter | seconds | Cumulative event-loop time, labeled as `idle` or `active` | | Node runtime | `nodejs.eventloop.utilization` | `nodejs_eventloop_utilization_ratio` | Gauge | 1 | Active share of the event loop during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.min` | `nodejs_eventloop_delay_min_seconds` | Gauge | seconds | Minimum event-loop delay during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.max` | `nodejs_eventloop_delay_max_seconds` | Gauge | seconds | Maximum event-loop delay during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.mean` | `nodejs_eventloop_delay_mean_seconds` | Gauge | seconds | Mean event-loop delay during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.stddev` | `nodejs_eventloop_delay_stddev_seconds` | Gauge | seconds | Standard deviation of event-loop delay during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.p50` | `nodejs_eventloop_delay_p50_seconds` | Gauge | seconds | 50th percentile event-loop delay during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.p90` | `nodejs_eventloop_delay_p90_seconds` | Gauge | seconds | 90th percentile event-loop delay during the last sampling interval | | Node runtime | `nodejs.eventloop.delay.p99` | `nodejs_eventloop_delay_p99_seconds` | Gauge | seconds | 99th percentile event-loop delay during the last sampling interval | | Network | `obi.network.flow.bytes` | `obi_network_flow_bytes_total` | Counter | bytes | Bytes submitted from a source network endpoint to a destination network endpoint | | Network | `obi.network.flow.packets` | `obi_network_flow_packets_total` | Counter | packets | Packets observed from a source network endpoint to a destination network endpoint | | Network | `obi.network.inter.zone.bytes` | `obi_network_inter_zone_bytes_total` | Counter | bytes | Bytes flowing between cloud availability zones in your cluster (Experimental, currently only available in Kubernetes) | | Network | `obi.stat.tcp.rtt` | `obi_stat_tcp_rtt_seconds` | Histogram | seconds | TCP round-trip time (RTT) latency observed between network endpoints (StatsO11y) | | Network | `obi.stat.tcp.failed.connections` | `obi_stat_tcp_failed_connections_total` | Counter | 1 | Failed TCP connection attempts between endpoints, labeled by failure reason (StatsO11y) | | Network | `obi.stat.tcp.retransmits` | `obi_stat_tcp_retransmits_total` | Counter | 1 | TCP retransmissions observed per connection (StatsO11y) | | Network | `obi.stat.tcp.io` | `obi_stat_tcp_io_bytes_total` | Counter | bytes | Bytes transferred at the socket layer per TCP connection and I/O direction (StatsO11y) | > [!NOTE] > > - v0.10.0 adopts the OpenTelemetry semantic convention names > `rpc.client.call.duration` and `rpc.server.call.duration`. The corresponding > Prometheus metric names include `_call_` as shown in the table. > - Starting with v0.12.1, server-side Redis and Memcached measurements use > `db.server.operation.duration` instead of `db.client.operation.duration`. > Update dashboards and alerts that query those server-side measurements. The .NET processor-count and heap-size instruments use UpDownCounters in OTLP and gauges in Prometheus. OBI can also export [Span metrics](https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/connector/spanmetricsconnector) and [Service graph metrics](https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/connector/servicegraphconnector), which you can enable via the [features](../configure/options/) configuration option. Starting with v0.14.0, the `application` feature includes the HTTP body-size histograms. Use `application_red` instead if you only want RED metrics, or combine `application_red` and `application_sizes` in the same feature list. See [metrics export features](../configure/export-data/#metrics-export-features). HTTP duration and body-size metrics now include `error.type` on failed requests by default. Their Prometheus series have an `error_type` label, with an empty value for successful requests. Update dashboards and alerts that match the old series. OTLP span metrics no longer put `host.id` on each data point; it remains on the resource and in `target_info`. OBI's Prometheus exporter is unaffected. Starting with v0.15.2, database duration metrics include `db.collection.name` by default. RPC client duration includes `server.port`, and failed RPC server duration includes `error.type`. Messaging duration metrics can also split by operation type and subscription name. MQTT SUBSCRIBE uses a subscription name instead of a destination name. Review these dimensions when estimating [cardinality](../cardinality/) or updating existing queries. The new ring-buffer counters in [Internal metrics](#internal-metrics) are also available through OTLP as `obi.bpf.ringbuf.writes` and `obi.bpf.ringbuf.write.failures`. ## Service graph limitations When combining Prometheus service graph metrics from multiple OBI instances, keep the scrape target's `instance` label or an equivalent external label. The metrics do not include a label that uniquely identifies the OBI instance. Destinations can appear as `outgoing` when Pod metadata is missing from the Kubernetes informer cache, or when non-HTTP destinations lack Kubernetes metadata. When OBI resolves a destination to a Kubernetes Service, it cannot match that Service to a Pod's overridden OpenTelemetry service name or namespace. The graph can then show separate nodes for the Kubernetes and OpenTelemetry names of the same service. ## Attributes of OBI metrics For the sake of brevity, the metrics and attributes in this list use the OTel `dot.notation`. When using the Prometheus exporter, the metrics use `underscore_notation`. In order to configure which attributes to show or which attributes to hide, check the `attributes`->`select` section in the [configuration documentation](../configure/options/). | Metrics | Name | Default | | ------------------------------------- | ----------------------------------------------------------- | ------------------------------------------------- | | Application (all) | `http.request.method` | shown | | Application (all) | `http.response.status_code` | shown | | Application (all) | `http.route` | shown if `routes` configuration section exists | | Application (all) | `k8s.daemonset.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.deployment.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.namespace.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.node.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.owner.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.pod.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.container.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.pod.start_time` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.pod.uid` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.replicaset.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.statefulset.name` | shown if Kubernetes metadata is enabled | | Application (all) | `k8s.cluster.name` | shown if Kubernetes metadata is enabled | | Application (all) | `container.id` | shown if Docker metadata is enabled | | Application (all) | `container.name` | shown if Docker metadata is enabled | | Application (all) | `cloud.provider` | shown if cloud metadata is enabled | | Application (all) | `cloud.platform` | shown if cloud metadata is enabled | | Application (all) | `cloud.region` | shown if cloud metadata is enabled | | Application (all) | `cloud.account.id` | shown if cloud metadata is enabled | | Application (all) | `cloud.availability_zone` | shown if cloud metadata is enabled | | Application (all) | `cloud.resource_id` | shown if cloud metadata is enabled (Azure only) | | Application (all) | `host.id` | shown if cloud metadata is enabled | | Application (all) | `host.type` | shown if cloud metadata is enabled | | Application (all) | `host.image.id` | shown if cloud metadata is enabled (AWS only) | | Application (all) | `gcp.gce.instance.name` | shown if cloud metadata is enabled (GCP only) | | Application (all) | `gcp.gce.instance.hostname` | shown if cloud metadata is enabled (GCP only) | | Application (all) | `service.name` | shown | | Application (all) | `service.namespace` | shown | | Application (all) | `target.instance` | shown | | Application (all) | `url.path` | hidden | | Application (client) | `server.address` | hidden | | Application (client) | `server.port` | hidden | | Application `rpc.*` | `rpc.response.status_code` | shown | | Application `rpc.*` | `rpc.method` | shown | | Application `rpc.*` | `rpc.system.name` | shown | | `rpc.client.call.duration` | `server.port` | shown | | `rpc.server.call.duration` | `error.type` | shown on failed requests | | Application (server) | `client.address` | hidden | | `obi.network.flow.bytes` | `obi.ip` | hidden | | `db.client.operation.duration` | `db.operation.name` | shown | | `db.client.operation.duration` | `db.collection.name` | shown | | `db.server.operation.duration` | `db.operation.name` | shown | | `db.server.operation.duration` | `db.collection.name` | shown | | `messaging.client.operation.duration` | `messaging.system` | shown | | `messaging.client.operation.duration` | `messaging.destination.name` | shown | | `messaging.client.operation.duration` | `messaging.operation.type` | shown | | `messaging.client.operation.duration` | `messaging.destination.subscription.name` | shown when present | | `messaging.process.duration` | `messaging.system` | shown | | `messaging.process.duration` | `messaging.destination.name` | shown | | `messaging.process.duration` | `messaging.destination.subscription.name` | shown when present | | `obi.network.flow.bytes` | `client.port` | hidden | | `obi.network.flow.bytes` | `direction` | shown | | `obi.network.flow.bytes` | `dst.address` | hidden | | `obi.network.flow.bytes` | `dst.cidr` | shown if the `cidrs` configuration section exists | | `obi.network.flow.bytes` | `dst.name` | hidden | | `obi.network.flow.bytes` | `dst.port` | hidden | | `obi.network.flow.bytes` | `dst.zone` (only Kubernetes) | hidden | | `obi.network.flow.bytes` | `iface` | hidden | | `obi.network.flow.bytes` | `k8s.cluster.name` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.dst.name` | hidden | | `obi.network.flow.bytes` | `k8s.dst.namespace` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.dst.node.ip` | hidden | | `obi.network.flow.bytes` | `k8s.dst.node.name` | hidden | | `obi.network.flow.bytes` | `k8s.dst.owner.type` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.dst.type` | hidden | | `obi.network.flow.bytes` | `k8s.dst.owner.name` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.src.name` | hidden | | `obi.network.flow.bytes` | `k8s.src.namespace` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.src.node.ip` | hidden | | `obi.network.flow.bytes` | `k8s.src.owner.name` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.src.owner.type` | shown if Kubernetes is enabled | | `obi.network.flow.bytes` | `k8s.src.type` | hidden | | `obi.network.flow.bytes` | `server.port` | hidden | | `obi.network.flow.bytes` | `src.address` | hidden | | `obi.network.flow.bytes` | `src.cidr` | shown if the `cidrs` configuration section exists | | `obi.network.flow.bytes` | `src.name` | hidden | | `obi.network.flow.bytes` | `src.port` | hidden | | `obi.network.flow.bytes` | `src.zone` (only Kubernetes) | hidden | | `obi.network.flow.bytes` | `transport` | hidden | | `obi.network.flow.bytes` | `network.type` | hidden | | `obi.network.flow.bytes` | `network.protocol.name` | hidden | | `obi.network.flow.bytes` | `src.country` | shown if the `geoip` configuration section exists | | `obi.network.flow.bytes` | `src.asn` | shown if the `geoip` configuration section exists | | `obi.network.flow.bytes` | `dst.country` | shown if the `geoip` configuration section exists | | `obi.network.flow.bytes` | `dst.asn` | shown if the `geoip` configuration section exists | | `obi.stat.tcp.rtt` | `obi.ip` | hidden | | `obi.stat.tcp.rtt` | `src.address` | hidden | | `obi.stat.tcp.rtt` | `dst.address` | hidden | | `obi.stat.tcp.rtt` | `src.port` | hidden | | `obi.stat.tcp.rtt` | `dst.port` | hidden | | `obi.stat.tcp.rtt` | `src.name` | hidden | | `obi.stat.tcp.rtt` | `dst.name` | hidden | | `obi.stat.tcp.rtt` | `src.zone` | hidden | | `obi.stat.tcp.rtt` | `dst.zone` | hidden | | `obi.stat.tcp.rtt` | `network.tcp.handshake.role` | hidden | | `obi.stat.tcp.failed.connections` | `obi.ip` | hidden | | `obi.stat.tcp.failed.connections` | `src.address` | hidden | | `obi.stat.tcp.failed.connections` | `dst.address` | hidden | | `obi.stat.tcp.failed.connections` | `src.port` | hidden | | `obi.stat.tcp.failed.connections` | `dst.port` | hidden | | `obi.stat.tcp.failed.connections` | `src.name` | hidden | | `obi.stat.tcp.failed.connections` | `dst.name` | hidden | | `obi.stat.tcp.failed.connections` | `src.zone` | hidden | | `obi.stat.tcp.failed.connections` | `dst.zone` | hidden | | `obi.stat.tcp.failed.connections` | `reason` | hidden | | `obi.stat.tcp.failed.connections` | `network.tcp.handshake.role` | hidden | | `obi.stat.tcp.retransmits` | Source/destination address, port, name, and zone attributes | hidden | | `obi.stat.tcp.io` | Source/destination address, port, name, and zone attributes | hidden | | `obi.stat.tcp.io` | `network.io.direction` | shown | | Traces (HTTP) | `url.query` | shown when a query string exists | | Traces (GraphQL) | `graphql.document` | hidden | | Traces (SQL, Redis) | `db.query.text` | hidden | > [!NOTE] > > The `obi.network.inter.zone.bytes` metric supports the same set of attributes > as `obi.network.flow.bytes`, but all of them are hidden by default, except > `k8s.cluster.name`, `src.zone` and `dst.zone`. > > The `obi.network.flow.packets` metric supports the same attributes and > defaults as `obi.network.flow.bytes`. JVM memory metrics include > `jvm.memory.type` and `jvm.memory.pool.name`. The `nodejs.eventloop.time` > metric includes `nodejs.eventloop.state`, with a value of `idle` or `active`. ## Internal metrics > [!WARNING] > > OBI v0.12.1 removes Prometheus-style suffixes from four internal OTLP metric > names. Update OTLP queries from `obi.bpf.map.entries_total` to > `obi.bpf.map.entries`, from `obi.bpf.map.max_entries_total` to > `obi.bpf.map.max_entries`, from `obi.bpf.network.packets.total` to > `obi.bpf.network.packets`, and from `obi.bpf.network.ignored.packets.total` to > `obi.bpf.network.ignored.packets`. The corresponding Prometheus counter names > remain unchanged. The Prometheus gauge names become `obi_bpf_map_entries` and > `obi_bpf_map_max_entries`. OBI can be [configured to report internal metrics](../configure/internal-metrics-reporter/) in Prometheus Format. | Name | Type | Description | | --------------------------------------- | ---------- | ---------------------------------------------------------------------------------------- | | `obi_ebpf_tracer_flushes` | Histogram | Length of the groups of traces flushed from the eBPF tracer to the next pipeline stage | | `obi_otel_metric_exports_total` | Counter | Length of the metric batches submitted to the remote OTel collector | | `obi_otel_metric_export_errors_total` | CounterVec | Error count on each failed OTel metric export, by error type | | `obi_otel_trace_exports_total` | Counter | Length of the trace batches submitted to the remote OTel collector | | `obi_otel_trace_export_errors_total` | CounterVec | Error count on each failed OTel trace export, by error type | | `obi_prometheus_http_requests_total` | CounterVec | Number of requests towards the Prometheus Scrape endpoint, faceted by HTTP port and path | | `obi_bpf_network_ignored_packets_total` | Counter | Number of network packets dropped by OBI network filters before flow accounting | | `obi_bpf_ringbuf_writes_total` | CounterVec | Attempted writes to each ring buffer, labeled by `bpf_map_name` | | `obi_bpf_ringbuf_write_failures_total` | CounterVec | Writes that failed because the ring buffer was full, labeled by `bpf_map_name` | | `obi_instrumented_processes` | GaugeVec | Instrumented processes by OBI, with process name | | `obi_internal_build_info` | GaugeVec | Version information of the OBI binary, including the build time and commit hash | | `obi_avoided_services` | GaugeVec | Services for which OBI suppressed telemetry after detecting OpenTelemetry SDK export |