v1.45.0

🌟 Highlights 🌟

• RPC spans and call duration metrics are now stable, including gRPC and Apache Dubbo.
• RC promotions cover Kubernetes and container metrics, JVM file descriptor metrics, hostmetrics attributes, and RPC headers and trailers.
• Other highlights include Kubernetes resource monitoring, clearer network endpoint guidance, MongoDB and Oracle context propagation, JVM executor metrics, and progress on Weaver schema v2.

🛑 Breaking changes 🛑

• android: Remove the android entity in favor of the android attribute group. Usage on resources stays the same. (#4168)

• aws: Remove the aws.ecs entity in favor of aws.ecs.task, aws.ecs.cluster, and aws.ecs.container entities identified by their required ARNs. (#4170)

• aws: Remove the AWS log entity while retaining its attributes for resource metadata. (#4171)

• cloudfoundry: Split cloudfoundry.system into cloudfoundry.system and cloudfoundry.system.instance, and define required identifying attributes for Cloud Foundry entities for Weaver schema v2 migration. (#4191)

• db: Rename db.client.connection.max to db.client.connection.limit (#4073)

• faas: This renames the faas attributes describing the invocation to be faas.invocation.* (#4228)
  The faas.time attribute has been renamed to faas.invocation.start_time.
  The faas.invocation_id attribute has been renamed to faas.invocation.id to complement the time attribute.

• gcp: Remove the gcp.gce and gcp.cloud_run entities in favor of gcp.gce.instance and gcp.cloud_run.job.task, and define required identifying attributes for GCP entities. (#4173)

• rpc: Replace rpc.request.metadata.<key> and rpc.response.metadata.<key> with rpc.request.header.<key>, rpc.response.header.<key>, and rpc.response.trailer.<key>. (#3924, #4072)
  gRPC and Connect RPC send response metadata both before and after the response payload and
  expose the two separately on their APIs. rpc.response.metadata.<key> did not record which
  of the two a value came from, so this change splits it into rpc.response.header.<key> for
  metadata sent before the response payload and rpc.response.trailer.<key> for metadata sent
  after it. Dubbo Triple has the same split: instrumentations record Triple response headers as
  rpc.response.header.<key> and Dubbo response attachments, which Triple sends as trailers, as
  rpc.response.trailer.<key>.

  No RPC system sends metadata after the request payload, so there is no rpc.request.trailer
  attribute and the request side is a plain rename.

  The already-deprecated rpc.grpc.* and rpc.connect_rpc.* metadata attributes now point at
  the new attributes directly instead of at the deprecated rpc.request.metadata.<key> and
  rpc.response.metadata.<key>.

• rpc: Rename rpc.response.status_code to rpc.status_code. (#3870)

• system: Rename system.device attribute to system.device.name (#1474)
  This namespaces system.device to system.device.name to clarify the property being described.
  The old system.device has been deprecated.

• system: system.memory.state now only defines the used and available states. (#2538)
  The new available state corresponds to MemAvailable on Linux, so that used + available
  equals system.memory.limit. The free, buffers and cached states are deprecated.

• zos: Remove z/OS entities and deprecate z/OS attributes in favor of open-telemetry/semantic-conventions-mainframe. (#4174, #3808)

🚩 Deprecations 🚩

• azure: Move the azure.resource.log event to schema v2 and deprecate section describing its body fields. (#4085)

• enduser: Deprecate enduser.id and enduser.pseudo.id in favor of open-telemetry/semantic-conventions-client-side. (#4238)

• messaging: Rename messaging.system to messaging.system.name and align cloud-provider values with their attribute namespaces. (#4250)

• messaging: Move system-specific messaging attributes into system-specific namespaces. (#4004)
  Rename messaging.kafka.*, messaging.rabbitmq.*, and messaging.rocketmq.*
  to kafka.*, rabbitmq.*, and rocketmq.*, respectively.
  Rename messaging.gcp_pubsub.* to gcp.pubsub.*,
  messaging.servicebus.* to azure.servicebus.*, and
  messaging.eventhubs.* to azure.eventhubs.*.
  The old attributes are deprecated.

• opentracing: Deprecate OpenTracing compatibility document and opentracing.ref_type attribute. (#4049)

• service: Deprecate service.peer.name and service.peer.namespace in favor of server.service.name/.namespace and client.service.name/.namespace. (#3472)

💡 Enhancements 💡

• aws: Define AWS Lambda, S3, and DynamoDB spans as refinements of the FaaS server and AWS SDK client spans. (#3808)
  The AWS Lambda span now lists the attributes it inherits from the FaaS server span, including
  the required faas.trigger.

• client: Add client.service.name and client.service.namespace attributes at development stability. (#3472)

• client: Add client.as.number and client.as.organization attributes for client network endpoints.
  (#1663, #3740)

• client: Add the client.mac.address attribute for client network endpoints.
  (#2416)

• client: Add the client.prefix attribute in CIDR notation for client network endpoints, with normalization by instrumentation or Collector receivers.
  (#3731)

• container: Define identifying attributes for container entities for Weaver schema v2 migration. (#4164)

• cpu: Clarify that cpu.logical_number is not the same as the number of cores. (#2934)
  cpu.logical_number attribute is not the same as the number of cores on a system, for example,
  a CPU with hyper-threading may have 8 cores and 16 logical CPUs. The cpu.logical_number
  attribute brief has been updated to include this information.

• db: Add trace context propagation for MongoDB through the comment command field (#3926)

• db: Add application context propagation to Oracle DB via piggybacking (#3863)
  Also recommends driver-provided APIs for updating V$SESSION.ACTION
  instead of issuing SQL or PL/SQL directly.

• db: Db spans have been switched to v2 syntax including spans now becoming refinements & adding db.system.name attribute. (#3808)
  azure.cosmosdb.client, db.cassandra.client, db.couchdb.client, db.elasticsearch.client, db.hbase.client,
  db.mariadb.client, db.mongodb.client, db.mysql.client, db.oracledb.client, db.postgresql.client, db.redis.client,
  db.sql_server.client, db.sql.client span defintiions have been switched to a refinement of the db.client span.

• db: Clarify that database client server.address identifies the configured database target, including multiple servers and service discovery.
  (#4041)

• db: Add db.system.name, db.namespace, server.address, and server.port attributes to database client connection pool metrics.
  (#4097)

• deployment: Define identifying attributes for deployment entities for Weaver schema v2 migration. (#4167)

• destination: Clarify that destination.address and destination.port describe the receiver of a network exchange and are recorded as the address available at the point of instrumentation. Instrumentation captures that address and does not try to resolve the address behind intermediaries such as proxies. They apply when there is no clear or known client/server role (for example low-level flow/packet telemetry or peer-to-peer protocols); use client.* / server.* for logical, de-proxied identity when the roles are known.
  (#3947, #1721)

• destination: Add destination.as.number and destination.as.organization attributes for destination network endpoints.
  (#1663, #3740)

• destination: Add the destination.mac.address attribute for destination network endpoints.
  (#2416)

• destination: Add the destination.prefix attribute in CIDR notation for destination network endpoints, with normalization by instrumentation or Collector receivers.
  (#3731)

• docs: Link the "Example Values" column directly to the full list of well-known values for enum-typed attributes. (#1085)

• docs: Stop rendering the redundant "well-known values" table on per-signal usage pages for enums whose only member is the generic fallback value, such as error.type. (#2141)
  The attribute registry page still shows this table.

• general: Support unbounded attributes with annotations.value.unbounded_size: true, add *_ref attribute conventions, and enforce both via Rego policies. (#1428, #3773, #4102)

• host: Define identifying attributes for host entities for Weaver schema v2 migration. (#4165)

• host: Define an explicit priority order for host.id sources, list the currently recognized cloud environments, and specify behavior around machine-id lookup and the user-configured fallback properties.
  (#3897)

• host: Define attribute roles for the host entity, make host.id identifying. (#4086)

• http: Add opt-in http.request.body.content and http.response.body.content attributes to capture HTTP request and response bodies (#857)
  These attributes may contain sensitive information.

• jvm: Promote jvm.file_descriptor.count and jvm.file_descriptor.limit metrics to release_candidate (#4019)
  The following metrics are being promoted to release_candidate:

  • jvm.file_descriptor.count
  • jvm.file_descriptor.limit
    These metrics are only available on Unix-like operating systems and are obtained
    from the UnixOperatingSystemMXBean JVM management interface.

• jvm: Define development semantic conventions for JVM executor metrics, including executor identities fixed for the lifetime of each registration, completed-task and rejected-submission counting semantics, and user-configured names for unnamed executors. (#3956)

• k8s: Add Pod level CPU limit and request metrics. (#4064)
  Introduces k8s.pod.cpu.limit.desired, k8s.pod.cpu.limit.current,
  k8s.pod.cpu.limit.utilization, k8s.pod.cpu.request.desired,
  k8s.pod.cpu.request.current and k8s.pod.cpu.request.utilization,
  aligning the k8s.pod namespace with the equivalent k8s.container metrics
  and with the K8s Pod level resources feature.

• k8s: Promote container.uptime, k8s.pod.uptime and k8s.node.uptime metrics to release_candidate (#3945)
  The following metrics are being promoted to release_candidate:

  • container.uptime
  • k8s.pod.uptime
  • k8s.node.uptime
    Notes referencing the Kubelet stats summary API fields (ContainerStats.StartTime,
    PodStats.StartTime, NodeStats.StartTime) have been added to each metric.

• k8s: Promote container.cpu.usage, k8s.pod.cpu.usage and k8s.node.cpu.usage metrics to release candidate (#4145)

• k8s: Add Pod and container memory limit and request utilization metrics. (#4133)
  Introduces k8s.pod.memory.limit.utilization, k8s.pod.memory.request.utilization,
  k8s.container.memory.limit.utilization and k8s.container.memory.request.utilization,
  aligning memory with the equivalent CPU utilization metrics. The utilization metrics use
  the memory working set as the numerator.

• k8s: Add Pod level memory limit and request metrics. (#4133)
  Introduces k8s.pod.memory.limit.desired, k8s.pod.memory.limit.current,
  k8s.pod.memory.request.desired and k8s.pod.memory.request.current,
  aligning Pod level memory with the equivalent Pod level CPU metrics.

• k8s: Promote k8s.pod.filesystem.*, k8s.node.filesystem.*, and container.filesystem.* metrics to release_candidate. Container filesystem metrics (container.filesystem.available, container.filesystem.capacity, container.filesystem.usage) also gain documentation notes mapping them to their cadvisor equivalents (container_fs_limit_bytes, container_fs_usage_bytes), enabling use cases where metrics are retrieved directly from cadvisor rather than through the Kubelet stats summary API.
  (#4053)

• k8s: Clarify the Linux working set calculation for Pod, Node, and container memory working set metrics and document the related kubelet eviction behavior. (#4133)

• k8s: Promote k8s.pod.network.io, k8s.pod.network.errors, k8s.node.network.io, and k8s.node.network.errors metrics to release_candidate (#3946)

• k8s: Promote paging fault metrics (k8s.node.paging.faults, k8s.pod.paging.faults, container.paging.faults) to RC (#4036)

• k8s: Add k8s.statefulset.pod.available metric, reporting the number of available replica pods (ready for at least minReadySeconds) targeted by a StatefulSet, aligning with the existing .pod.available metrics on k8s.deployment, k8s.replicaset, and k8s.replicationcontroller.
  (#4028)

• k8s: Add k8s.node.pid.usage and k8s.node.pid.limit metrics to track Kubernetes node PID limit and usage. (#3864)

• log: Remove docs/general/logs.md and attribute groups covered there (log.record, attributes.log, and attributes.log.file) in favor of attributes defined in the registry.
  (#4083)

• network: Remove the network connection and carrier attributes section from docs/general/attributes.md. The network.carrier.* and network.connection.* attributes remain available in the attribute registry.
  (#4048)

• network: Add guidance for choosing between client.* / server.*, source.* / destination.*, and network.local.* / network.peer.* in a dedicated Network document (docs/network/choosing-network-attributes.md).
  (#3947, #1721)

• network: Add public network.peer and network.local attribute groups, replacing the internal network-core group.
  (#4084)

• network: Add network.local.as.number, network.local.as.organization, network.peer.as.number, and network.peer.as.organization attributes for local and peer network endpoints.
  (#1663, #3740)

• network: Add network.local.mac.address and network.peer.mac.address attributes for local and peer network endpoints.
  (#2416)

• network: Add network.local.prefix and network.peer.prefix attributes in CIDR notation for local and peer network endpoints, with normalization by instrumentation or Collector receivers.
  (#3731)

• os: Define identifying attributes for os entities for Weaver schema v2 migration. (#4165)

• otel: Mark error.type as conditionally required on otel.sdk.processor.{span,log}.processed and otel.sdk.exporter.{span,log,metric_data_point}.exported, matching otel.sdk.exporter.operation.duration.
  (#3954)

• otel: Define otel.scope.name as required and identifying and keep otel.scope.version recommended and descriptive for Weaver schema v2 migration. (#4193)

• otel: Clarify that otel.sdk.log.created counts log records at intake on an enabled Logger, before per-record filtering.
  (#3506)

• profile: Extend the list of known frame types with a value for CUDA (#4079)

• rpc: Update RPC call-duration histogram bucket recommendations and align gRPC boundaries with gRPC's existing OpenTelemetry metrics. (#316)
  General RPC (including Dubbo) uses 18 boundaries from 0.0001 to 60 seconds with 1, 2.5, 5 subdivisions through 10 seconds;
  gRPC uses the 41 latency boundaries defined in gRPC A66, including zero.

• rpc: Mark rpc.request.header.<key>, rpc.response.header.<key>, and rpc.response.trailer.<key> as release candidate. (#4276)

• rpc: Clarify that RPC instrumentations may use additional context or provide configuration options to customize which status or error codes are considered errors.
  (#3704)

• rpc: Clarify how to populate server.address and server.port on RPC server spans. Change requirement level of server.address to opt_in on rpc.server.call.duration metric and to recommended on server spans.
  (#3784)

• rpc: Add Apache Dubbo-specific metric refinements for rpc.server.call.duration and rpc.client.call.duration, and define Dubbo spans as refinements of the generic RPC spans with shared attribute definitions. The rpc.method requirement level differs between client (required) and server (conditionally required), matching the corresponding Dubbo span conventions.
  (#4120)
  Require client server.port only when server.address is set and a port is
  specified in the direct provider target URL, excluding registry discovery targets.
  Dubbo spans now explicitly include rpc.system.name.

• rpc: Add gRPC-specific metric refinements for rpc.server.call.duration and rpc.client.call.duration, rendered on the gRPC page. The rpc.method requirement level differs between client (required) and server (conditionally required), matching the corresponding gRPC span conventions. Convert gRPC spans to refinements and share gRPC attribute definitions across spans and metrics.
  (#3868)

• rpc: Mark RPC spans and call duration metrics as stable, including the gRPC and Apache Dubbo conventions. (#3477)

• rpc: Define Connect RPC and JSON-RPC spans as refinements of the generic RPC client and server spans. (#3808)
  JSON-RPC spans now list the opt-in rpc.request.header.<key>, rpc.response.header.<key>, and
  rpc.response.trailer.<key> attributes inherited from the generic RPC spans.

• server: Add server.service.name and server.service.namespace attributes at development stability. (#3472)

• server: Add server.as.number and server.as.organization attributes for server network endpoints.
  (#1663, #3740)

• server: Add the server.mac.address attribute for server network endpoints.
  (#2416)

• server: Add the server.prefix attribute in CIDR notation for server network endpoints, with normalization by instrumentation or Collector receivers.
  (#3731)

• service: Add service.instance.cost_center.id and service.instance.cost_center.name attributes to the service registry. (#3754)

• source: Clarify that source.address and source.port describe the sender of a network exchange and are recorded as the address available at the point of instrumentation. Instrumentation captures that address and does not try to resolve the address behind intermediaries such as proxies. They apply when there is no clear or known client/server role (for example low-level flow/packet telemetry or peer-to-peer protocols); use client.* / server.* for logical, de-proxied identity when the roles are known.
  (#3947, #1721)

• source: Add source.as.number and source.as.organization attributes for source network endpoints.
  (#1663, #3740)

• source: Add the source.mac.address attribute for source network endpoints.
  (#2416)

• source: Add the source.prefix attribute in CIDR notation for source network endpoints, with normalization by instrumentation or Collector receivers.
  (#3731)

• system: Promote attributes used by hostmetrics system metrics to RC (#1804)

• v8js: Add members for the current V8 heap space set to v8js.heap.space.name and note that map_space is only reported by Node.js <= 19, since V8 merged the map space into the old space. (#4057)

• webengine: Define webengine.name as identifying and keep webengine.version recommended and descriptive for Weaver schema v2 migration. (#4194)

🧰 Bug fixes 🧰

• db: Correct PostgreSQL namespace examples and handle pipe escaping in Markdown templates. (#4222)

• docs: Fix broken Markdown tables when an attribute, metric, or enum-member brief spans multiple lines in YAML. (#3073)

• k8s: Fix description of k8s.container.cpu.request.utilization to allow values above 1.0. (#4126)
  The previous description incorrectly stated the value range is [0.0,1.0].
  Since a CPU request is a scheduling reservation and not a hard cap, values
  above 1.0 are possible and indicate the container is using more CPU than requested.

• url: Restore AWSAccessKeyId and Signature to the default list of sensitive query parameter keys. (#3989)

New Contributors

• @om7057 made their first contribution in #4023
• @svencowart made their first contribution in #4046
• @Aman-goel-04 made their first contribution in #4108
• @pinoOgni made their first contribution in #4057
• @peschinskiy made their first contribution in #3916
• @riccardomenegazzo made their first contribution in #4148
• @Abhimanyu9988 made their first contribution in #4158
• @martinkuba made their first contribution in #4238
• @udsmicrosoft made their first contribution in #4056
• @YaoYingLong made their first contribution in #3970
• @ajimenez1503 made their first contribution in #4115

Full Changelog: v1.44.0...v1.45.0