v1.44.0

🛑 Breaking changes 🛑

• browser: Reworked the browser.web_vital event - moved name/value/delta/id from the event body to browser.web_vital.* attributes, and added new attributes per current instrumentation. (#3401)
• k8s: Rename k8s.node.memory.paging.faults, k8s.pod.memory.paging.faults and container.memory.paging.faults metrics by dropping the memory namespace segment.
  (#3805)
  Renamed metrics:
  • k8s.pod.memory.paging.faults -> k8s.pod.paging.faults
  • k8s.node.memory.paging.faults -> k8s.node.paging.faults
  • container.memory.paging.faults -> container.paging.faults
• k8s: Change {container, k8s.pod, k8s.node}.memory.usage to updowncounter (#3889)

💡 Enhancements 💡

• cloud: Add Scaleway Cloud to cloud.provider and scaleway_cloud_compute to cloud.platform (#2773)

• db: Clarify db.query.parameter.<key> semantics around sensitive data handling.
  (#3831)

• faas: The general FaaS attributes are now defined on the FaaS spans that use them, so each span documents its full attribute set. (#3905)

• host: Update the recommendations to use full path to commands when collecting host.id attribute value.
  (#3896)

• hw: Per-component hw.errors and hw.status metrics now document the full set of component-specific attributes (e.g. hw.model, hw.vendor) alongside the generic ones.
  (#3808)

• k8s: Add k8s.node.filesystem.inode.count and k8s.node.filesystem.inode.free metrics to track Kubernetes node root filesystem inode capacity and usage. (#3858)

• k8s: Promote a selection of container and K8s memory metrics to release_candidate (#3882)
  The following metrics are being promoted to release_candidate:

  • container.memory.usage
  • container.memory.available
  • container.memory.rss
  • container.memory.working_set
  • k8s.pod.memory.usage
  • k8s.pod.memory.available
  • k8s.pod.memory.rss
  • k8s.pod.memory.working_set
  • k8s.node.memory.usage
  • k8s.node.memory.available
  • k8s.node.memory.rss
  • k8s.node.memory.working_set

• k8s: Clarify how cpu.usage is calculated for K8s and container entities. (#2418)
  CPU usage is measured in cores/CPUs and is calculated based on the formula:
  usageCores = (cpuTimeEnd - cpuTimeStart) / elapsedSeconds
  This PR improves the description making it aligned with the corresponding
  system and process metrics.

• messaging: Add messaging.kafka.cluster.id attribute identifying the Kafka cluster a client is connected to, and reference it from Kafka spans. (#32)

• messaging: Define a span per messaging operation type (create, send, receive, process, settle) and refine them for each messaging system. (#3904)

• network: Promote network.interface.name attribute to release_candidate (#3953)

• otel: Clarify that otel.sdk.processor.span.processed and otel.sdk.processor.log.processed are recorded when the processor passes items to the exporter, not when an item is accepted into a queue or when the export operation completes, and that the export outcome does not affect these metrics.
  (#3902)

• process: Add process.disk.operations, process.memory.utilization, and process.signals_pending metrics (#3937)
  Process namespace has been made release candidate. During migration of process metrics to release
  candidate in hostmetrics receiver, it was discovered that the following process metrics were not
  present in process semantic conventions: process.disk.operations, process.memory.utilization,
  and process.signals_pending.

• server, client, source, destination: The descriptions of the server, client, source, and destination attribute groups are now documented in General attributes instead of the attribute registry pages.
  (#3935)

• system: Add system.process.limit metric representing the maximum number of concurrent tasks/threads allowed by the operating system. (#3859)

🧰 Bug fixes 🧰

• otel: Clarify that already_shutdown on otel.sdk.processor.span.processed and otel.sdk.processor.log.processed is a value-only requirement for the error.type attribute: whether and when a processor drops records because it has already been shut down is governed by the SDK specification, not by this metric.
  (#3957)
• rpc: Fix wording in server.address note to say "reverse DNS lookup" instead of "reverse proxy lookup".
  (#3871)
• rpc: Remove tracing-only rpc.method_original guidance from the shared rpc.method definition.
  (#3872)

New Contributors

• @Jaegeonyoon made their first contribution in #3843
• @shashank-reddy-nr made their first contribution in #3819
• @lachmatt made their first contribution in #3896
• @jaredDlewis made their first contribution in #3851

Full Changelog: v1.43.0...v1.44.0

v1.43.0

🚩 Deprecations 🚩

• docs: Add a generic signal requirement level document and deprecate the metric-specific one.
  (#3810, #3278)

💡 Enhancements 💡

• azure: Add azure.resource_group.name attribute for the name of the Azure resource group the resource belongs to and document that cloud.account.id maps to the Azure subscription ID. (#3697, #3708)

• cicd: Promote CI/CD span, metric, log, attribute, and entity conventions to release_candidate (#3674, #3675, #3677)

• db: Clarify db.operation.batch.size semantics for database batch operations, including multi-operand operations, parameterized operation batch APIs, and empty batches.
  (#3811)

• db: Promote Oracle Database connection attributes to release candidate so the RC db.oracledb.client span does not reference lower-stability attributes. (#3753)

• disk, network, system: Promote disk.io.direction, network.io.direction, and system.paging.fault.type attributes to RC (#3796)
  The network.io.direction brief was updated based on review feedback to clarify that
  it describes traffic at the observing interface boundary and is not intended to represent
  logical connection or network flow direction.

• otel: Clarify that the otel.span.sampling_result attribute of otel.sdk.span.live can only take values RECORD_ONLY and RECORD_AND_SAMPLE, not DROP, since the metric does not count non-recording spans.
  (#3782)

• process: Promote process metrics and entity to release_candidate (#3041)

• rpc: Clarify how to populate server.address and server.port for Dubbo consumers using service discovery. (#3408)

• schema: Add registry manifest. (#3824)

• telemetry: Add kotlin as a value for the telemetry.sdk.language attribute. (#3552)

• vcs: Promote VCS attribute, entity, and metric conventions to release_candidate (#3676)

🧰 Bug fixes 🧰

• db: Align Oracle Database db.query.text guidance with other SQL database conventions. (#3790)
  Parameterized query text is now recommended to be collected by default, while
  query parameter values remain opt-in.

New Contributors

• @kathiehuang made their first contribution in #3722

Full Changelog: v1.42.0...v1.43.0

v1.42.0

📣 Highlights 📣

• Generative AI Migration: All gen_ai.* attributes, metrics, events, and spans have been deprecated and moved to the dedicated OpenTelemetry GenAI Semantic Conventions Repository. (#3696)
• Graduations to Stable: A core selection of Kubernetes and container registry resource attributes have officially graduated to stable status. (#3382)
• Promotions to Release Candidate: Several conventions advanced to release_candidate, including cpu.mode (#3694), process attributes (#3041), and core CPU time metrics (k8s.pod.cpu.time, k8s.node.cpu.time, and container.cpu.time) (#3001).

Changelog

🛑 Breaking changes 🛑

• gen-ai: Move Generative AI semantic conventions to a dedicated repository. (#3696)
  All gen_ai.* attributes, metrics, events, and spans previously defined under
  model/gen-ai/, model/openai/, and model/mcp/ (and documented under
  docs/gen-ai/) are deprecated in this repository and have moved to the
  OpenTelemetry GenAI semantic conventions repository.

  Instrumentations following the new repository's conventions should refer to
  it for the corresponding schema_url to use.

• http: Change network.peer.address attribute requirement level from Recommended to Opt-In on the http.client.open_connections and http.client.connection.duration metrics. (#3279)
  network.peer.address is typically high-cardinality (one time series per remote IP)
  and causes unbounded growth of metric streams over the lifetime of the process when
  exported with cumulative temporality (always the case for http.client.open_connections
  since it is an UpDownCounter, and possible for http.client.connection.duration when
  the histogram is configured for cumulative). Operators that need this attribute can
  still opt-in.

• v8js: Rename v8js.memory.heap.limit to v8js.memory.heap.space.size (per-space pre-allocated size from v8.getHeapSpaceStatistics()),
  and add a new v8js.memory.heap.limit UpDownCounter representing the absolute heap size limit from v8.getHeapStatistics().heap_size_limit
  (controlled by --max-old-space-size or V8 defaults).
  (#3476)

🚀 New components 🚀

• browser: Introduce browser.document.url.full attribute and browser.document entity to capture the absolute URL of the current browser document. (#174)
  The browser.document entity is introduced as a separate entity from browser because
  the browser runtime attributes (brands, platform, language) are immutable for the SDK
  lifetime, while the document URL changes on every navigation. The new attribute follows
  RFC3986 and has Recommended requirement level at Development stability.

💡 Enhancements 💡

• app: Defines a basic crash of an end-user facing app that requires a minimal amount of information. (#3448)

• cpu: Promote cpu.mode attribute and its enum members to release_candidate (#3694)

• docs: Add guidance for defining events in semantic conventions. (#1855)

• docs: Don't allow signal definition to reference attributes with lower stability than the signal itself (unless they are opt-in).
  (#3752)

• gcp: Add GCE instance labels as resource attributes. (#2617)

• k8s: Promote k8s.pod.cpu.time, k8s.node.cpu.time and container.cpu.time metrics to release_candidate (#3001)

• k8s: Add metric k8s.container.ephemeral_storage.usage to track ephemeral storage usage by containers. (#3681)
  This metric includes an attribute k8s.container.ephemeral_storage.fs_type to distinguish between rootfs and logs usage.

• k8s: Promote a selection of k8s and container registry attributes to stable (#3382)

• otel: Clarify semantics of SDK exporter metrics when the exporter performs retries: an export operation is considered finished only after the final attempt has concluded, items are counted exactly once per operation, and otel.sdk.exporter.operation.duration records one observation per operation covering all attempts and backoff. Also drop a spurious error.type note from the *.inflight metrics, which do not carry an error.type attribute.
  (#3770)

• otel: Reserve already_shutdown as a value for error.type on the otel.sdk.processor.span.processed and otel.sdk.processor.log.processed metrics, used when a processor drops items because it has already been shut down.
  (#3710)

• process: Promote process attributes to release_candidate (#3041)

• profile: Extend the list of known frame types with a value for LuaJIT (#3707)

• service: Mark service.criticality attribute as alpha. (#2986)
  The service.criticality attribute and its allowed values (critical, high, medium, low)
  are now in alpha status.

• system, file: Add support for filesystem lock counts (#3611)

• v8js: Update v8js.gc.duration metric to match bucket boundaries used by Node.js instrumentation. (#3761)

🧰 Bug fixes 🧰

• browser: remove the condition to include user_agent.original attribute in browser resource (#3738)
• cicd: Consistent spelling of "CI/CD" throughout the CICD registry. (#3634)
• container: Remove kernel from cpu.mode enum and clarify container cpu states (#3700)
• k8s: Exclude k8s.container.cpu.limit_utilization and k8s.container.cpu.request_utilization metrics from code generation (#3711, #3705)

New Contributors

• @mquentin made their first contribution in #3633
• @Hronom made their first contribution in #3481
• @Dipanshusinghh made their first contribution in #3739
• @mcollina made their first contribution in #3761
• @david-luna made their first contribution in #3738
• @Ayushi-12 made their first contribution in #3781

Full Changelog: v1.41.0...v1.42.0

v1.41.1

What's Changed

• Exclude k8s.container.cpu.limit_utilization and k8s.container.cpu.request_utilization metrics from code generation (#3711)

Full Changelog: v1.41.0...v1.41.1

v1.41.0

🛑 Breaking changes 🛑

• gen-ai: Execute tool call requires the name of the tool used for span naming. (#3595)

• graphql: Change graphql.document attribute requirement level from Recommended to Opt-In due to sensitive data, cardinality, and size concerns (#2985)
  The graphql.document attribute contains user-inputted data that is often sensitive,
  potentially unbounded in length, and high-cardinality. Moving from Recommended to Opt-In
  ensures operators make an informed decision before enabling this attribute.
  Also updated the note to use SHOULD for stronger normative language regarding sanitization.

• process: Move process.executable to its own entity. (#3535)

  • Move process.executable to its own entity
  • Identifying attribute:
    • process.executable.build_id.htlhash
      • We need to update its description in another PR.
  • Descriptive attributes:
    • process.executable.build_id.go
    • process.executable.build_id.gnu
    • process.executable.path
    • process.executable.name

• process: Update requirement levels for process attributes to ensure consistent identification and description across platforms. (#864)

  • Changed process.pid and process.creation.time to be Required (Identity).
  • Changed process descriptive attributes to be Optional.

• rpc: Remove client.address and client.port attributes from RPC server spans. (#3487, #3488)

💡 Enhancements 💡

• Go: Add opt-in go.memory.gc.pause.duration histogram metric. (#3353)

• cicd: Clarify that cicd.pipeline.task.run.id must be unique within a pipeline run. (#3566)
  This id may be used for correlating results of a task across several pipeline runs.

• configuration: Defined declarative configuration for semantic convention version selection. (#3424)

• configuration: Clarify that the experimental flag should not be used to gate incomplete implementations of stable convention versions. (#3505)

• db: Include db.system.name value changes in database migration guide. (#3557)

• db: Clarify db.response.returned_rows when captured on spans. (#3480)

• db: Clarify database client span duration. (#3479)

• deployment: Stabilize deployment.environment.name attribute. (#3339)
  The deployment.environment.name attribute is promoted from development to stable.

• deployment: Add enum values for deployment.environment.name attribute. (#2910)
  The deployment.environment.name attribute now has well-known enum values: Production, Staging, Test, and Development.

• exception: Improve exception event definition: add guidance on how to set event name and severity.
  (#3311)

• exceptions: Add guidance that error.type and exception.type may unwrap uninformative wrapper types. (#3588, #3623)

• faas: Define domain-specific exception event for FaaS invocations (#3451)
  This domain-specific exception event provides a structured way to record
  exceptions that occur during FaaS function invocations:

  • faas.invocation.exception for exceptions during function execution

• gen-ai: Enhance the definition of gen_ai.tool.definitions attribute. (#2721, #1835)
  The schema of gen_ai.tool.definitions attribute is now enhanced to:

  • Add JSON schema of gen_ai.tool.definitions attribute.
  • Document the behavior of capturing tool definitions.
  • Capture the tool definitions in a simplified format if the content capturing is disabled.

• gen-ai: Add gen_ai.usage.reasoning.output_tokens attribute for reasoning tokens (#3194)

  • Add gen_ai.usage.reasoning.output_tokens for tokens used for reasoning

• gen-ai: Define gen_ai.client.operation.exception event for recording exceptions during Generative AI client operations (#3436)
  This domain-specific exception event provides a structured way to record
  exceptions that occur during Generative AI client operations (API errors,
  rate limiting, timeouts).

• gen-ai: Add invoke workflow operation name support to the gen-ai semantic conventions. (#2912)

• gen-ai: Add gen_ai.client.operation.time_to_first_chunk and gen_ai.client.operation.time_per_output_chunk metrics. (#3113)

  • Add gen_ai.client.operation.time_to_first_chunk metric to track time from client request to first chunk of response.
  • Add gen_ai.client.operation.time_per_output_chunk metric to track time per output chunk of response.

• gen-ai: Split invoke_agent into separate client and internal spans, and split attributes from inference hierarchy. (#2632)
  The invoke_agent client span now extends a dedicated attributes.gen_ai.invoke_agent.client
  attribute group instead of attributes.gen_ai.inference.client. This removes
  gen_ai.response.id, gen_ai.response.model, and gen_ai.response.finish_reasons
  and adds gen_ai.agent.version.
  A new invoke_agent internal span is added for in-process agents (e.g., LangChain, CrewAI),
  which does not include server.address, server.port, or token usage attributes.

• gen-ai: Added gen_ai.response.model attribute to embeddings spans. (#3499)

• gen_ai: Add streaming attributes for GenAI inference spans (#3598)
  Adds gen_ai.request.stream (boolean) and gen_ai.response.time_to_first_chunk (double)
  attributes for streaming inference operations.

• go: Add the go.cpu.time opt-in metric, and add go.cpu.detailed_state and go.memory.detailed_type attributes to CPU and memory metrics respectively with wildcard values. (#3354)

• go: Add the opt-in go.memory.gc.cycles metric. (#3353)

• http: Clarify that intentional client-side cancellation of HTTP requests SHOULD NOT be treated as an error. (#3495)

• k8s: Add PersistentVolume and PersistentVolumeClaim entities with phase and storage metrics. (#3375)
  Entities:

  • k8s.persistentvolume
  • k8s.persistentvolumeclaim
    Metrics:
  • k8s.persistentvolume.status.phase, k8s.persistentvolume.storage.capacity
  • k8s.persistentvolumeclaim.status.phase, k8s.persistentvolumeclaim.storage.request,
    k8s.persistentvolumeclaim.storage.capacity

• k8s: Update CPU metrics for container CPU limit and request to handle resize. (#3558)
  Introduced new metrics for container CPU limit and request that account for
  KEP 1287 allowing for in-place updates of container resources.

• k8s: Promote a selection of k8s/container attributes to release_candidate (#3491, #3380)

• k8s: Split container memory limit/request into desired/current to handle in-place resize (#3558)
  Introduced new metrics for container memory limit and request that account for
  KEP 1287 allowing for in-place updates of container resources.

• k8s: Add k8s.node.system_container.* metrics to capture node-level system daemon metrics. (#3531)

• messaging: Define domain-specific exception events for messaging (#3435)
  These domain-specific exception events provide a structured way to record
  exceptions that occur during messaging operations:

  • messaging.create.exception for messaging create operations
  • messaging.send.exception for messaging send operations
  • messaging.receive.exception for messaging receive operations
  • messaging.settle.exception for messaging settle operations
  • `messaging.pro...

v1.40.0

🛑 Breaking changes 🛑

• db: Oracle's db.namespace updated to represent only the database’s
  unique identifier. Information previously bundled into db.namespace —including
  PDB name, instance name, and service name — has been moved to dedicated attributes:
  oracle.db.pdb, oracle.db.instance.name, and oracle.db.service.
  Introduced a new oracle.db.domain attribute to capture the database domain
  (DB_DOMAIN).
  (#2147)
• error, feature-flags: Deprecate error.message in favor of using domain-specific codes such as feature_flag.error.message.
  (#3307, #3308)
• rpc: Remove network.protocol.name, network.protocol.version, and network.transport attributes from RPC spans and metrics.
  (#3349, #3350)
• system: Rename system.memory.shared to system.memory.linux.shared (#3260)
  The system.memory.shared metric has been renamed to system.memory.linux.shared to align with other Linux-specific memory metrics (system.memory.linux.available and system.memory.linux.slab.usage). This metric reports shared memory used, primarily by tmpfs, and is only available on Linux systems.

🚩 Deprecations 🚩

• exceptions: Update exception recording guidelines to not use span events. (#3256)
• rpc: Deprecate experimental rpc.server.request.size, rpc.server.response.size, rpc.client.request.size, and rpc.client.response.size metrics due to ambiguous definitions and inconsistent implementation.
  (#3267)
• rpc: Deprecate the rpc.message event and its attributes (#3266, #3283)
  Per-message events within RPC streaming calls are of limited utility in their current form. They are recorded as span events (also being deprecated). There is no direct replacement at this time.

💡 Enhancements 💡

• cloud: Add oracle_cloud.realm vendor-specific attribute for Oracle Cloud Infrastructure (OCI) realm identifier. (#3121)
  An Oracle Cloud realm is a logical collection of cloud regions that are isolated from each other, used for compliance and data residency requirements.

• db: Mark the Oracle Database client span as Release Candidate. (#2989)

• db: Modify area to oracledb from oracle (#3316)

• db: Clarified db.query.summary behavior for batch operations. (#3287)

• db, rpc: Define domain-specific exception events for database and RPC (#3426)
  These domain-specific exception events provide a structured way to record
  exceptions that occur during operations in each area:

  • db.client.operation.exception for database client operations
  • rpc.client.call.exception and rpc.server.call.exception for RPC

• docs: Add how-to guide for defining spans (#3240)

• event: Allow event body to represent a display message. (#3343)

• exception: Indicate that exception.message attribute may contain sensitive information. (#2967, #3310)

• exceptions: Introduce OTEL_SEMCONV_EXCEPTION_SIGNAL_OPT_IN environment variable for transitioning exception recording from span events to logs. (#3363)

• gcp: Add GCP GCE Instance Group Manager attributes (#905)

• gen-ai: Add retrieval span support to the gen-ai semantic conventions. (#2907)

• gen-ai: Enhance how tool calls are captured to extend support for server side tools.
  (#2585)

• gen-ai: Add gen_ai.agent.version attribute to capture the version of an agent. (#3428)

• gen-ai: Add cache token attributes and provider-specific normalization guidance for GenAI usage metrics (#1959)

  • Add gen_ai.usage.cache_read.input_tokens attribute for tokens served from provider cache
  • Add gen_ai.usage.cache_creation.input_tokens attribute for tokens written to provider cache
  • Add provider-specific token handling notes to OpenAI span
  • Add Anthropic span with computation guidance for gen_ai.usage.input_tokens

• gen_ai: Added sampling-relevant flag to relevant GenAI span attributes to indicate their importance for sampling decisions. (#2994)

• http: Defined declarative configuration for overriding known HTTP methods. (#3394)

• http: Define declarative configuration for sensitive query parameters. (#3403)

• http: Define http.client.request.exception and http.server.request.exception events for recording exceptions during HTTP operations (#3414)
  These domain-specific exception events provide a structured way to record
  exceptions that occur during HTTP client requests (network failures, timeouts)
  and HTTP server request processing (application errors, internal failures).

• jvm: Add jvm.file_descriptor.limit as an in-development metric to track the maximum number of open file descriptors as reported by the JVM. (#3430)

• k8s: Promote a selection of k8s and container attributes to beta (#3120)

• k8s: Add k8s.service entity with endpoint and load balancer ingress metrics (#3294)
  New metrics:

  • k8s.service.endpoint.count: Number of endpoints by condition, address type, and zone
  • k8s.service.load_balancer.ingress.count: Number of load balancer ingress points
    New attributes:
  • k8s.service.endpoint.condition, k8s.service.endpoint.address_type, k8s.service.endpoint.zone
  • k8s.service.traffic_distribution, k8s.service.publish_not_ready_addresses
    New entity:
  • entity.k8s.service: k8s service entity with identifying and descriptive attributes

• openai: Add openai.api.type attribute to distinguish between different OpenAI API types for similar operations. (#3337)

• pprof: add pprof.scope.default_sample_type and pprof.scope.sample_type_order attributes to support lossless conversion of pprof profiles. (#3078)

• rpc: Mark core RPC (plus gRPC and Apache Dubbo) semantic conventions as release candidate. (#3446)

• rpc: Add Metrics sections to JSON-RPC, gRPC, and Connect RPC system-specific conventions (#3319, #3320)

• rpc: Change rpc.server.call.duration and rpc.client.call.duration metrics from recommended to required (#3284)

• rpc: Add RPC migration guide documenting changes from v1.37.0 to latest version. (#3204)

• rpc: Clarify that network.peer.address should indicate the last contacted address in case of retries.
  (#3392, #3427)

• rpc: Add semantic conventions for Apache Dubbo RPC system. (#3291)

• rpc: Clarify how to populate server.address and server.port when using client-side load balancing or service discovery.
  (#3258, #3317)

• service: Add service.criticality attribute to classify services based on operational importance (#2986)
  This attribute enables observability platforms to implement criticality-aware tracing, monitoring,
  and sampling strategies. Supports four levels: critical, high, medium, and low.

• service: Stabilize service.instance.id attribute. (#1519)
  The...

v1.39.0

🛑 Breaking changes 🛑

• rpc: Rename rpc.client|server.duration to rpc.client|server.call.duration, change RPC duration metrics from milliseconds to seconds, and clarify metric and span duration semantics for streaming. (#383, #2961)
• rpc: Rename rpc.grpc.request.metadata and rpc.connect_rpc.request.metadata to rpc.request.metadata
• rpc.grpc.response.metadata and rpc.connect_rpc.response.metadata to rpc.response.metadata
  (#2869, #3169)
• rpc: Merge rpc.method and rpc.service into fully-qualified rpc.method attribute. Clarify how to handle possible high cardinality of rpc.method in edge cases. Clarify rpc.method usage in span names. (#2863, #3196, #3223)
• rpc: Deprecate rpc.grpc.status_code, rpc.connect_rpc.error_code and rpc.jsonrpc.error_code attributes in favor of the more general rpc.response.status_code attribute.
  (#1504, #2920)
• rpc: Align RPC conventions with naming guidelines. Renames:
  • rpc.system to rpc.system.name. The values are also updated to match
    naming guidelines: connect_rpc is renamed to connectrpc;
    apache_dubbo is renamed to dubbo;
    java_rmi, dotnet_wcf, and onc_rpc are not included in the new enum.
  • rpc.jsonrpc.request_id to jsonrpc.request.id.
  • rpc.jsonrpc.version to jsonrpc.protocol.version.
    (#2703, #2921)
• system, linux: Rename *.linux.memory metrics and attributes to *.memory.linux (#1661)

🚩 Deprecations 🚩

• peer: The peer.service attribute has been deprecated in favor of service.peer.name. (#2945)
  The peer.service attribute has been renamed to service.peer.name to align with the service.{name|namespace} resource attributes.
• process: Introduce process.unix.file_descriptor.count and process.windows.handle.count metrics.
  Deprecate process.open_file_descriptor.count
  (#3188)

💡 Enhancements 💡

• cloud: Add gcp.agent_engine as a value for cloud.platform (#2957)

• cloud: Add Oracle Cloud Infrastructure to SemConv Areas list (#3190)

• cloud: Add Hetzner Cloud to cloud.provider and cloud.platform (#2758)

• cloud: Add Linode (Akamai Cloud) to cloud.provider and linode_cloud_compute to cloud.platform (#2756)

• cloud: Add Vultr Cloud to cloud.provider and cloud.platform (#2757)

• k8s: Add missing entity associations to k8s and container metrics (#3213)

• k8s: Adds k8s.pod.ip, k8s.pod.hostname and k8s.pod.start_time attributes (#3171)

• k8s: Define missing roles for entity attributes (#3135)

• k8s: Promote a selection of k8s and container attributes to alpha (#3120)
  The selected attributes being promoted are used by the Collector Contrib components
  that are targeting stability. See opentelemetry-collector-contrib/#44130

• k8s: Define roles for k8s entities' attributes (#3017)

• mcp: Add MCP semantic conventions (#2043, #2083)

• otel: Add otel.event.name attribute for use by non-OTLP exporters and logging libraries. (#2913)
  The otel.event.name attribute can be used by:

  • Non-OTLP exporters to emit the EventName
  • Applications using existing logging libraries to add event name information that can be used to set the EventName field by Collector or SDK components

• pprof: add various pprof attributes to support lossless conversion from and to OTel profiles. (#3078)

• rpc: Use consistent set of attributes between RPC spans and metrics (#2922, #3197)

  • Use rpc.response.status_code on common metrics and spans.
  • Make error.type note consistent between metrics and spans.
  • Promote notes on rpc.method and rpc.service to attribute definition.

• rpc: Mark rpc.system.name, rpc.method, rpc.service, server.address, and server.port as sampling relevant attributes in RPC spans.
  (#3202)

• service: Splits service into three entities: namespace, service and instance. (#2880)

• service: service.peer.name and service.peer.namespace have been introduced as the replacement for peer.service to describe remote services that use both a name and a namespace. (#2945)
  The peer.service attribute could not fully represent the service.{name|namespace} resource attributes;
  service.peer.name and service.peer.namespace provide a more complete representation.

🧰 Bug fixes 🧰

• rpc: Demote server.address from required to conditionally required (when available) on RPC spans and metrics. Update server.port requirement level accordingly. (#3236)

New Contributors

• @paulojmdias made their first contribution in #2756
• @codeboten made their first contribution in #3164
• @marleypowell made their first contribution in #3166
• @mmanciop made their first contribution in #3097

Full Changelog: v1.38.0...v1.39.0

v1.38.0

🛑 Breaking changes 🛑

• k8s: Rename k8s.node.allocatable.* metrics to k8s.node.*.allocatable (#2973)
  Renamed metrics:

  • k8s.node.allocatable.cpu -> k8s.node.cpu.allocatable
  • k8s.node.allocatable.memory -> k8s.node.memory.allocatable
  • k8s.node.allocatable.ephemeral_storage -> k8s.node.ephemeral_storage.allocatable

• k8s: Fix plural-named updowncounter k8s metrics (#2301)
  Renamed metrics:

  • k8s.node.allocatable.pods -> k8s.node.pod.allocatable
  • k8s.daemonset.current_scheduled_nodes -> k8s.daemonset.node.current_scheduled
  • k8s.daemonset.desired_scheduled_nodes -> k8s.daemonset.node.desired_scheduled
  • k8s.daemonset.misscheduled_nodes -> k8s.daemonset.node.misscheduled
  • k8s.daemonset.ready_nodes -> k8s.daemonset.node.ready
  • k8s.job.active_pods -> k8s.job.pod.active
  • k8s.job.failed_pods -> k8s.job.pod.failed
  • k8s.job.successful_pods -> k8s.job.pod.successful
  • k8s.job.desired_successful_pods -> k8s.job.pod.desired_successful
  • k8s.job.max_parallel_pods -> k8s.job.pod.max_parallel
  • k8s.cronjob.active_jobs -> k8s.cronjob.job.active
  • k8s.deployment.desired_pods -> k8s.deployment.pod.desired
  • k8s.deployment.available_pods -> k8s.deployment.pod.available
  • k8s.replicaset.desired_pods -> k8s.replicaset.pod.desired
  • k8s.replicaset.available_pods -> k8s.replicaset.pod.available
  • k8s.replicationcontroller.desired_pods -> k8s.replicationcontroller.pod.desired
  • k8s.replicationcontroller.available_pods -> k8s.replicationcontroller.pod.available
  • k8s.statefulset.desired_pods -> k8s.statefulset.pod.desired
  • k8s.statefulset.ready_pods -> k8s.statefulset.pod.ready
  • k8s.statefulset.current_pods -> k8s.statefulset.pod.current
  • k8s.statefulset.updated_pods -> k8s.statefulset.pod.updated
  • k8s.hpa.desired_pods -> k8s.hpa.pod.desired
  • k8s.hpa.current_pods -> k8s.hpa.pod.current
  • k8s.hpa.max_pods -> k8s.hpa.pod.max
  • k8s.hpa.min_pods -> k8s.hpa.pod.min

• process: Renames process.context_switch_type to process.context_switch.type (#1501)

• rpc: Attributes for rpc metric are now explicitly associated with metrics. (#2311)
  The breaking change is due to the removal of network.type from the metrics.

• system: Mark cpu.logical_number attribute as opt-in for system.cpu.time and system.cpu.utilization metrics (#2932)

• system, process: Rename system.process.status to process.state (#1803)

• system, process: Replace system.paging.type and process.paging.fault_type with system.paging.fault.type (#1803)

• v8js: Rename metric v8js.heap.space.available_size to v8js.memory.heap.space.available_size.
  Rename metric v8js.heap.space.physical_size to v8js.memory.heap.space.physical_size.
  (#2856)

🚩 Deprecations 🚩

• rpc: Deprecate rpc..requests_per_rpc and rpc..responses_per_rpc metrics. (#2784)
• system: Fix deprecation of system.cpu.logical_number (#2700)

💡 Enhancements 💡

• app: Add app.screen.name attribute to capture application screen names. (#2743)

• cicd: Add guidance on per pipeline run metrics (#1184)

• cicd: Improve the description of cicd.worker entities. (#1191)

• cicd: Add guidance on CI/CD logs (#1714)

• db: Add database context propagation via V$SESSION.ACTION for SQL Server (#2610)

• gcp: Document gcp.apphub_destination.* attributes (#2649)

• gen-ai: Add participant's name on generative AI chat message. (#2943)

• gen-ai: Add tool definitions and execution details for Single and Multi-Agent systems

• Add 'gen_ai.tool.definitions' attribute to 'invoke_agent' span

• Add 'gen_ai.tool.definitions' attribute to 'attributes.gen_ai.inference.client' attribute group.

• Add 'gen_ai.tool.call.arguments' and 'gen_ai.tool.call.result' attributes to 'execute_tool' span.
  (#2444, #591)

• gen-ai: Introducing Evaluation Event in GenAI Semantic Conventions to represent and capture evaluation results.
  (#2563)

• gen-ai: invoke_agent spans now include span kind (CLIENT/INTERNAL) guidance and clarify when server.* attributes should be set. (#2837)

• gen-ai: Add reasoning content message parts for chat messages. (#1965)

• gen_ai: Add gen_ai.embeddings.dimension.count attribute to capture the dimension count of embeddings. (#2361)
  The gen_ai.embeddings.dimension.count is added to the span.gen_ai.embeddings.client span
  and gen_ai.client.operation.duration metric.

• gen_ai: Add multimodal uri, file, and blob parts to GenAI JSON Schemas (#1556)

• gen_ai, db, enduser, messaging: Fix deprecation reasons - use uncategorized when attribute type is changing. Fix missing/invalid deprecations on (deprecated) gen_ai.system attribute.
  (#2688)

• http: Adds in the HTTP QUERY method. (#2642)

• http: Clarify value for http.route attribute may contain static and dynamic segments. (#2616)

• k8s: Add container CPU limit/request utilization metrics. (#1489)

• k8s: Add k8s.pod.status.phase and k8s.pod.status.reason metrics (#2075)

• k8s: Add memory metrics for k8s.node, k8s.pod and container resources (#2776)

• nfs: Add nfs.* metrics, along with supporting rpc.onc.*/rpc.nfs.* attributes. (#39978)

• openshift: Adds metrics for openshift's clusterquota (#2078)

• pprof: Introduce pprof attributes. (#2522)

• pprof: add attribute pprof.profile.comment. (#2861)

• process: Increases the amount of attributes which are used to describe the running process. (#2655)

• rpc: JSON-RPC now has its own span definition. (#2228)

• rpc: Adds in network.protocol.name and network.protocol.version to all RPC spans and metrics. (#2719)

• rpc: Adds in explicit span definitions for Connect RPC. (#2720)

• rpc: Remove network.type attribute from RPC spans and metrics (#2851)
  The network.type attribute is no longer included in RPC semantic conventions.
  It remains available in the attribute registry for opt-in usage when needed.

• rpc: Adds in explicitly defined grpc spans. (#2720)

• rpc: Update the RPC stability warning and guidance for instrumentation authors (#2718)

• rpc: The rpc spans now contains the error.type attribute to report the type of error that occurred. (#2812)

🧰 Bug fixes 🧰

• dns: Simplify dns.question.name note to remove incorrect custom encoding requirements (#2143)
  The note now simply states that the domain name should be captured as it appears in the DNS query
  without any additional normalization, removing the previous incorrect custom encoding requirements.

• nfs: nfs/metrics.yaml: nfs.server.net.count: add missing network.transport attribute. Implementation not merged, so not a breaking change. (#39978)

• system: The system.network.dropped and system.network.packets metrics have been added as deprecated rather than being removed. ([#2828](https://github.com/open-te...

v1.37.0

🛑 Breaking changes 🛑

• gen-ai: Revamp how GenAI chat history is recorded. Instead of per-message events,
  we now have gen_ai.system_instructions, gen_ai.input.messages, and
  gen_ai.output.messages attributes that can appear on GenAI spans or the new
  gen_ai.client.inference.operation.details event.
  New attributes are not recorded by default when content capturing is disabled.
  Deprecations:
  • gen_ai.system.message event - use gen_ai.system_instructions or
    gen_ai.input.messages attributes instead.
  • gen_ai.user.message, gen_ai.assistant.message, gen_ai.tool.message events
    (use gen_ai.input.messages attribute instead)
  • gen_ai.choice event (use gen_ai.output.messages attribute instead)
    (#2010, #2179, #1913, #1621, #1912)
• container: The container runtime can now also have the version and description provided (#2342)
• gen-ai: Follow system-specific naming policy in GenAI semantic conventions.
  • Rename gen_ai.system to gen_ai.provider.name
  • Remove gen_ai prefix from gen_ai.openai.* attributes.
  • Rename az.ai.* attribute names to azure.ai.*.
    (#2046)
• system: Rename system.network.dropped to system.network.packet.dropped and system.network.packets to system.network.packet.count (#1700)

💡 Enhancements 💡

• all: Add metric_value_type code-generation annotation to all metric definitions. (#2444, #591)
• http: Clarify how server.address and server.port should be populated by HTTP client instrumentations. (#2443, #2463)
• otel: Add otel.scope.schema_url attribute to non-OTLP exporters that represents an instrumentation scope schema URL. (#2320, #2489)
• aspnetcore: Add ASP.NET Core Identity metrics and update the registry. (#2509)
• gen-ai: Add operation name and tool type attributes for Execute tool span. (#2525, #2526)
• aspnetcore: Add ASP.NET Core authentication and authorization metrics and update the registry. (#2530)
• aspnetcore: Add ASP.NET Core memory pool metrics and update the registry. (#2532)
• app: Defines a new app.build_id attribute. (#2591)
• k8s: Add k8s node, pod, container filesystem metrics (#1488)
• k8s: Add k8s Pod Volume metrics (#1485)
• messaging: Added in AWS SNS as messaging systems (#2266)
• system: Used memory should be based on operating system's internals (#2534)
• db: Add context propagation via SQL commenter for databases (#2162)
• hardware: Move text hardware metrics to the hardware components, yaml and md files (#2380)
• app: Defines a new jank event in the app domain (#2552)
• docs: Document the "T-Shaped" notion for defining semantic conventions. (#2232)

New Contributors

• @martincostello made their first contribution in #2500
• @joschi made their first contribution in #2539
• @qnnn made their first contribution in #2526
• @LikeTheSalad made their first contribution in #2591
• @Cirilla-zmh made their first contribution in #2679

Full Changelog: v1.36.0...v1.37.0

v1.36.0

🚩 Deprecations 🚩

• os: Adds the 'deprecated:' tag/attribute to the z_os enum value of os.type. This value was recently deprecated in v1.35.0. (#2479)

💡 Enhancements 💡

• otel: Replaces otel.sdk.span.ended with otel.sdk.span.started and allow differentiation based on the parent span origin (#2431)
• db: Add database context propagation via SET CONTEXT_INFO for SQL Server (#2162)
• entities: Adds support for Entity registry and Entity stabilization policies. (#2246)

🧰 Bug fixes 🧰

• cloud: Exclude deprecated Azure members from code generation on the cloud.platform attribute (#2477, #2455)

Full Changelog: v1.35.0...v1.36.0