The OpenTelemetry Collector Contrib contains everything in the opentelemetry-collector release, be sure to check the release notes there as well.

Unmaintained Components

These components are marked as Unmaintained and will eventually be removed from our community distributions. If you depend on these components we need your help to support them.

• receiver/bigip
• exporter/alibabacloudlogservice

End User Changelog

🛑 Breaking changes 🛑

• all: Increase minimum Go version to 1.25 (#46000)

• exporter/elasticsearch: Remove ecs mode log enrichment for agent.name and agent.version (#45849) The removed log enrichments are duplicate code that already exists in the github.com/elastic/opentelemetry-collector-components/processor/elasticapmprocessor. It is recommended to use the github.com/elastic/opentelemetry-collector-components/processor/elasticapmprocessor when using mapping mode ecs to ensure index documents contain all required Elastic fields to power the Kibana UI.

• processor/akamaidetector: Update semantic conventions to v1.39.0 and use the convention for cloud.platform in Akamai detector (#45511)

• processor/k8s_attributes: Introduce semantic conventions compliant feature gate pair for k8sattributes processor (#44693)

  • Added processor.k8sattributes.EmitV1K8sConventions feature gate to enable stable semantic convention attributes (singular form: k8s.<workload>.label.<key> and k8s.<workload>.annotation.<key>)
  • Added processor.k8sattributes.DontEmitV0K8sConventions feature gate to disable legacy non-compliant attributes (plural form: k8s.<workload>.labels.<key> and k8s.<workload>.annotations.<key>)
  • Both feature gates are in alpha stage and disabled by default
  • The processor now validates that legacy attributes cannot be disabled without enabling stable attributes
  • Deprecated k8sattr.labelsAnnotationsSingular.allow feature gate in favor of the new semconv-compliant gates (will be removed in v0.150.0)
  • During migration period, both legacy and stable attributes can coexist when EmitV1K8sConventions is enabled but DontEmitV0K8sConventions is not

• processor/resourcedetection: Promote processor.resourcedetection.propagateerrors feature gate to Stable and is now always enabled (#44609)

• receiver/hostmetrics: process.context_switches will now properly count context switches for all threads. (#36804) Previously, only the lead thread's context switches would be counter. We believe this was a bug, but are marking it as a breaking change since the values of these metrics may change drastically compared to previous expectations. However, the values you get now will be more accurate than before.

• receiver/kafkametrics: receiver.kafkametricsreceiver.UseFranzGo feature gate is now in Beta stage and enabled by default (#44600)

🚩 Deprecations 🚩

• connector/signal_to_metrics: Rename signaltometrics connector to signal_to_metrics and add deprecated alias signal_to_metrics (#45551)

• exporter/alibabacloud_logservice: Marking component unmaintained (#46029)

• processor/datadogsemantics: Deprecate the datadogsemantics processor. (#46052) If you are using this component, please contact Datadog support.

• processor/k8s_attributes: Rename k8sattributes processor to k8s_attributes processor and add deprecated alias k8sattributes. (#45894)

• receiver/kubeletstats: The GCEPersistentDisk, AWSElasticBlockStore, and Glusterfs have been deprecated as these have been depreacted in k8s (#40477) K8s documentation for the three deprecated volume types is below:

  • GCEPersistentDisk: https://v1-32.docs.kubernetes.io/docs/concepts/storage/volumes/#gcepersistentdisk
  • AWSElasticBlockStore: https://v1-32.docs.kubernetes.io/docs/concepts/storage/volumes/#awselasticblockstore
  • Glusterfs: https://v1-32.docs.kubernetes.io/docs/concepts/storage/volumes/#glusterfs

• receiver/signalfx: This receiver is deprecated. Please use the OTLP receiver instead (#46079) This component will be removed in a future release.

• receiver/wavefront: Deprecate the wavefront receiver (#46087) There is no replacement for the wavefront receiver. Wavefront is EOL: https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/release-announcements/0/25153

🚀 New components 🚀

• receiver/vcr: First PR for new receiver (#42877)

💡 Enhancements 💡

• cmd/golden: Golden tool now returns all comparison errors with attempt numbers instead of only the last error (#45424) When multiple metric payloads fail validation, the golden tool now displays all errors with their attempt numbers, making it easier to debug test failures. Previously, only the last error was shown.

• connector/signal_to_metrics: Add error_mode configuration to control OTTL evaluation error handling. Prevents single bad record from failing entire batch. (#38826, #45746) Adds three error modes:

  • propagate (default): Returns error, entire batch fails (backward compatible)
  • ignore: Logs error, skips bad record, continues processing valid records
  • silent: Skips bad record silently, continues processing valid records

• exporter/datadog: Add new alpha feature gate 'exporter.datadogexporter.DisableAllMetricRemapping' to disable all metric remapping in the Datadog exporter. (#45943) The feature gate is marked as alpha pending changes in Datadog's backend.

• exporter/elasticsearch: Update the ECS mode metrics data point hasher to exclude the elasticsearch.mapping.hints attribute (#45887) Excluding the elasticsearch.mapping.hints attribute will allow similar metric data points to be grouped together and indexed to the same document.

• exporter/elasticsearch: Add traces_dynamic_id config to dynamically set document IDs for traces and span events (#43649) Adds traces_dynamic_id configuration option to allow setting document IDs based on span and span event attributes using the elasticsearch.document_id attribute. This prevents duplicate documents from being created when the same span is sent multiple times, similar to the existing logs_dynamic_id feature. Disabled by default.

• exporter/file: Add support for rotation when group_by is enabled in file exporter (#43143)

• exporter/googlecloudstorage: Add compression support for Google Cloud Storage exporter (#45337) The Google Cloud Storage exporter now supports compression of log data before uploading to GCS. Supported compression algorithms: gzip and zstd.

• exporter/kafka: Add conn_idle_timeout configuration option to control when idle connections are not reused and may be closed. (#45321) Defaults to 9 minutes.

• extension/awscloudwatchmetricstreams_encoding: Add support for extracting percentile statistics (p50, p90, p99, etc.) from CloudWatch Metric Streams JSON format (#45855) The JSON unmarshaler now extracts percentile fields from CloudWatch Metric Streams data and converts them to OpenTelemetry Summary quantiles. This enables feature parity with the embedded cwmetricstream unmarshaler in awsfirehosereceiver.

• extension/awslogs_encoding: Support CloudWatch Logs extracted fields (@aws.account, @aws.region) for centralized logging (#45792) CloudWatch Logs subscription filter unmarshaler now supports extracted fields (@aws.account and @aws.region) that are automatically added when using CloudWatch Logs centralization and enabling emitSystemFields in subscription filters. This enhancement enables proper resource attribution in OpenTelemetry when processing logs from multiple AWS accounts and regions. Logs with different extracted field values are automatically grouped into separate ResourceLogs for proper semantic convention mapping:

  • @aws.account maps to cloud.account.id
  • @aws.region maps to cloud.region

• extension/awslogs_encoding: Handle multiple concatenated JSON objects for AWS CloudWatch Log subscription (#46120)

• extension/azureauth: Add and implement new method Token(context.Context) (*oauth2.Token, error). (#45064)

• extension/encoding: Introduce streaming support for encoding extensions (#38780)

• extension/oidc: Adds public_keys_file to the provider config. When set, keys are loaded from a local JWKS file instead of using remote discovery. (#44899) The file is watched for changes and keys are automatically reloaded on update. Supported key types are RSA, ECDSA, and ED25519.

• pkg/ottl: Add IsInCIDR function to check if IP belongs to given list of CIDR (#42215)

• pkg/sampling: Optimize OTel tracestate parsing by replacing regex validation with hand-written validator (10-21x faster). (#45539)

• pkg/sampling: Replace regex-based W3C tracestate validation with hand-written validator for 30-65x performance improvement (#45734)

• pkg/stanza: Ensure filter operator does not split batches of entries (#42391)

• processor/filter: Introduces inferred context conditions for filtering (#37904) Introduces three new top-level config fields [metric_conditions, log_conditions, trace_conditions]. A user can supply OTTL conditions for each without needing to supply context.

• processor/k8s_attributes: Added container.image.tags resource attribute with feature gate controls according to OpenTelemetry semantic conventions. (#44589)

• processor/lookup: Add lookup processor implementation and YAML source (#41816) Adds the core lookup processor implementation for enriching telemetry data using external lookups. Includes YAML file source for loading lookup tables from local files.

• processor/vultrdetector: Update semantic conventions to v1.39.0 and add support for cloud.platform in Vultr detector (#45512)

• receiver/datadog: Add support for handling the /api/v0.2/stats endpoint to receive and process APM trace stats payloads from the Datadog Agent. (#45778) The Datadog Receiver can now process APM trace stats payloads sent by the Datadog Agent via the /api/v0.2/stats endpoint. The handler correctly processes gzipped msgpack payloads, decodes them into StatsPayload, translates them into OpenTelemetry-compatible metrics, and forwards them to the configured metrics consumer. This enables the complete APM metrics flow: Application → Datadog SDK → Datadog Agent → OpenTelemetry Collector → OTEL Backends.

• receiver/hostmetrics: Add optional system.memory.linux.shared metric (#32712) This metric reports shared memory usage, including tmpfs filesystems, System V shared memory, and POSIX shared memory. Currently only available on Linux systems due to platform-specific data availability. This corresponds to the Shmem field in /proc/meminfo.

• receiver/k8s_cluster: Add opt-in service metrics derived from k8s Service and EndpointSlice API (#45620) New metrics (disabled by default):

  • k8s.service.endpoint.count: Number of endpoints by condition (ready, serving, terminating), address type, and zone
  • k8s.service.load_balancer.ingress.count: Number of load balancer ingress points assigned to the service New resource attributes:
  • k8s.service.name: The k8s service name
  • k8s.service.uid: The k8s service uid
  • k8s.service.type: The k8s service type
  • k8s.service.traffic_distribution: The service's traffic routing preference
  • k8s.service.publish_not_ready_addresses: Whether the service publishes endpoints before pods are ready

• receiver/kafka: Add conn_idle_timeout configuration option to control when idle connections are not reused and may be closed. (#45321) Defaults to 9 minutes.

• receiver/mongodb: Add support for auth_mechanism, auth_source, and auth_mechanism_properties configuration options (#40686) Users can now specify the authentication mechanism (e.g., SCRAM-SHA-256, GSSAPI, MONGODB-AWS), auth source database, and auth mechanism properties when connecting to MongoDB instances. This is particularly useful for MongoDB servers that require specific authentication mechanisms. For example, GSSAPI (Kerberos) may require SERVICE_NAME, and MONGODB-AWS may require AWS_SESSION_TOKEN when using temporary AWS credentials.

• receiver/pprof: Implement the functionality of transforming pprof to OTel Profiles (#45411)

• receiver/prometheusremotewrite: Improved performance when parsing Remote Write v2 requests. (#45623)

• receiver/prometheusremotewrite: Add exemplar support to the Prometheus Remote Write receiver (#44983)

• receiver/redfish: Change system.host_name and base_url as resource attributes. (#45470)

• receiver/sqlquery: Add support for initial_delay in logs collection. (#29671) Log collection now applies initial_delay (previously ignored). If initial_delay is not set, the first log collection now occurs at 1 second, instead of occurring after collection_interval time has passed.

• receiver/sqlserver: Add the sqlserver.procedure_id and sqlserver.procedure_name attributes to TopQuery and Sample Events (#44656) Refined query and reported events to include stored procedure information when applicable. Additionally, the maximum number of active queries reported by default has been increased from 200 to 250 to account for record deaggregation introduced by this change, ensuring the effective limit remains consistent with the previous 200-query baseline.

• receiver/statsd: Discard StatsD metrics with NaN or infinite values to prevent invalid data from entering the metric pipeline (#44288)

• receiver/syslog: Add facility_text attribute to syslog parser output (#45641) The syslog parser now outputs a facility_text attribute containing the human-readable facility name (e.g., "auth", "kern", "local0") in addition to the existing numeric facility attribute.

🧰 Bug fixes 🧰

• exporter/datadog: OTLP logs now support array type attributes. Arrays containing primitive values or nested maps are now correctly preserved in the log output. (#45708)

• exporter/datadog: Fix data race in the Datadog exporter which could cause a crash with error message "concurrent map iteration and map write". (#46051) Specifically, when processing spans with the datadog.host.use_as_metadata attribute.

• exporter/elasticsearch: Fix ECS mode to properly protect known schema fields from getting .value suffix when conflicting with nested attributes (#37211) Previously, when ECS mode was enabled and attributes like process.executable.name conflicted with the known ECS field process.executable, the deduplication logic would incorrectly add a .value suffix to the known field, resulting in process.executable.value. This fix ensures protected ECS fields remain unchanged and conflicting nested attributes are properly ignored.

• exporter/opensearch: Fix sending_queue not using default values for num_consumers and queue_size when only batch is configured (#45016)

• exporter/syslog: Update the timestamp when using the RFC 3164 formatter to space-pad the day of month for single digit days (#46115)

• extension/awslogs_encoding: Fix duplicate resource attributes in subscription filter unmarshaler (#45792) The aws.log.group.names and aws.log.stream.names resource attributes were incorrectly being set twice: first as array values, then immediately overwritten as string values. This fix removes the duplicate string assignments, ensuring the attributes are correctly set only as arrays per OpenTelemetry semantic conventions.

• extension/oauth2client: Fix oauth2clientauth client-credentials grant type (#45786)

• extension/text_encoding: Fix text encoding extension to not split large messages when no separator is configured. (#45845)

• pkg/stanza: Fix recombine operator logging errors at ERROR level when on_error is set to quiet mode (#42646)

• pkg/translator/prometheusremotewrite: Fix export of Instrumentation Scope attributes as Prometheus labels. (#45912) Instrumentation Scope attributes (name, version, and other attributes) are now correctly translated to Prometheus labels with the otel_scope_ prefix.

• pkg/xk8stest: Fix IPv6 gateway handling in HostEndpoint to avoid invalid address formatting in e2e tests (#46082) Prefer IPv4 gateways when resolving the Docker kind network gateway. Fall back to bracketed IPv6 if no IPv4 gateway is found, so that appending :port produces a valid address (e.g. [::1]:4317).

• processor/k8s_attributes: Fix concurrent map access panic by cloning pod labels and annotations before extraction. (#46112)

• processor/k8s_attributes: Allow key_regex to work without tag_name by using the default tag name format (#45719) When using key_regex with capturing groups but without specifying tag_name, the processor now correctly uses the default tag name format (e.g., k8s.pod.labels.<label_key>) instead of producing empty tag names.

• processor/redaction: Improve database sanitization with system-aware obfuscation, span name sanitization, and URL path parameter redaction. (#44229)

  • Database sanitization now validates span kind (CLIENT/SERVER/INTERNAL ) and requires db.system.name/db.system attribute for traces/metrics
  • Implemented span name obfuscation for database operations based on db.system
  • Added URL path parameter sanitization for span names with configurable pattern matching
  • Improved query validation database sanitizers
  • Fix issue ensuring no spans with ... name can be generated due to enabling multiple sanitizers
  • If something went wrong during span name sanitization, original span name is used

• receiver/azure_event_hub: Fixes a bug where the receiver would stop receiving messages after a parsing error. (#45898)

• receiver/faro: Updates Faroreceiver to return HTTP 202 Accepted status code upon successful data ingestion to comply with the OpenAPI specification. (#45648)

• receiver/fluentforward: handle uint64 to int64 overflow by changing to string (#45252) FluentD supports record entries with uint64 types. OpenTelemetry log attributes only support int64 and no uint64. The old solution would overflow with uint64 values greater than math.MaxInt64 and result in negative attribute values. This fix changes that behaviour by storing only those large values as string attributes instead.

• receiver/googlecloudpubsub: Fix compression detection when both encoding and compression are set in the config (#45810)

• receiver/mongodb: Check if metrics are enabled before collecting them to prevent errors when metrics are disabled. (#41465)

• receiver/postgresql: Updated the default value for top_n_query (200) to match with other db receivers (#45612)

API Changelog

🚩 Deprecations 🚩

• pkg/stanza: Package "github.com/open-telemetry/opentelemetry-collector-contrib/pkg/stanza/operator/parser/json" has been deprecated. Use "github.com/open-telemetry/opentelemetry-collector-contrib/pkg/stanza/operator/parser/jsonparser" instead (#45006)
• pkg/stanza: Package "github.com/open-telemetry/opentelemetry-collector-contrib/pkg/stanza/errors" has been deprecated. Use "github.com/open-telemetry/opentelemetry-collector-contrib/pkg/stanza/stanzaerrors" instead (#45006)
• pkg/stanza: Package "github.com/open-telemetry/opentelemetry-collector-contrib/pkg/stanza/operator/parser/time" has been deprecated. Use "github.com/open-telemetry/opentelemetry-collector-contrib/pkg/stanza/operator/parser/timeparser" instead (#45006)

💡 Enhancements 💡

• processor/filter: Introduces inferred context conditions for filtering (#37904) Introduces three new top-level config fields [metric_conditions, log_conditions, trace_conditions]. A user can supply OTTL conditions for each without needing to supply context.

• receiver/pprof: Implement the functionality of transforming pprof to OTel Profiles (#45411)

🧰 Bug fixes 🧰

• processor/sumologic: Export config types to allow programmatic configuration via Go API (#45880)
• receiver/filestats: Ensure that bsd build tags are respected by renaming filestats_darwin.go to filestats_bsd.go (#42645)

We are thrilled to welcome our first-time contributors to this project. Thank you for your contributions @himanshu130700, @plantfansam, @akshaysw, @ShaanveerS, @sugaf1204, @hardik-choksi, @SoumyaRaikwar, @Alex-Souslik, @robertgustafsonsplunk, @MrWong99, @ahayworth, @odorT, @natasha-aleksandrova, @yuanyuanzhao3, @pradhyum6144, @alyshawang, @RaphaelManke, @kyyril, @vanugrah, @cjksplunk ! 🎉