open-telemetry/opentelemetry-collector-contrib v0.100.0

on GitHub

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

End User Changelog

🛑 Breaking changes 🛑

• receiver/hostmetrics: enable feature gate receiver.hostmetrics.normalizeProcessCPUUtilization (#31368)
  This changes the value of the metric process.cpu.utilization by dividing it by the number of CPU cores.
  For example, if a process is using 2 CPU cores on a 16-core machine,
  the value of this metric was previously 2, but now it will be 0.125.

• testbed: Remove deprecated GetAvailablePort function (#32800)

🚀 New components 🚀

• healthcheckv2extension: Introduce the skeleton for the temporary healthcheckv2 extension. (#26661)
• intervalprocessor: Implements the new interval processor. See the README for more info about how to use it (#29461)
• OpenTelemetry Protocol with Apache Arrow Receiver: Implementation copied from opentelemetry/otel-arrow repository @v0.20.0. (#26491)
• roundrobinconnector: Add a roundrobin connector, that can help single thread components to scale (#32853)

💡 Enhancements 💡

• telemetrygen: Add support to set metric name (#32840)

• exporter/kafkaexporter: Enable setting message topics using resource attributes. (#31178)

• exporter/datadog: Introduces the Datadog Agent logs pipeline for exporting logs to Datadog under the "exporter.datadogexporter.UseLogsAgentExporter" feature gate. (#32327)

• elasticsearchexporter: Add retry.retry_on_status config (#32584)
  Previously, the status codes that trigger retries were hardcoded to be 429, 500, 502, 503, 504.
  It is now configurable using retry.retry_on_status, and defaults to [429, 500, 502, 503, 504] to avoid a breaking change.
  To avoid duplicates, it is recommended to configure retry.retry_on_status to [429], which would be the default in a future version.

• exporter/splunkhec: add experimental exporter batcher config (#32545)

• windowsperfcountersreceiver: Returns partial errors for failures during scraping to prevent throwing out all successfully retrieved metrics (#16712)

• jaegerencodingextension: Promote jaegerencodingextension to alpha (#32699)

• kafkaexporter: add an ability to publish kafka messages with message key based on metric resource attributes - it will allow partitioning metrics in Kafka. (#29433, #30666, #31675)

• cmd/opampsupervisor: Switch the OpAMP Supervisor's bootstrap config to use the nopreceiver and nopexporter (#32455)

• otlpencodingextension: Move otlpencodingextension to alpha (#32701)

• prometheusreceiver: Prometheus receivers and exporters now preserve 'unknown', 'info', and 'stateset' types. (#16768)
  It uses the metric.metadata field with the 'prometheus.type' key to store the original type.

• ptracetest: Add support for ignore scope span instrumentation scope information (#32852)

• sqlserverreceiver: Enable direct connection to SQL Server (#30297)
  Directly connecting to SQL Server will enable the receiver to gather more metrics
  for observing the SQL Server instance. The first metric added with this update is
  sqlserver.database.io.read_latency.

• connector/datadog: The Datadog connector now has a config option to identify top-level spans by span kind. This new logic can be enabled by setting traces::compute_top_level_by_span_kind to true in the Datadog connector config. Default is false. (#32005)
  traces::compute_top_level_by_span_kind needs to be enabled in both the Datadog connector and Datadog exporter configs if both components are being used.
  With this new logic, root spans and spans with a server or consumer span.kind will be marked as top-level. Additionally, spans with a client or producer span.kind will have stats computed.
  Enabling this config option may increase the number of spans that generate trace metrics, and may change which spans appear as top-level in Datadog.

• exporter/datadog: The Datadog exporter now has a config option to identify top-level spans by span kind. This new logic can be enabled by setting traces::compute_top_level_by_span_kind to true in the Datadog exporter config. Default is false. (#32005)
  traces::compute_top_level_by_span_kind needs to be enabled in both the Datadog connector and Datadog exporter configs if both components are being used.
  With this new logic, root spans and spans with a server or consumer span.kind will be marked as top-level. Additionally, spans with a client or producer span.kind will have stats computed.
  Enabling this config option may increase the number of spans that generate trace metrics, and may change which spans appear as top-level in Datadog.

• exporter/datadog: Support stable semantic conventions for HTTP spans (#32823)

• cmd/opampsupervisor: Persist collector remote config & telemetry settings (#21078)

• cmd/opampsupervisor: Support AcceptsRestartCommand Capability. (#21077)

• telemetrygen: Add headers to gRPC metadata for logs (#32668)

• sshcheckreceiver: Add support for running this receiver on Windows (#30650)

• zipkinencodingextension: Move zipkinencodingextension to alpha (#32702)

🧰 Bug fixes 🧰

• prometheusremotewrite: Modify prometheusremotewrite.FromMetrics to only generate target_info if there are metrics, as otherwise you can't deduce the timestamp. (#32318)

• prometheusremotewrite: Change prometheusremotewrite.FromMetrics so that the target_info metric is only generated if at least one identifying OTel resource attribute (service.name and/or service.instance.id) is defined. (#32148)

• k8sclusterreceiver: Fix container state metadata (#32676)

• sumologicexporter: do not replace . with _ for prometheus format (#31479)

• pkg/stanza: Allow sorting by ascending order when using the mtime sort_type. (#32792)

• opampextension: Add a new ppid parameter that can be used to enable orphan detection for the supervisor. (#32189)

• awsxrayreceiver: Retain CloudWatch Log Group when translating X-Ray segments (#31784)

• pkg/stanza: Fix issue when exclude_older_than is enabled without ordering_criteria configured (#32681)

• awskinesisexporter: the compressor was crashing under high load due it not being thread safe. (#32589)
  removed compressor abstraction and each execution has its own buffer (so it's thread safe)

• filelogreceiver: When a flush timed out make sure we are at EOF (can't read more) (#31512, #32170)

• vcenterreceiver: Adds the vcenter.cluster.name resource attribute to resource pool with a ClusterComputeResource parent (#32535)

• vcenterreceiver: Updates vcenter.cluster.memory.effective (primarily that the value was reporting MiB when it should have been bytes) (#32782)

• vcenterreceiver: Adds warning to vcenter.cluster.memory.used metric if configured about its future removal (#32805)

• vcenterreceiver: Updates the vcenter.cluster.vm.count metric to also report suspended VM counts (#32803)

• vcenterreceiver: Adds vcenter.datacenter.name attributes to all resource types to help with resource identification (#32531)

• vcenterreceiver: Adds vcenter.cluster.name attributes warning log related to Datastore resource (#32674)

• vcenterreceiver: Adds new vcenter.virtual_app.name and vcenter.virtual_app.inventory_path resource attributes to appropriate VM Resources (#32557)

• vcenterreceiver: Adds functionality for vcenter.vm.disk.throughput while also changing to a gauge. (#32772)

• vcenterreceiver: Adds initially disabled functionality for VM Templates (#32821)

• remotetapprocessor: Fix memory leak on shutdown (#32571)

• haproxyreceiver: Fix reading stats larger than 4096 bytes (#32652)

• connector/count: Fix handling of non-string attributes in the count connector (#30314)

• datadogexporter: Fix nil pointer dereference when using beta infrastructure monitoring offering (#32865)
  The bug happened under the following conditions:

  • Setting datadog.host.use_as_host_metadata to true on a payload with data about the Datadog exporter host
  • Running using the official opentelemetry-collector-contrib Docker image

• pkg/translator/jaeger: translate binary attribute values to/from Jaeger as is, without encoding them as base64 strings (#32204)

• awscloudwatchreceiver: Fixed a bug where autodiscovery would not use nextToken in the paginated request (#32053)

• awsxrayexporter: make comma, as invalid char for x-ray segment name (#32610)

API Changelog

🛑 Breaking changes 🛑

• pkg/stanza: Pass TelemetrySettings to the Build method of the Builder interface (#32662, #31256)
  The reason for this breaking change is to pass in the component.TelemetrySettings
  so as to use them later in various ways:
  • be able to report state statistics and telemetry in general
  • be able to switch from SugaredLogger to Logger

🚩 Deprecations 🚩

• confmap/provider/s3: Deprecate s3provider.New in favor of s3provider.NewFactory (#32742)
• confmap/provider/secretsmanager: Deprecate secretsmanagerprovider.New in favor of secretsmanagerprovider.NewFactory (#32743)

🚀 New components 🚀

• roundrobinconnector: Add a roundrobin connector, that can help single thread components to scale (#32853)

💡 Enhancements 💡

• opampextension: Added support for other components to register custom capabilities and receive custom messages from an opamp extension (#32021)
• kafkaexporter: add an ability to publish kafka messages with message key based on metric resource attributes - it will allow partitioning metrics in Kafka. (#29433, #30666, #31675)
• sshcheckreceiver: Add support for running this receiver on Windows (#30650)