This release includes version 0.119.0 of the upstream Collector components.
The individual upstream Collector changelogs can be found here:
v0.119.0:
- https://github.com/open-telemetry/opentelemetry-collector/releases/tag/v0.119.0
- https://github.com/open-telemetry/opentelemetry-collector-contrib/releases/tag/v0.119.0
Highlights from the upstream Collector changelog
🛑 Breaking changes 🛑
exporterhelper
: Change queue to embed the async consumers. (#12242)cmd/mdatagen
: Allow passing OTel Metric SDK options to the generatedSetupTelemetry
function. (#12166)exporterhelper
: Rename exporter span signal specific attributes (e.g. "sent_spans" / "send_failed_span") to "items.sent" / "items.failed". (#12165)cmd/mdatagen
: Remove dead fieldtelemetry::level
(#12144)exporterhelper
: Change exporter ID to be a Span level attribute instead on each event. (#12164)
This does not have an impact on the level of information emitted, but on the structure of the Span.cmd/mdatagen
: Removelevel
field from metrics definition (#12145)
This mechanism will be added back once a new views mechanism is implemented.
💡 Enhancements 💡
-
configtls
: Allow users to mention their preferred curve types for ECDHE handshake (#12174) -
otelcol
: Adds support for listing config providers in components command's output (#11570) -
pdata/pprofile
: Add new helper methodFromAttributeIndices
to build apcommon.Map
out ofAttributeIndices
. (#12176) -
component
: Allowcomponent.ValidateConfig
to recurse through all fields in a config object (#11524) -
component
: Show path to invalid config in errors returned fromcomponent.ValidateConfig
(#12108) -
processor/transformprocessor
: Add support for flat configuration style. (#29017)
The flat configuration style allows users to configure statements by providing a list of statements instead of a
structured configuration map. The statement's context is expressed by adding the context's name prefix to path names,
which are used to infer and to select the appropriate context for the statement. -
resourcedetectionprocessor
: Expose additional configuration parameters for the AWS metadata client used by the EC2 detector (#35936)
In some cases, you might need to change the behavior of the AWS metadata client from the standard retryerBy default, the client retries 3 times with a max backoff delay of 20s.
We offer a limited set of options to override those defaults specifically, such that you can set the client to retry 10 times, for up to 5 minutes, for example:
processors: resourcedetection/ec2: detectors: ["ec2"] ec2: max_attempts: 10 max_backoff: 5m
-
cumulativetodeltaprocessor
: Add metric type filter for cumulativetodelta processor (#33673) -
resourcedetectionprocessor
: Addfail_on_missing_metadata
option on EC2 detector (#35936)
If the EC2 metadata endpoint is unavailable, the EC2 detector by default ignores the error.
By settingfail_on_missing_metadata
to true on the detector, the user will now trigger an error explicitly,
which will stop the collector from starting. -
resourcedetectionprocessor
: Thegcp
resource detector will now detect resource attributes identifying a GCE instance's managed instance group. (#36142) -
jaegerreceiver
: Log the endpoints of different servers started by jaegerreceiver (#36961)
This change logs the endpoints of different servers started by jaegerreceiver. It simplifies debugging by ensuring log messages match configuration settings. -
receiver/hostmetrics
: Added support for tracking process.uptime (#36667) -
processor/transformprocessor
: Replace parser collection implementations withottl.ParserCollection
and add initial support for expressing statement's context via path names. (#29017) -
prometheusreceiver
: Addreceiver.prometheusreceiver.UseCollectorStartTimeFallback
featuregate for the start time metric adjuster to use the collector start time as an approximation of process start time as a fallback. (#36364) -
tailsamplingprocessor
: Reworked the consume traces, sampling decision, and policy loading paths to improve performance and readability (#37560)
🧰 Bug fixes 🧰
-
exporterhelper
: Fix bug that the exporter with new batcher may have been marked as non mutation. (#12239)
Only affects users that manually turned onexporter.UsePullingBasedExporterQueueBatcher
featuregate. -
confighttp
: confighttp.ToServer now sets ErrorLog with a default logger backed by Zap (#11820)This change ensures that the http.Server's ErrorLog is correctly set using Zap's logger at the error level, addressing the issue of error logs being printed using a different logger.
-
exporterhelper
: Fix context propagation for DisabledBatcher (#12231) -
mdatagen
: Fix bug where Histograms were marked as not supporting temporaly aggregation (#12168) -
exporterhelper
: Fix MergeSplit issue that ignores the initial message size. (#12257) -
filelogreceiver
: Fix issue where flushed tokens could be truncated. (#35042) -
k8sattributesprocessor
: Wait for the other informers to complete their initial sync before starting the pod informers (#37056) -
pkg/stanza
: Fix default source identifier in recombine operator (#37210)
Its defualt value is now aligned with the semantic conventions:attributes["log.file.path"]
-
tailsamplingprocessor
: Fixed sampling decision metricsotelcol_processor_tail_sampling_sampling_trace_dropped_too_early
andotelcol_processor_tail_sampling_sampling_policy_evaluation_error_total
, these were sometimes overcounted. (#37212)
As a result of this change non-zero values ofotelcol_processor_tail_sampling_sampling_trace_dropped_too_early
andotelcol_processor_tail_sampling_sampling_policy_evaluation_error_total
metrics will be lower.
Before this fix, errors got counted several times depending on the amount of traces being processed
that tick and where in the batch the error happened.
Zero values are unaffected.