signalfx/splunk-otel-collector v0.142.0 on GitHub

This Splunk OpenTelemetry Collector release includes changes from the opentelemetry-collector v0.142.0
and the opentelemetry-collector-contrib v0.142.0 releases where appropriate.

🛑 Breaking changes 🛑

(Splunk) configserver: Removed the deprecated localhost config endpoint (http://localhost:55554/debug/configz) and its related environment variables. (#6984)
The collector's default configuration files enable the zpages extension and the respective debug/expvarz endpoint,
by default at http://localhost:55679/debug/expvarz. This endpoint provides a JSON map containing the collector
configuration details. The keys splunk.config.initial and splunk.config.effective within this map
hold, respectively, the initial and effective YAML configurations as strings.
To retrieve the configuration in YAML format using this endpoint, you can use the following commands:
On Windows PowerShell 5.1:
```
(Invoke-WebRequest http://localhost:55679/debug/expvarz).Content | ConvertFrom-Json | Select-Object -ExpandProperty splunk.config.effective
```
On Bash (with curl and jq):
```
curl http://localhost:55679/debug/expvarz --silent | jq -r '.["splunk.config.initial"]'
```
(Contrib) exporter/prometheusremotewrite: Updated to Remote Write 2.0 spec rc.4, requiring Prometheus 3.8.0 or later as the receiving endpoint. (#44861)
The upstream Prometheus library updated the Remote Write 2.0 protocol from rc.3 to rc.4 in
prometheus/prometheus#17411. This renamed CreatedTimestamp to StartTimestamp and moved it
from the TimeSeries message to individual Sample and Histogram messages. This is a
wire-protocol incompatibility, so if you are exporting to a Prometheus server, you must upgrade
it to version 3.8.0 or later to receive data from this exporter.
(Contrib) extension/googlecloudlogentry_encoding: Parse External Application Load Balancer logs into log record attributes instead of placing it in the body as is. (#44438)
(Contrib) pkg/stanza: Allow max_batch_size of 0 for unlimited batching in recombine operator (#43982)
The recombine operator now supports setting max_batch_size: 0 to disable batch size limits.
This allows unlimited batching, letting entries be combined based only on max_log_size and matching conditions.
If you have max_batch_size: 0 in your config and want to keep the behavior unchanged, change the configuration to max_batch_size: 1.
(Contrib) processor/cumulativetodelta: Change default max_staleness from 0 (infinite) to 1 hour (#44427)
The processor now defaults to a max_staleness of 1 hour instead of 0 (infinite retention).
This prevents unbounded memory growth in long-running collector instances, especially when tracking metrics with high cardinality or frequently changing attribute values.
To restore the previous behavior of infinite retention, explicitly set max_staleness: 0 in your configuration.
(Contrib) processor/resourcedetection: Promote processor.resourcedetection.propagateerrors feature gate to beta (#44609)
Note: Splunk OTel Collector has disabled this feature gate by default to prevent unexpected crashes due to resource detection errors
(Contrib) processor/resourcedetection: Remove deprecated attributes configuration option (#44610)
(Contrib) receiver/dockerstats: Upgrades default Docker API version to 1.44 to be compatible with recent Docker Engine versions. (#44279)
Users requiring an older Docker API version can set the api_version in the docker stats receiver config. The minimum supported API level is not changed, only default.
(Contrib) receiver/filelog: Move filelog.decompressFingerprint to stable stage (#44570)
(Contrib) receiver/prometheus: Promote the receiver.prometheusreceiver.RemoveStartTimeAdjustment feature gate to stable and remove in-receiver metric start time adjustment in favor of the metricstarttime processor, including disabling the created-metric feature gate. (#44180)
Previously, users could disable the RemoveStartTimeAdjustment feature gate to temporarily keep the legacy start time adjustment behavior in the Prometheus receiver.
With this promotion to stable and bounded registration, that gate can no longer be disabled; the receiver will no longer set StartTime on metrics based on process_start_time_seconds, and users should migrate to the metricstarttime processor for equivalent functionality.
This change also disables the receiver.prometheusreceiver.UseCreatedMetric feature gate, which previously used the <metric>_created series to derive start timestamps for counters, summaries, and histograms when scraping non OpenMetrics protocols.
However, this does not mean that the _created series is always ignored: when using the OpenMetrics 1.0 protocol, Prometheus itself continues to interpret the _created series as the start timestamp, so only the receiver-side handling for other scrape protocols has been removed.
(Contrib) receiver/prometheus: Native histogram scraping and ingestion is now controlled by the scrape configuration option scrape_native_histograms. (#44861)
The feature gate receiver.prometheusreceiver.EnableNativeHistograms is now stable and enabled by default.
Native histograms scraped from Prometheus will automatically be converted to OpenTelemetry exponential histograms.
To enable scraping of native histograms, you must configure scrape_native_histograms: true in your Prometheus
scrape configuration (either globally or per-job). Additionally, the protobuf scrape protocol must be enabled
by setting scrape_protocols to include PrometheusProto.

🚩 Deprecations 🚩

(Splunk) signalfx-agent: The collectd/zookeeper plugin is deprecated. Please use the zookeeper receiver instead. This plugin will be removed by the end of March 2026. (#7016)
(Contrib) processor/k8sattributes: Removes stable k8sattr.fieldExtractConfigRegex.disallow feature gate (#44694)
(Contrib) receiver/kafka: Deprecate default_fetch_size parameter for franz-go client (#43104)
The default_fetch_size parameter is now deprecated for the franz-go Kafka client and will only be used with the legacy Sarama client.
Users should configure max_fetch_size instead when using franz-go.
This deprecation is marked as of v0.142.0.
(Contrib) receiver/kafka: Support configuring a list of topics and exclude_topics; deprecate topic and exclude_topic (#44477)
(Contrib) receiver/prometheus: Deprecate use_start_time_metric and start_time_metric_regex config in favor of the processor metricstarttime (#44180)

💡 Enhancements 💡

(Core) exporter/debug: Add logging of dropped attributes, events, and links counts in detailed verbosity (#14202)
(Contrib) exporter/kafka: Adds server.address attribute to all Kafka exporter metrics. (#44649)
(Contrib) exporter/prometheusremotewrite: Add option to remove service.name, service.instance.id, service.namespace ResourceAttribute from exported metrics (#44567)
(Contrib) exporter/signalfx: Support setting default properties for dimension updates to be set lazily as part of configuration (#44891)
(Contrib) extension/googlecloudlogentry_encoding: Add support for GCP VPC Flow Log fields for MIG (Managed Instance Group) and Google Service logs. (#44220)
Adds support for the following GCP VPC Flow Log fields:
- Add support for gcp.vpc.flow.{source,destination}.google_service.{type,name,connectivity}
- Add support for gcp.vpc.flow.{source,destination}.instance.managed_instance_group.{name,region,zone}
(Contrib) extension/health_check: Added extension.healthcheck.useComponentStatus feature gate to enable v2 component status reporting in healthcheckextension while maintaining backward compatibility by default. (#42256)
(Contrib) pkg/translator/azurelogs: Added support for Activity Logs Recommendation category (#43220)
(Contrib) processor/resourcedetection: Add support for dynamic refresh resource attributes with refresh_interval parameter (#42663)
(Contrib) processor/transform: New Transform Processor function set_semconv_span_name() to overwrite the span name with the semantic conventions for HTTP, RPC, messaging, and database spans. (#43124)
In other cases, the original span.name is unchanged.
The primary use of set_semconv_span_name() is alongside the
Span Metrics Connector
to address high-cardinality issues in span metrics when span.name does not comply with the OTel requirement
that span names be low cardinality.
(Contrib) receiver/azuremonitor: Scrape storage account sub types (#37220)
(Contrib) receiver/googlecloudpubsub: Exponential backoff streaming restarts (#44741)
(Contrib) receiver/kafka: Make session_timeout, heartbeat_interval, max_partition_fetch_size, and max_fetch_wait unconditional in franz-go consumer (#44839)
(Contrib) receiver/kafka: Validate that exclude_topics entries in kafkareceiver config are non-empty. (#44920)
(Contrib) receiver/oracledb: Added independent collection interval config for Oracle top query metrics collection (#44607)
(Contrib) receiver/prometheusremotewrite: Map.PutStr causes excessive memory allocations due to repeated slice expansions (#44612)
(Contrib) receiver/splunk_hec: Support parsing JSON array payloads in Splunk HEC receiver (#43941)
(Contrib) receiver/sshcheck: Promote sshcheck receiver to beta stability (#41573)
(Contrib) pkg/ottl: Accept string trace/span/profile IDs for TraceID(), SpanID(), and ProfileID() in OTTL. (#43429)
This change allows for a more straightforward use of string values to set trace, span, and profile IDs in OTTL.
(Contrib) pkg/stanza: New featuregate filelog.windows.caseInsensitive introduced. It will make glob matching is case-insensitive on Windows. (#40685)
Previously, any include pattern that included some manner of wildcard (* or **) would
be case-sensitive on Windows, but Windows filepaths are by default case-insensitive. This meant
that in a directory with the files a.log and b.LOG, the pattern *.log would previously only
match a.log. With the filelog.windows.caseInsensitive featuregate enabled, it will match both a.log
and b.LOG when on Windows. The behaviour is the same as always on other operating systems, as all other
currently supported platforms for the Collector have case-sensitive filesystems.
(Contrib) processor/k8sattributes: Updates semconv version to v1.37.0 (#44696)
(Contrib) processor/resourcedetection: Update semconv dependency to 1.37.0 which updates the schema url in the data, but no other impact is expected. (#44726)

🧰 Bug fixes 🧰

(Splunk) resourcedetection: The processor.resourcedetection.propagateerrors feature gate setting has been disabled by default to prevent unexpected crashes due to resource detection errors. (#7030)
In upstream release v0.142.0, the processor.resourcedetection.propagateerrors feature gate has been changed from disabled to enabled by default.
This change may lead to unexpected crashes in environments where resource detection errors occur.
To mitigate this issue, we have disabled the propagateerrors setting by default in our distribution.
Users who require strict error propagation can manually enable this feature in their configuration.
(Contrib) processor/cumulativetodelta: Check whether bucket bounds are the same when verifying whether histograms are comparable (#44793)
(Contrib) processor/cumulativetodelta: Fix logic handling ZeroThreshold increases for exponential histograms (#44793)
(Contrib) processor/filter: Fix context initialization for metric/datapoint context (#44813)
(Contrib) processor/k8sattributes: Fix k8sattr.labelsAnnotationsSingular.allow feature gate to affect config default tag names in addition to runtime extraction (#39774)
(Contrib) processor/tail_sampling: Fix a memory leak introduced in 0.141.0 of the tail sampling processor when not blocking on overflow. (#44884)
(Contrib) receiver/googlecloudpubsub: Acknowledge messages at restart (#44706)
Rewrote the control flow loop so the acknowledgment of messages is more reliable. At stream restart, the messages
ackIds are resent immediately without an explicit acknowledgment. Outstanding ackIds are only cleared when the
acknowledgment is sent successfully.
(Contrib) receiver/kafka: Use max_fetch_size instead of default_fetch_size in franz-go client (#43104)
The franz-go Kafka consumer was incorrectly using default_fetch_size (a Sarama-specific setting) instead of max_fetch_size when configuring kgo.FetchMaxBytes.
This fix ensures the correct parameter is used and adds validation to prevent max_fetch_size from being less than min_fetch_size.
The default value for max_fetch_size has been changed from 0 (unlimited) to 1048576 (1 MiB) to maintain backward compatibility with the previous (incorrect) behavior.
(Contrib) receiver/prometheus: Fix HTTP response body leak in target allocator when fetching scrape configs fails (#44921)
The getScrapeConfigsResponse function did not close resp.Body on error paths.
If io.ReadAll or yaml.Unmarshal failed, the response body would leak,
potentially causing HTTP connection exhaustion.
(Contrib) receiver/prometheus: Fixes yaml marshaling of prometheus/common/config.Secret types (#44445)

Docker Image Manifests:

Linux (amd64, arm64, ppc64le) and Windows (2019 amd64, 2022 amd64):

quay.io/signalfx/splunk-otel-collector:0.142.0
digest: sha256:2017e680b48be7f6cfe3b41c912e2f895489e9ae882844176ed020e014b24c40