This Week’s Cloud Native Pulse: Dec 13-19 – OTel Memory Leak Fix, K8s 1.35 GA Blitz, ArgoCD Shields Up

Last week was packed with important releases across the tools many of us rely on daily: OpenTelemetry, Kubernetes, ArgoCD, ArgoCD Image Updater, Prometheus, and Grafana. This post highlights the changes that are most likely to impact your clusters, dashboards, and pipelines, with direct links to deeper release notes on https://www.relnx.io/

OpenTelemetry Collector Contrib v0.142.0

OpenTelemetry Collector Contrib v0.142.0 was released on December 17, 2025, and it comes with a mix of critical fixes and useful quality‑of‑life improvements for production pipelines. This is a release worth prioritizing if you use tail sampling, Prometheus Remote Write, GCP networking, or Datadog integrations.

Key highlights:

1. Tail sampling memory leak fix
   A critical memory leak introduced in 0.141.0 for the tail sampling processor (when not blocking on overflow) has been fixed, which is essential if you rely on tail sampling for high-volume traces.
   Details: fix tail sampling memory leak. ​https://www.relnx.io/features/fix-a-memory-leak-introduced-in-01410-of-the-tail-sampling-processor-when-not-blocking-on-overflow-1450

2. Remote Write 2.0 rc.4 (breaking change)
   The collector now targets Remote Write 2.0 spec rc.4, which requires Prometheus 3.8.0 or later, so environments using Prometheus Remote Write must ensure compatibility before upgrading.

Details: Remote Write 2.0 spec rc.4 change https://www.relnx.io/features/updated-to-remote-write-20-spec-rc4-requiring-prometheus-380-or-later-the-upstream-prometheus-library-updated-the-remote-write-20-protocol-from-rc3-to-rc4-in-prometheusprometheus17411-1475

1. filelog.decompressFingerprint is now stable The filelog.decompressFingerprint feature for identifying and decompressing log files has graduated to stable, improving confidence in processing compressed logs at scale for better storage and transfer efficiency

Details:https://www.relnx.io/features/move-filelogdecompressfingerprint-to-stable-stage-1472

1. Better GCP External HTTP(S) LB logs
   External Application Load Balancer logs can now be parsed into log record attributes instead of being left as raw body payloads, increasing readability and query power for GCP users.

2. Simplified cache lifecycle management
   Cache lifecycle handling has been simplified by removing unnecessary WaitGroup complexity, which reduces internal complexity and the chances of subtle lifecycle bugs.

Details: https://www.relnx.io/features/simplified-cache-lifecycle-management-by-removing-unnecessary-waitgroup-complexity-1457

1. Datadog receiver: multi-tag parsing flag A new receiver.datadogreceiver.EnableMultiTagParsing feature gate controls how Datadog tags are converted into OpenTelemetry attributes, giving more precise control over tag-to-attribute mapping.

Details: https://www.relnx.io/features/add-receiverdatadogreceiverenablemultitagparsing-feature-gate-the-feature-flag-changes-the-logic-that-converts-datadog-tags-to-opentelemetry-attributes-1438

1. Datadog receiver: AWS SDK semantic conventions The Datadog receiver improves compliance with OpenTelemetry Semantic Conventions for AWS SDK spans, bringing more consistent, interoperable tracing data across services using the AWS SDK

Details: https://www.relnx.io/features/improve-the-compliance-with-otel-semantic-conventions-for-aws-sdk-spans-in-the-datadog-receiver-compliance-improvements-on-spans-received-via-the-datadog-receiver-when-applicable-1436

1. Datadog tag runtime remapped The Datadog runtime tag now maps to container.runtime.name instead of container.runtime, aligning better with OpenTelemetry attribute naming and improving trace and metric consistency.

Details: https://www.relnx.io/features/the-datadog-tag-runtime-is-now-mapped-to-the-otel-attribute-containerruntimename-instead-of-containerruntime-1435

1. New transform: set_semconv_span_name()
   A new transform processor function, set_semconv_span_name(), can rewrite span names according to semantic conventions for HTTP, RPC, messaging, and database spans, helping tackle high-cardinality span names

2. GCP VPC Flow Logs: MIG & Google Service fields
   Support was added for GCP VPC Flow Log fields for Managed Instance Groups and Google Service logs, enabling more granular visibility and troubleshooting for GCP network traffic.

Everything else in this release: https://www.relnx.io/releases/opentelemetry-collector-contrib-v0-142-0

Kubernetes v1.35.0

Kubernetes v1.35.0 contains several observability, metrics, and UX changes, along with some deprecations and GA features that may affect day‑to‑day operations. This is a good release to review from both SRE and platform governance perspectives.

Highlights:

1. Improved kube-proxy /statusz The /statusz page for kube-proxy now includes a list of exposed endpoints, making debugging and introspection of network behavior easier.

Details: https://www.relnx.io/features/updated-the-statusz-page-for-kube-proxy-to-include-a-list-of-exposed-endpoints-making-debugging-and-introspection-easier-1699

1. Deprecated metrics hidden by policy Deprecated metrics are now hidden according to the metrics deprecation policy, helping teams avoid relying on outdated signals while keeping their metric surface area clean.

Details: https://www.relnx.io/features/deprecated-metrics-will-be-hidden-as-per-the-metrics-deprecation-policy-httpskubernetesiodocsreferenceusing-apideprecation-policydeprecating-a-metric-1597

1. Excluded dry-run requests from apiserver_request_sli_duration_seconds Dry‑run requests are excluded from this SLI metric, ensuring latency measurements better reflect real user-impacting operations.

Details: https://www.relnx.io/features/metrics-excluded-dryrun-requests-from-apiserver-request-sli-duration-seconds-1570

1. New kubelet metrics for secret-pulled images

New kubelet metrics for the “Ensure Secret Pulled Images” KEP provide visibility into pulling images from private registries with secrets, improving troubleshooting of image pull performance.

Details: https://www.relnx.io/features/introduced-new-kubelet-metrics-for-the-ensure-secret-pulled-images-kep-including-1557

1. Metrics for StatefulSet MaxUnavailable New metrics expose how many pods can be unavailable during a StatefulSet update, which helps control and reason about downtime during rolling updates

Details: https://www.relnx.io/features/added-metrics-for-the-maxunavailable-feature-in-statefulset-1535

1. More events during Pod resizing Additional events are emitted during pod resizing, providing clearer visibility into resize status changes and helping debug vertical scaling operations.

Details: https://www.relnx.io/features/added-additional-event-emissions-during-pod-resizing-to-provide-clearer-visibility-when-a-pods-resize-status-changes-1533

1. New kubelet image manager metric
   The kubelet_image_manager_ensure_image_requests_total{present_locally, pull_policy, pull_required} counter exposes detailed information on how often kubelet must ensure images are present, which can inform image placement strategies.

2. In‑place Pod resource updates are GA
   In‑place updates of Pod CPU and memory resources have graduated to GA, enabling nondisruptive vertical scaling for many workloads that previously required recreating pods.

3. HPA performance improvement for container metrics
   Container-specific HPA metrics now use an optimized lookup that exits early when the target container is found, reducing overhead in pods with multiple containers.

4. Dropped certificates/v1beta1 CSR support in kubectl
   kubectl no longer supports certificates/v1beta1 CertificateSigningRequest, nudging users to use stable API versions.

5. Stricter kubectl exec syntax
   kubectl exec [POD] [COMMAND] is no longer supported; kubectl exec [POD] -- [COMMAND] is now required, which aligns with long‑established best practices and avoids parsing ambiguities

Details: https://www.relnx.io/features/changed-kubectl-exec-syntax-to-require-before-the-command-the-form-kubectl-exec-pod-command-is-no-longer-supported-use-kubectl-exec-pod-command-instead-1594

1. UserNamespacesPodSecurityStandards gate removed The UserNamespacesPodSecurityStandards feature gate has been removed now that the minimum supported kubelet version is v1.31, making the enhanced pod security behavior default and reducing configuration complexity.

Details: https://www.relnx.io/features/removed-the-usernamespacespodsecuritystandards-feature-gate-the-minimum-supported-kubernetes-version-for-kubelet-is-now-v131-so-the-gate-is-no-longer-needed-1687

Full Kubernetes v1.35.0 release highlights are available on: https://www.relnx.io/releases/kubernetes-v1-35-0

ArgoCD v3.2.2

ArgoCD v3.2.2, released on December 18, 2025, is a smaller but meaningful bug‑fix release targeting authentication, secret management, and ApplicationSet behavior.​

Key fixes:

1. AuthMiddleware: check userinfo endpoint
   The AuthMiddleware now checks the userinfo endpoint, improving validation of authenticated users and strengthening the security model around who can access ArgoCD

2. Read and write secrets for the same URL
   Support for separate read and write secrets on the same URL provides more granular access control, which is useful for tightening permissions around sensitive resources

3. AppSet preserves annotations during hydration
   ApplicationSet now preserves annotations when hydration is requested, ensuring that attached metadata remains intact and usable by downstream tools and automation.

Read the full ArgoCD 3.2.2 breakdown on: https://www.relnx.io/releases/argocd-v3-2-2

Argocd-image-updater V1.0.2

ArgoCD Image Updater v1.0.2, released on December 16, 2025, focuses on making deployments more predictable and reducing surprise behavior around tags and annotations.​

Highlights:

1. Installed into argocd namespace by default
   Installing the Image Updater into the argocd namespace by default simplifies setup and improves integration between the controller and ArgoCD itself.​

2. Preserve existing Helm tag parameter when image has no tag
   When an image has no explicit tag, the existing Helm tag parameter is preserved, reducing the risk of unintentionally changing image versions during updates.

3. Fix infinite commit loop with digest strategy
   A bug where digest strategy inconsistently wrote tag names and caused infinite commit loops has been fixed, eliminating noisy commits and wasted CI/CD cycles.​

4. Default argocd-image-updater-controller annotation

Using argocd-image-updater-controller as a default container annotation makes automatic image management simpler and helps keep workloads on up‑to‑date images with less manual effort.

More details are available in the full ArgoCD Image Updater v1.0.2 notes on https://www.relnx.io/releases/argocd-image-updater-v1-0-2

Prometheus v3.8.1

Prometheus v3.8.1, released on December 16, 2025, is a focused bug‑fix release that is especially relevant if you rely on Remote Write.

Highlights:

1. Remote Write receiver bug fix The Remote Write receiver now avoids sending incorrect response headers for the v1 flow, which previously caused senders to emit false partial error logs and metrics, improving the accuracy and trustworthiness of your monitoring data.​

Full Prometheus 3.8.1 release summary is available on https://www.relnx.io/releases/prometheus-v3-8-1

---

Grafana v12.3.1

Grafana v12.3.1, released on December 17, 2025, is a UI and UX‑focused update that cleans up dashboard behavior and improves Azure log exploration.

Highlights:

1.Fixed empty space under time controls
Dashboards with many variables no longer show a large empty space under the time controls, giving back valuable screen real estate for panels and visualizations.​

1. Clearing hideSeriesFrom on query edit
   The QueryEditorRows behavior now clears hideSeriesFrom overrides when a query is edited, helping prevent accidental hiding of relevant series after query changes.​

2. Azure logs: aggregate columns in logs builder
   Azure users can now include aggregate columns directly in the logs builder, making it easier to derive and visualize higher-level metrics from log data.​

More Grafana 12.3.1 details can be found on https://www.relnx.io/releases/grafana-v12-3-1

---

That wraps up a busy week across OpenTelemetry, Kubernetes, ArgoCD, Prometheus, and Grafana. If you want to keep up with these changes and benefit from automated upgrade guidance, join the community at relnx.io, where you can track releases for your favorite tools and explore auto‑upgrade workflows tailored to your stack.