Last 7 days
1
Features: 0
Changes: 0
Fixes: 1
Deprecations: 0
Datadog Agent
Open-source Datadog Agent releases for metrics, traces, logs, and integrations.
Latest Datadog Agent changelog updates, official release notes, breaking changes, security patches, pricing changes, and developer reactions in one product feed.
Follow this Datadog Agent release-notes page to spot useful features, risky migrations, noisy announcements, and source links before they hit your backlog.
Where Datadog Agent updates are checked
Changes.Watch links back to official changelog and release-note sources so summaries stay easy to verify.
Related changelog channels
Use channels to follow groups of tools around a stack, workflow, or topic.
Datadog Agent updates by type
Rolling windows show how many product updates landed in the last 7, 30, 90, and 365 days, grouped by existing changelog semantics.
1
4
14
52
Latest Datadog Agent changelog updates
Auto‑enable logon duration when infrastructure mode end‑user device is set, still overridable via config.
Fixed Windows Agent install/upgrade where DDAGENTUSER arguments were overridden by stale registry or service account values.
Enable automatic multi‑line log detection by default; can be disabled via logs.config.auto_multi_line_detection or the corresponding env var.
Fixed Windows DDAGENTUSER KEEP RIGHTS opt-out loss during Fleet automation upgrades by preserving the setting across uninstall/reinstall steps.
Default metric ingestion now uses Datadog v3 intake for Datadog destinations, with configurable flags to enable or disable v3 globally or per endpoint.
Added a new reflector‑based Kubernetes event collection path, enabled via the "watch" event collection mode.
Metrics default to the more compact Datadog v3 intake (bandwidth reduction); v2 intake can be re‑enabled via config flags, and the exporter.datadogexporter.metricremappingdisabled flag is removed.
Added more trace collection during SSI installation on Linux hosts.
Agents are now built with Go 1.25.11.
CIS Docker rules now skip evaluation on Kubernetes nodes using non‑Docker runtimes (e.g., containerd, CRI‑O), eliminating false positives on GKE COS and similar platforms.
Updated the embedded Python runtime from 3.13.13 to 3.13.14.
Health Platform API changed: ReportIssue now requires an IssueReport struct and persistence files are version‑2, older files are logged as incompatible and dropped (breaking).
Updated containerd dependencies (v1.7.32, v2.0.9) to mitigate CVE‑2026‑46680.
Upgrade Prometheus library to v0.311.4, addressing CVE‑2026‑42151 and CVE‑2026‑42154.
macOS Agent now installs system‑wide as a LaunchDaemon under a dedicated dd‑agent user; existing per‑user installations must be removed and re‑installed (breaking change for macOS users).
Upgraded github.com/moby/spdystream to v0.5.1 to patch CVE‑2026‑35469, fixing unchecked memory allocations in SPDY/3 frame parsing.
Upgraded OpenTelemetry Go SDK to v1.43.0 to remediate CVE‑2026‑39883 (PATH hijacking on BSD/Solaris).
Added `datadog agent otel` command to install/remove DDOT from OCI packages
Updated embedded Python to 3.13.13 and rebuilt agents with Go 1.25.9
Added AWS‑based API key authentication, vertical pod resizing, a new KSM CRD configuration provider, and extensive dd‑procmgrd enhancements (CLI, gRPC, process ordering) plus automatic SBOM and CWS policy generation.
Fix: Cloud Network Monitoring now correctly resolves NAT'd cluster IPs when Cilium replaces kube-proxy.
Hide MacOS agent GUI by default for per‑user installs
Agents are now built with Go 1.25.8.
Added many new features: dd-procmgrd Rust daemon, CRD listener, logs pipeline failover, memory‑pressure metrics, Private Action Runner integration (including Windows service), Autodiscover template variables, and OTLP container‑tags v2 f...
Updated circl library to remediate CVE‑2026‑1229 and patched an out‑of‑bounds memory read/DoS vulnerability in the Windows kernel driver, adding extra overflow safety checks.
Fix: infra mode tag now correctly added to system.cpu.user on Windows when not in full mode, aligning behavior with Linux checks
Added ACL command obfuscation to APM span tags (security).
Added instance‑id filter for agent checks, a privateactionrunner binary for running actions, and health‑platform reporting of integration failures.
Updated agents to be built with Go 1.25.7.
Released Agent Prelude and Datadog Cluster Agent Prelude version 7.75.3, including updates to core checks.
Upgrade OpenJDK to 11.0.30, which may break TLS‑enabled JMX integrations
Updated agents to be built with Go 1.25.6.
Added many new capabilities: infrastructure mode for end‑user devices, Kafka Data Streams intake, APM edge mode, enhanced secret backend (Kubernetes, Docker, file secrets), system battery health checks, Azure IMDS version config, discove...
Fixed fatal concurrent map access errors in system probe when evaluating SSH monitoring security rules.
Critical Netflow bug in 7.74.0 (missing flushConfig) can cause 2‑30× flush frequency and billing spikes; upgrade to 7.75.1+.
Fixed diskv2 check LVM device tagging to use friendly /dev/mapper names, aligning with Python psutil behavior and preserving existing dashboards.
Disk check now gracefully handles partition loading errors, continuing to collect metrics for successfully loaded partitions.
Released Agent Prelude and Cluster Agent Prelude version 7.73.1 on 2025‑12‑17
Introduced new core checks (raw Kubelet manifests, AWS ECS Managed Instances support) and enabled Remote Agent Management by default, plus many feature enhancements like OTLP exporter helper, Go disk/network checks, and improved APM trac...
Updated Cluster Agent to use the latest schema for the DatadogPodAutoscaler CRD.
Removed unused span trace loader binary from the Heroku build to reduce package size.
Fix GPU metrics bug where gpu.sm_active and related metrics emitted values lower than expected.
Updated the WDK version used by ddprocmon.
Updated embedded Python to 3.13.7 and built the Agent with Go 1.24.9; added new features such as kube distribution tags, extra Kubernetes metrics, Windows OpenTelemetry service, network path support, and fine‑grained secret‑scope controls.
Imported from changelog source; review and generate a concise summary before publishing.
Fix configuration parsing for extended container metric collection in the container core check.
Added eBPF support for the prctl syscall, new Kubernetes metrics (terminating duration, deployment rollout), RUM support in OpenTelemetry mapping, and low‑port binding via AmbientCapabilities in the Agent.
Added new scrubbers for numerous API keys and auth tokens in HTTP headers and YAML files, enhancing secret redaction.
Released Datadog Agent Prelude version 7.70.1 on 2025-09-10, updating core checks and pinning to the datadog‑agent v7.70.1 tag.
Added GPU metadata autodiscovery (enabled by default) with a new `gpu.enabled` flag, renamed GPU metrics, and improved GPU monitoring for Docker containers and Deeplearning4j programs.
Released Agent Prelude (v7.69.4) on 2025‑08‑27
Updated agents to be compiled with Go 1.24.6 and upgraded JMXFetch to 0.50.0, mitigating CVE‑2025‑52999.
Upgraded libxml2 dependency to version 2.14.5
Added pre‑update validation for custom Agent passwords on Windows domain‑joined hosts, preventing failed updates that could leave agents uninstalled.
Core Agent: cilium conntracker enabled by default (requires /host/sys/fs/bpf mount); new JMXFetch JVM options, ephemerals container collection, GPU host tag now default, secret backends (AWS SSM/Secrets, Vault, Azure), Data Streams Kafka...
Upgrade JMXFetch to 0.49.9, reverting a change that broke Java 7 support.
Fixed pre‑install script bug that unintentionally removed integrations during Agent upgrades.
Released Agent Prelude (v7.68.1) with updated Core Checks and security notes
Bumped core runtimes: Python upgraded to 3.12.11, Go to 1.24.4, and introduced unified attribute precedence across span and resource attributes.
Agent binary now compiled with Go 1.23.10.
Upgrade core components: Python 3.12.11, Go 1.23.9, JMXFetch 0.49.7 with YAML 1.2 support, and default log compression switched to ZSTD.
Fix parsing pod list from kubelet when InPlacePodVerticalScaling feature gate is enabled
Incompatible with Kubernetes 1.33+ due to InPlacePodVerticalScaling; upgrade to 7.66.1 for compatibility
Ensure post‑install script always rebuilds fipsmodule.cnf for the Datadog FIPS agent, fixing FIPS‑related issues
Restored compatibility with the deprecated v1 auto‑instrumentation webhook, preventing forced migration to v2.
Added Windows Agent MSI support for .NET APM instrumentation with IIS auto‑configuration, introduced auto multi‑line detection for logs, and a new OpenLineage proxy endpoint; container‑based primary tags are now enabled by default.
Agents are now built with Go 1.23.8.
Fix: Process Agent files are now added to the flare archive instead of showing "no session token provided" error.
Fixed Python package dependency issue that blocked integrations on Windows
Enabled IMDSv2 by default on EC2 hosts, turned on ECS task collection, added GPU tagging and built agents with Go 1.23.6.
Disable the X25519Kyber768Draft00 key exchange to prevent firewall compatibility issues (e.g., AWS Network Firewall, Suricata).
TLS change may prevent Agent communication with backend when using AWS Network Firewalls
Introduces a TLS change that may disrupt Agent communication when used behind AWS Network Firewalls, prompting users to downgrade to v7.61 or wait for v7.63.3
Security updates include OpenSSL 3.3.3 fixing CVE-2024-12797 and tighter Windows named-pipe access controls.
TLS change introduced in v7.62.3 can prevent the Agent from communicating with Datadog backend when used behind AWS Network Firewalls.
TLS change may block Agent communication through AWS Network Firewalls; downgrade to v7.61 or upgrade to v7.63.3 when available.
TLS change may block Agent communication through AWS Network Firewalls; recommended downgrade to v7.61 or upgrade to v7.63.3 when available.
Added TLS metric tags, new system.cpu.* metrics, external data resolution for APM, and a Logs Agent "analyze" subcommand; extended telemetry with histogram +Inf bucket, p75/p95/p99 values, HTTP zstd compression, and Go 1.23.3 build.
Agents rebuilt with Go 1.21.11 and container images (agent and cluster agent) switched to Ubuntu 24.04 (noble) base.
Imported from changelog source; review and generate a concise summary before publishing.
Updated golang.org/x/crypto to address CVE-2024-45337
Introduced several new features: Aurora cluster DB monitoring autodiscovery, ECS orchestrator check, OTLP top‑level span computation flag, host‑benchmarks enabled by default, network‑policy collection, and default credit‑card obfuscation...
Enabled peer tags aggregation and compute‑stats‑by‑span‑kind by default, adding trace metrics for inferred services and a new check‑tag‑cardinality option.
Added a dedicated temporary directory for JMXFetch, reusing the Agent's temp location
Added ability to run process/container collection on the core Agent (Linux) and introduced new sqllexer flag, GPU vendor tag, and file log collection from rootless Podman containers.
Reverted the cloud‑provided hostname default for the Agent in AKS (added in 7.56.0).
Removed noisy log statement from Network Path logs.
Network Path timeout semantics changed to per‑hop with a default of 1000 ms and traceroute now uses UDP port 33434, affecting existing behavior.
Updated agents to be compiled with Go 1.22.7.
APM trace agent now logs UDS listener creation errors instead of crashing
Added JSON output for `agent diagnose`, device metadata integration, configurable securityContext for auto‑instrumentation, kube runtime class tags, host‑tag API for Python checks, static Kubernetes events allowlist, reverse‑DNS enrichme...
Release Agent and Cluster Agent version 7.56.2
Imported from changelog source; review and generate a concise summary before publishing.
Add support for multiple configuration files (‑E flag) and new config options for Kubernetes events, Docker/Kubernetes event bundling, and source detection.
Agents are now built with Go 1.21.12.
Fixed CVE‑2024‑6257 vulnerability in core checks
Fixed regression that prevented the agent from starting on SysVinit systems.
Added multiple new features: Data Jobs Monitoring flag, configurable Lambda Logs API buffering, Linux‑only service‑discovery check, Kubernetes namespace‑annotation tagging, experimental SBOM on Windows, and OTLP span conversion.
Released Datadog Agent Prelude version 7.54.1 on 2024-06-18.
Added many new features: Probabilistic sampler for APM, Windows CWS write/delete/chdir/mount/umount events, DogStatsD pipeline autoadjust strategies, OTLP ingestion aligned with v0.97.0, Cisco SD‑WAN integration, Podman SQLite backend, a...
Added Aurora cluster DB monitoring autodiscovery, new ECS core check, and OTLP top‑level span flag for richer APM data.
Added a Windows installer check to verify the caller's membership before installing the Agent.
Added security and tracing features: Windows agent now supports CWS process and network threat detection, file‑integrity monitoring, chdir event for container escape, data scrubbing, and span‑links in the trace agent; APM stats include G...
Added chdir event type to CWS in Core Checks.
Orchestrator check moves from Process Agent to Core Agent (configuration migration needed) and Python is upgraded from 3.9 to 3.11; Process Agent pod check will be deprecated in a future release.
Fix incorrect system probe metadata sent to Inventory and Fleet Automation products.
Updated agents to be built with Go 1.20.12.
Fix bug introduced in 7.50.0 that prevented DD_TAGS from being added to kubernetes_state metrics
Migrated win32 event log check from Python to Go (breaking regex changes, legacy mode fallback) and added resume‑position handling.
Add arch field to agent context in CWS events.
Added numerous new checks and metrics (kubelet stats.summary, image pull metrics, GPU metrics, containerd page fault counters, Windows crash detection, Aerospike on ARM, Oracle support, CWS enhancements) and expanded APM tagging and inst...
Upgrade Python to 3.9.18.
Added numerous new core checks and language detection capabilities (e.g., kubernetes pod tolerations, kubelet healthz, process language detection, Oracle custom queries, OTLP config options, and new telemetry metrics).
Fixed NPM driver restart failure on Windows causing a "File Not Found" error
Added ability to send an agent flare via the Datadog app, a remote process metadata collector, and new APM Redis argument obfuscation with configurable environment variables.
Imported from changelog source; review and generate a concise summary before publishing.
Bumped ncurses to 6.4 in the embedded agent, fixing CVE‑2023‑29491.
Added extensive new observability features: topology data collection via CDP, Confluent Kafka support, XCCDF benchmarks for CSPM, custom resource definition metrics, OpenTelemetry span links, semconv 1.17.0, and histogram min/max support...
Agents rebuilt using Go 1.19.8
Added numerous new features: HTTP/2 parsing for Universal Service Monitoring, OTLP probabilistic sampling, .NET APM auto‑instrumentation, CWS SBOM generation, container image metadata collection, Kafka parsing, extra Kubernetes and Farga...
Upgraded JMXFetch to version 0.47.8
Set DD_TRACE_MANAGED_SERVICES default to true
Release 7.43.1 (2023‑03‑07) updates agents to be built with Go 1.19.6.
Added SNMP reachable/unreachable metrics, container image long‑running check, SBOM collection, and default orchestrator manifest collection; updated Windows Agent Manager CLI to use double‑dash arguments (breaking for custom scripts).
Prelude release dated 2023-02-16.
Prelude release version 7.42.1 published on 2023‑02‑02.
Added multiple new integration capabilities, including kube api version tagging, cluster tagger support, OTLP APM compatibility, AWS Lambda telemetry API metrics, process context extraction, and custom resource collection for the orchest...
Agents are now built using Go 1.18.9
Default TLS upgraded to 1.2 and the force_tls_12 flag removed; users must set min_tls_version to use older TLS versions (breaking/security).
Updated agents to be built with Go 1.18.8.
Agent now aborts start if hostname cannot be resolved, causing start failures in containers and requiring Linux kernel >= 4.14 for Universal Service Monitoring (breaking change).
Bug fix: Cloud Workload Security now defaults Traffic Control filter priority to 10.
Updated OPA to v0.43.1 fixing CVE‑2022‑36085
Imported from changelog source; review and generate a concise summary before publishing.
Fixes bug where containerized agents (docker images 7.38.x) spawned zombie (defunct) processes when launched without hostPID:true
Imported from changelog source; review and generate a concise summary before publishing.
Imported from changelog source; review and generate a concise summary before publishing.
Added NetFlow collection, CWS thread‑filtering, CI Visibility payload support, and extended OTLP ingest (exponential histograms, debug log level) plus new Kubernetes and journald metrics.
Cluster Agent followers now forward queries directly to leaders, reducing connections and balancing load
Add multiple Admission Controller injection modes and failure‑policy configuration; enable internal profiling and a new KSM cronjob completion check.
Imported from changelog source; review and generate a concise summary before publishing.
Breaking change: removed deprecated OTLP environment variables and config fields; use new otlp config and DD_OTLP_CONFIG_RECEIVER_* variables.
Fix proxy configuration being ignored by the trace agent
Added IBM Cloud detection and host alias collection; introduced Helm check event collection, service check, and kube QoS tags; made OTLP ingest endpoint GA and rotated package signing keys.
Fix regression impacting CSPM metering
Collect ResourceRequirements from Deployments, StatefulSets, ReplicaSets, and Daemonsets
Fixed weak dependency on the datadog signing keys package for agent, iot agent, and dogstatsd DEB packages.
• Breaking change: APM head‑based sampling default now sends 0 % of traces when the max‑traces‑per‑second setting is 0; OTLP ingest endpoint is stable and old experimental OTLP env vars are removed.
Add external metrics provider.endpoints parameter with fallback and file‑based endpoint checks
Introduced many new capabilities including Podman autodiscovery, Oracle Cloud detection, APM proxy endpoint, SECL rule support for process ancestors, runtime reload command, default process discovery check, mmap/mprotect/ptrace monitorin...
Fix occasional panic when handling tags for deleted containers or pods.
Added PVC tag collection for pending pods and check‑name filtering in cluster checks output
Added several APM enhancements: credit‑card obfuscation (off‑by‑default), configurable error/rare samplers, and a new compact trace intake endpoint.
Removed Log4j dependency from JMXFetch
Updated log4j to version 2.12.2 in JMXFetch.
Set Dlog4j2.formatMsgNoLookups=True for JMXfetch startup.
Reduced excessive ListTagsForResource calls to prevent ECS API throttling.
Add collection of RBAC resources: ClusterRole, ClusterRoleBinding, Role, RoleBinding, ServiceAccount.
Added many new features: Kubernetes priority class tags, autodiscovery variable resolution, container environment tags, Windows support for containerd, CSPM Rego checks, multi‑line log detection, experimental OTLP metrics, a lightweight ...
Pinned to Datadog Agent v7.31.1
Fixed CSPM intake protocol transmission issue
Enable StatefulSet collection by default in the orchestrator check
Adds DaemonSet and StatefulSet collection to the orchestrator check, and enables DaemonSet collection by default.
Added collection of DaemonSet resources for the orchestrator explorer and introduced the `external.metrics.datadog` metric to monitor DatadogMetric validity.
Added numerous new configuration options and endpoints (hostname file, auto‑exit, dogstatsd capture destination, zstd compression, API server TLS params, Cloud Foundry auto‑tagging, runtime security self‑test, container exclusion age, PO...
Windows-only release fixing bugs.
Prelude release dated August 20, 2021.
Added APM internal profiling, experimental OpenTelemetry OTLP support, SELinux event collection, and new Kubernetes autodiscovery variables and tags.
Added correct computation, aggregation, and tagging for Fargate stats and traces in APM
Pinned Cluster Agent to Datadog Agent v7.29.0.
Added support for extra host tags, host aliases, network device metadata, and a new time‑comparison operator in SECL rules, plus APM obfuscator enhancements (Postgres dollar‑quoted strings and optional SQL alias handling).
Provides the same feature set as version 7.28.1
Prelude release announced on 2021-05-31
Provides the same features as version 7.28.0
Switched Agent Docker base image to Ubuntu 20.10, added Windows Core images and ARM64 support for runtime security, plus new config options (OID batch size, ECS tag colon replace, sampler bucket expiry).
Provides the same features as version 7.27.1
Windows‑only release (MSI and Chocolatey installers).
Added templating for kube cluster name and any environment variable as tags in DatadogMetric queries; cluster agent now removes ".." folders from /conf.d mounts when copying configs.
Provides the same feature set as version 7.27.0
Adds extensive platform support: APM on AIX/ppc64, Kubernetes namespace label tags, SNMP corecheck in Go, Windows NPM, new disk read/write metrics, forwarder disk storage, and synthetic Kubernetes state metrics.
Aligns with the feature set of 7.26.0
Added extensive APM, agent autodiscovery, gRPC tagger, log streaming, SQL obfuscation, Prometheus autodiscovery, and ARM64 network tracer support.
Version 6.25.1 includes the same feature set as 7.25.1.
Fixed fatal error from concurrent map reads/writes in inventories payload JSON marshaling.
Mirrors all features from 7.25.0 release
Added numerous APM and integration enhancements such as auto‑discovery labels for AWS Fargate, Windows pipe support, increased tag key length, new obfuscation functions, pause‑container exclusion, and Kubernetes owner‑reference tags.
Datadog Agent 6.24.1 mirrors the feature set of 7.24.1, except for supported Python versions.
Fixed version parsing bug that prevented upgrades from alpha/beta prerelease integration versions.
Added `datadog cluster agent health` command, node‑info collection for the orchestrator explorer, custom orchestrator tags, and AutoscalerReferences field for DatadogMetrics
Provides the same feature set as Datadog Agent 7.24.0 for the 6.x line
Added a new conntrack‑all‑namespaces flag (now enabled by default) and regex‑wildcard support for Kubernetes pod annotations and node‑label tags, plus JMXFetch dedicated logging, UDS support, and env‑var defined dogstatsd mapper profiles.
Reverted the JSON environment variable secret handling introduced in 7.23.0 due to parsing side effects (e.g., whitespace strings becoming lists).
Fixed EC2 metadata tag fetching to respect the ec2_prefer_imdsv2 setting and removed erroneous warning logs
Collect node and cluster resources for Orchestrator Explorer; add resolve option for endpoint checks, health command, and external metrics server configuration.
Aligns 6.23.1 feature set with 7.23.1.
Added numerous APM, network, and logging enhancements including default DNS stats, SQL obfuscation cache, UTF‑16 log support, UDP stats fix, span normalization speed boost, new auto‑discovery tags, GCP project ID tag, IMDSv2 support, and...
Provides all features from 7.22.1 but with a different set of supported Python versions
Added default log file (security agent.log) for the security agent.
Provides the same feature set as version 7.22.0, differing only in supported Python versions.
Added agent‑side compliance rule evaluation, SNMP trap ingestion, and a new Runtime Security Agent with file integrity monitoring (Linux only)
Provides the same feature set as version 7.21.1
Upgrade JMXFetch to 0.38.2 to restore Java 7 compatibility.
Provides the same features as version 7.21.0
Introduced many new capabilities including DatadogMetric CRD autoscaling, compliance agent, 32‑bit ARM IoT Agent, gRPC server, RFC3339 log format, docker/containerd CPU limit metrics, and more.
Release 7.20.2 published on 2020-06-16.
Prelude release (2020-06-11) for version 7.20.1.
Provides the same features as version 7.20.0 but with altered Python version support.
Added multiple‑value pod/container tags, container include/exclude filters, SNMP autodiscovery, custom Docker tags, and a cloud‑provider metadata config option.
Aligns feature set with 7.19.2
Prelude release version 7.19.2 published on 2020‑05‑12.
Fix panic in the DogStatsD standalone package when running in a containerized environment.
Default logs transport switches to HTTPS with compression (TCP fallback remains, configurable via logs config.use_tcp/http)
Default logs transport now defaults to HTTPS with compression (fallback to TCP), which is a breaking change for existing log configurations.
Provides the same features as version 7.18.0
Added top‑level “env” config, new SystemD SUSE integration, and uptime metrics for Docker, CRI and containerd.
Fixes a Windows-only issue where the Agent 7 installer could trigger a system reboot if the C runtime was upgraded during installation.
Includes all features present in version 7.17.1.
Fixed a panic in the log agent when ECS Fargate auto‑discovery reported new containers and the agent couldn't connect to the Docker daemon.
Aligns with 7.17.0 feature set, differing only in supported Python versions.
Updated the agent base image to Debian bullseye and changed hostname handling for containers running in a dedicated UTS namespace (breaking change).
Updated UnixODBC to version 2.3.7.
Imported from changelog source; review and generate a concise summary before publishing.
Datadog Agent v7.16.0 defaults to Python 3 only, introducing a breaking change for custom checks.
Provides all features introduced in 7.16.0
Collect IP address from containers in awsvpc mode.
Added new tagging and API features like persistent volume claim tags for StatefulSets, a Python persistent cache API, tracemalloc whitelist/blacklist, JVM container support, DNS enrichment in network data, and cluster‑level check stats.
Disabled debug log lines for the `hostname` command to prevent noisy output.
Added many new features: Python 3 runtime, container tagging for APM, histogram support, CLC Runner API stats, enhanced systemd check (unit names required, private socket), and numerous new metrics.
Added Unix Domain Socket support for APM receiver, environment‑variable interpolation in check configs, new systemd integration, and ability to set global tags in Fargate via DD_TAGS.
Windows‑only 6.12.2 release aggregating all changes from 6.12.0 and 6.12.1
Fixed bug in kubelet and Fargate integrations that prevented collection of `kubernetes.cpu.*` and `kubernetes.memory.*` metrics.
Breaking change: Kubernetes log collection now prioritizes the Docker socket; use the logs config flag to force file-based collection.
Upgrade JMXFetch to version 0.27.1.
Add option to send the OS hostname as an alias when using the BOSH agent on Cloud Foundry.
Prioritized Kubernetes log collection to try '/var/log/pods' first and fall back to the Docker socket if needed
Added several new features: Secrets beta on Windows, JSON logging support, container thread count metrics, CRI‑O log collection, and AutoDiscovery of unready pods.
Fixes a race condition in Autodiscovery that prevented proper unscheduling of checks on container exit.
Allowed mixed-case service and operation names in APM config by treating them case-insensitively.
Security: Agent now defaults to yaml.safe_load and yaml.safe_dump for all checks, mitigating CVE‑2017‑18342 and fixing unsafe usage in the kubernetes client library.
Added configurable tag cardinality for checks and DogStatsD and new environment variables for extra listeners and config providers.
Fixed a bug that could duplicate the last log entry when a container stopped emitting new logs.
Fixed panic that could occur when a container stopped while the agent was reading its logs.
Fixed the default kubelet check configuration introduced in 6.8.0.
Introduced eBPF network tracer, docker config provider, Kerberos support, healthcheck endpoint, and added numerous metrics and Windows enhancements.
Simplified Trace Search configuration in the trace agent.
Introduced new checks and integrations such as disk check for the puppy agent, cluster‑level checks via the clusterchecks provider, a CRI core check for container resource metrics, and added SysVinit support on Debian.
Fixed a crash in the logs package that could occur when Docker tailer initialization failed.
Fix possible deadlocks when new Docker sources/services are pushed, covering cases where the Docker socket is closed at agent setup, not mounted, or the Kubernetes integration is enabled.
Critical bug causing agent deadlock on Linux/Docker containers leads to removal of packages for those platforms; users are advised to downgrade to 6.4.2.
flare command no longer captures the agent container's environment variables
Added configurable payload exclusion, new Docker and Kubernetes metrics options, SOCKS5 proxy support for logs, and other feature enhancements.
Added new `system.mem.buffered` metric for Linux systems.
Service mapper now groups pod-to-service mappings by namespace to prevent incorrect kube_service tags across namespaces.
Upgrade JMXFetch to 0.20.1 with tagging bugfixes
Introduces numerous new features: Docker memory soft limit metric, host tags for Docker Swarm node role, multi‑URL import command, read buffer size option for DogStatsD, histogram‑to‑distribution flags (beta), encrypted password fetching...
Fixed host tag collection from GCP projects and instance attributes.
Added new metrics and capabilities such as a Docker CPU shares gauge, memory commit metrics, protobuf C++ implementation, TLS support for the Cluster Agent API, and expanded Kubernetes integration (namespace configuration, kubelet filter...
Fixed unauthorized local access to the /agent/check config endpoint
Fixed Docker network tag incorrectly appearing on non-network metrics and autodiscovery tags.
Fixed edge cases where flare could inadvertently expose secrets when they were wrapped in quotes
Fixed Docker size‑check crash for images without repository tags;
Added numerous platform and integration features: agent version in flare, DD_CHECK_RUNNERS env var, status command, configurable process/logs/APM enablement, JMX cgroup memory limit, Kubernetes pod annotation tags, bind host option, YAML...
Fixed issue preventing Docker tags when collecting container logs.
Restores proper download packages for Docker, Windows, Linux, and macOS (DMG & MSI).
Add Log Agent check information to the GUI General Status page.
Agent 6.0 is a full rewrite in Go, adding higher DogstatsD throughput, check concurrency, and a native Prometheus metrics ingest.