Skip to content

Changelog

2.5.0(2026/07/08)

This release is an iterative release, with the following main updates:

New Features

  • Oracle collector added ASM diskgroup metrics (#3136)
  • Oracle collector added database open mode, resource limits, and datafile limit metrics (#3136)
  • Flameshot added Go language pprof collection support (#3132)
  • Flameshot added proactive Heap Dump and hprof upload to OSS/S3 (#3122)

Bug Fixes

  • Fixed gpu_smi infinite loop risk in GPU drop warning logic (#3144)
  • Fixed aggregation tail sampling not grouping by sinker header (#3095)

Improvements

  • Redis added role_status metric field for master/replica role change detection (#3141)
  • Added global measurement_version config for unified measurement version during install/upgrade, covering MySQL, Oracle, PostgreSQL, SQLServer, Redis, JVM collectors (#3142)
  • Added /v1/health endpoint and CRI runtime auto-recovery for better K8s liveness detection (#3140)

2.4.0(2026/07/02)

This release is an iterative release, with the following main updates:

New Features

  • DataKit now supports adaptive IPv4/IPv6 dual-stack connections to DataWay, preferring IPv6 by default and automatically falling back to IPv4, with configurable IP family policies (#3131)
  • Pipeline added the enable_grok_fast_path option to explicitly control the Grok fast path optimization, which is disabled by default (#3133)
  • SNMP added the user Profile directory conf.d/snmp/extra_profiles, allowing user Profiles to override or extend built-in Profiles without being overwritten during upgrades (#3128)

Bug Fixes

  • Fixed gpu_smi compatibility with NVIDIA SMI v12/v13 XML, including newer power fields and additional GPU metrics (#3129)

Improvements

  • SNMP built-in Profile metrics now support tags_ignore and tags_ignore_regexp, reducing unnecessary time series caused by high-cardinality tags (#3130)
  • Corrected types, units, and descriptions for MySQL, MySQL user status, DBM, and InnoDB metrics to distinguish cumulative counters from current-state values (#3135)

Compatibility Adjustments

  • The OpenTelemetry collector now supports newer LoongSuite/Alibaba ARMS semantic fields, adding DB, HTTP/Network, messaging, and RPC attribute mappings (#3085)

2.3.0(2026/06/24)

This release is an iterative release, with the following main updates:

New Features

  • Added GKE Autopilot integration. DataKit can now be deployed as a Deployment through the datakit-gke-autopilot Helm Chart/YAML and collect container metrics plus stdout/stderr logs through GCP Cloud APIs (#3107)

Bug Fixes

  • Fixed PostgreSQL table object collection compatibility across versions and duplicate index data during batched collection (#3127)

Improvements

  • Optimized SNMP collection scheduling: duplicate jobs of the same type for the same IP are skipped, while different collection types remain serialized, reducing duplicate collection and wait overhead (#3121)
  • Improved communication efficiency between eBPF and DataKit Operator, reducing query and parsing overhead in large clusters. This requires DataKit Operator v1.8.9 or later (#3125)
  • Added the new Kafka Dashboard and Monitor entries to the Kafka integration (#3040)
  • Upgraded DCA to 0.1.7, improving management UI stability and interaction experience, with additional frontend and backend unit test coverage (#3065)
  • Improved measurement metadata: measurements-meta.json now includes desc_i18n, and units, tags, and field descriptions have been completed for multiple collectors (!3956)

Compatibility Adjustments

  • DataKit metric points now include a __input_source tag with values like dk.<input> to identify the source collector. Metrics generated through HTTP API and Pipeline also include the corresponding source (#3124)

2.2.1(2026/06/18)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed the PostgreSQL collector excluding the default postgres database by default, which prevented metrics for that database from being collected. To exclude it, configure ignored_databases explicitly (#3123)

2.2.0(2026/06/17)

This release is an iterative release, with the following updates:

New Features

  • Added IBM AS/400 (IBM i) external collector, collecting system, disk, job, memory pool, subsystem, job queue, and message queue metrics via ODBC (#3082)
  • vSphere collector added VM-level disk storage metrics disk_used_latest, disk_provisioned_latest, disk_unshared_latest (#3111)
  • Dialtesting collector added SSL/TLS certificate check task, supporting certificate expiration, TLS version detection, etc. (#3106)
  • Pipeline added json_all and pt_kvs_set_map functions (#3109)
  • SNMP collector added oid_batch_size and bulk_max_repetitions configs, for SNMP agents sensitive to GetBulk requests like iDRAC (#3093)
  • PodMonitor/ServiceMonitor switched to informer architecture; YAML modifications take effect dynamically without restarting collection (#2832)

Bug Fixes

  • Fixed APM auto-injection failure in certain scenarios (#3120)
  • Fixed ICMP dial test misidentifying 0ns response as packet loss on low-latency Windows environments (#3118)
  • Fixed PostgreSQL 9.1+ replication delay metrics not being reported due to numeric type conversion (#3114)
  • Fixed prom_remote_write collector continuing to process invalid data due to missing return after parse failure (#3105)
  • Fixed PayloadType field inconsistency after compact Body cache dump/load (#3102)
  • Fixed diskio collector unit test occasional doubled read/write rate (#3101)
  • Fixed disk usage calculation anomaly under certain conditions (#3090)
  • Fixed HTTP API reload losing rate limit and timeout configs; hot-reloaded config now matches initial startup (#3079)

Improvements

  • Container log collection now keeps the last config and warns on duplicate paths instead of discarding the task (#3119)
  • Browser dial test results now hide low-level Lightpanda startup errors to avoid exposing runtime paths (#3103)
  • Dial test node name changes are now automatically synced to task reporting data (#3099)
  • Added SNMP custom YAML template documentation (#3116)
  • Updated database integration Dashboard path (#3113)
  • CI added Go module/golangci-lint cache reuse, Docker buildx supports registry cache (#3096)

Compatibility Adjustments

  • DK self-metric collection changed to all-or-nothing mode, whitelist filtering removed; Profile collection changed to manual control (#3110)
  • ddtrace collector telemetry tag now compatible with DD_TRACE_TAGS field name, correctly writing to JVM metric tags (#3108)

2.1.5(2026/06/16)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed a case where logs could be lost for short-lived Pods (#3117)

2.1.4(2026/06/12)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed an issue where log collection could lose data when a Pod shuts down (#3115)

2.1.3(2026/06/12)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed time series expansion caused by collector_source_ip auto-injection in metrics; OpenTelemetry metrics now support disabling global tags via configuration (#3112)

Improvements

  • Adjusted the default behavior of datakit debug --bug-report: it now only generates and keeps the local zip file by default instead of uploading automatically. Use --bug-report-dataway explicitly for Dataway upload, or continue using --oss for OSS upload (#3089)

2.1.2(2026/06/08)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed an issue where the process collector could miss the container_id field in special environments. DataKit now scans all lines in /proc/{pid}/cgroup to extract a valid container ID (#3092)
  • Fixed excessive Prometheus metric series for log file path scan duration when containers are created and deleted frequently, reducing high memory usage and OOM risk (#3097)
  • Fixed a race condition in container log collection that could prevent the Single goroutine from exiting when a container is deleted, avoiding goroutine leaks (#3100)

2.1.1(2026/06/04)

This release is a hotfix release, contents are as follows:

Compatibility Adjustments

  • Browser dial testing now uses the Lightpanda engine by default, with Lightpanda built into the DataKit image (#3098)

2.1.0(2026/06/03)

This release is an iterative release, with the following main updates:

New Features

  • Added browser dial testing tasks. inputs.dialtesting can now pull and execute BROWSER tasks to simulate real browser page access, interactions, and result reporting (#3072)
  • DataKit CLI added the datakit completion command to generate completion scripts for bash, zsh, fish, and PowerShell (#3066)

Improvements

  • Optimized exponential Pipeline performance regression, reducing processing latency for Grok parsing patterns with consecutive GREEDYDATA segments (#3094)

Compatibility Adjustments

2.0.0(2026/05/27)

This release is the first DataKit 2.x mainline release and officially enables the independent datakit-v2 install and upgrade source, with the following main updates:

New Features

  • BPF network logs now support recording HTTP headers, preserving trace-related context in L7 logs (#3081)
  • HTTP API now supports the pull interface and allows filter processing to be disabled for a single data upload (#3076)
  • Prometheus Remote Write collector added the keep_exist_metric_name option to preserve original metric names (#3075)
  • Logfwd now supports configuring from_beginning_threshold_size through environment variables (#3074)
  • DDTrace collector implemented the info interface to improve collector information output (#3073)
  • hostobject added Kingsoft Cloud KEC metadata support (#3068)
  • OceanBase collector added compatibility for 4.x system views (#3067)
  • DataKit can now automatically dump its own profile and upload it to the center for online troubleshooting (#3064)
  • Redis collector added database object collection and reporting (#2984)

Bug Fixes

  • Fixed abnormal HTTP/3 dial testing timing data that caused Download duration to display an invalid long value (#3080)
  • Fixed Kubernetes Prometheus collector errors when the ServiceAccount token file expires (#3071)
  • Fixed the 6060 pprof service not restarting after it exits following hot reload (#3063)
  • Fixed delayed disk and hostobject filtering that could trigger automatic mounting on special mount points and affect object reporting (#3044)
  • Fixed Flameshot jcmd handling and duplicated profile tags (#3062)

Improvements

  • Adjusted DataKit v2 file paths on OSS/CDN to support the major version release flow (#3083)
  • Optimized default traceroute configuration for the dial testing collector (#3078)
  • Adjusted DataKit Sinker header handling so global tags are only carried by point write requests (#3070)
  • Improved profile reporting observability by adding info data and related metrics (#3055)
  • Added support for newer Docker API versions to improve container environment compatibility (#2991)

Compatibility Adjustments

  • DataKit upgraded to major version v2 while keeping a permanent v1 staging branch; upgrade logic now checks whether the current environment supports upgrading to v2 (#3008)
  • DataKit v2 now uses an independent install and upgrade source at https://static.guance.com/datakit-v2/, isolated from the 1.x source https://static.guance.com/datakit/.
  • Automatic upgrades from 1.x will not cross major versions to 2.x; to upgrade from 1.x to 2.x, manually replace datakit with datakit-v2 in the install or upgrade URL.
  • DataKit v2 build toolchain has been upgraded to Go 1.26.2; system requirements are now Linux kernel >= 3.2, Windows Server 2016(included)+, or macOS 12+.

1.94.0(2026/05/13)

This release is an iterative release, with the following main updates:

New Features

  • Socket log collection now records the source IP address using the collector_source_ip tag (#3061)
  • HTTP dial testing now supports custom protocol versions, covering more HTTP compatibility scenarios (#3041)
  • PostgreSQL DBM now reports SQL execution count and QPS metrics (#3046)
  • MongoDB now supports database object collection (#3045)
  • Doris now supports object collection (#3043)
  • Bug report upload now defaults to Dataway-based upload while keeping the existing OSS direct-upload mode (#3028)
  • AWS Lambda collection improved function invocation tracing by correlating runtime events with invocation context (#2961)

Bug Fixes

  • Fixed 403/404 handling for the 9529 HTTP API when routes are not registered or related inputs are not enabled, reducing troubleshooting ambiguity (#3054)
  • Fixed DataKit service status when the HTTP service fails to start; the main process now exits on HTTP startup errors (#3052)
  • Fixed APM auto-injection issues caused by arm64 shared library cross-compilation and replacement failures (#3050)
  • Fixed occasional concurrent map writes in the Redis collector and improved host tag priority handling (#3039)

Improvements

  • StatsD now disables DogStatsD event and service check log collection by default, enabling them only through explicit configuration to avoid extra default log volume (#3059)
  • Improved DK_HTTP_LISTEN parsing so ip:port can be used directly, with clearer precedence when DK_HTTP_PORT is also configured (#3053)
  • Improved Pipeline Grok and JSON processing performance for higher log processing throughput (#3051)
  • Removed CGO dependency from the eBPF collector and optimized memory usage across netlog, netflow, L7 flow, and exporter paths with better runtime observability (#3049)
  • Continued improving automatic multiline log rules to enhance default multiline detection (#3048)
  • Added hot reload support for cat, xfsquota, windowsremote, logfwdserver, and related collectors (#3042)
  • Optimized DataKit startup by removing unnecessary collection work during initialization, reducing startup time (#3038)
  • Refactored the OpenTelemetry collector to reuse shared cliutils/otlp parsers, consolidating metrics/logs/traces parsing loops while preserving DataKit-specific semantics (#3026)

1.94.1(2026/07/08)

This release is a hotfix release, contents are as follows:

Compatibility Adjustments

  • Added global measurement_version config for unified measurement version during install/upgrade, covering MySQL, Oracle, PostgreSQL, SQLServer, Redis, JVM collectors (#3142)

1.93.0(2026/04/22)

This release is an iterative release, with the following main updates:

New Features

  • Switched MySQL and PostgreSQL DBM collectors to a new implementation, improving database performance and object collection flows (#2998)
  • Pipeline scripts now support wildcard matching on source, making rule reuse easier (#3036)
  • Added jcmd and hprof collection support to Flameshot for better Java troubleshooting (#3031)

Bug Fixes

  • Fixed issues in aggregate send and metric accounting paths, restoring compatibility metrics and concurrent sending behavior (#3024)
  • Fixed vSphere event collection time and timeout handling, and added related unit tests (#3037)
  • Fixed several stability issues in the dial testing module and completed unit test coverage for core paths (#3018)

Improvements

  • Optimized multiline log processing strategy, unifying manual and automatic matching behavior while expanding the default rule set (#3029)
  • Refactored the eBPF collector path to use cilium-related capabilities directly, with improved stability and compatibility (#3016)
  • Updated DCA documentation and usage guidance (#3032)

1.92.1(2026/04/16)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed the Pipeline add_key path not serializing list/map values, and now normalizes composite values into strings before writing them (#3035)
  • Fixed SQLServer and Oracle object collection not strictly honoring the configured collection interval, preventing collection from still running before the next window is reached (#3034)
  • Fixed incorrect k8s requests field values in container metric collection, ensuring container resource requests are reported correctly (#3033)
  • Fixed datakit import replay not inheriting DataWay settings from datakit.conf, while still keeping command-line URL overrides available (#3030)

Improvements

  • Adjusted DK external collector and related component build flows, improving multi-architecture build paths and compilation compatibility (#3013)

1.92.0(2026/04/09)

This release is an iterative release, with the following main updates:

New Features

  • Added support for data pre-aggregation, covering aggregation and tail-sampling flows (#2892)
  • Pipeline now supports processing llm data type inputs (#3001)
  • Dial testing now reports SSL certificate validity fields, including certificate expiration time and remaining days (#3003)

Bug Fixes

  • Fixed OpenTelemetry compatibility issue where newer clients require a non-trivial response body, by completing the gRPC response payload (#3017)
  • Fixed DDTrace memory leak by improving large trace recycling logic to avoid OOM (#3012)
  • Fixed goroutine leak in log collection, preventing Tailer shutdown paths from blocking on cross-instance waits (#3010)
  • Fixed wrapped-url-error caused by an unencoded X-Global-Tags-V2 header in datakit sinker v2 (#3009)
  • Fixed NTP time diff not being cleared automatically after system clock recovery (#3006)

Improvements

  • Optimized database_instance priority and DBM object naming for SQLServer and Oracle to avoid cross-node data confusion (#3011)
  • Removed election tag injection from the NewPoint stage across several database collectors to reduce overhead, and improved Oracle slow query obfuscation (#3004)
  • Added a default multiline rule for TiDB slow logs (#3005)
  • Added compatibility for higher-version systemd libraries in Journald (#2996)
  • Refactored GitLab collector Prometheus metric classification, completed metric fields, and unified measurement naming (#2988)

Compatibility Adjustments

  • Removed the HTTP web service from upgrade; upgrade management is now handled through DCA (#3007)

1.91.0(2026/03/26)

This release is an iterative release, with the following main updates:

New Features

  • Kingbase collector added server field configuration support, allowing explicit server identification, defaults to host:port format (#3002)
  • Bug report now collects external collector logs, automatically gathering .log files from [DataKit Install Dir]/externals directory (#2989)
  • SQLServer and Oracle collectors added database_instance dimension, querying database to obtain instance identifier and write as tag (#2999)
  • Monitor command added -Q (--quantile) option, supporting quantile selection from summary metrics (#2968)

Bug Fixes

  • Fixed FireLens log streaming support for nested map/list types, now serializing complex types to JSON strings (#3000)
  • Fixed Kingbase collector singleton mode limitation, now supporting multiple concurrent instances (#2995)
  • Fixed logfwd 1.86.0 configuration compatibility issue, supporting deprecated LOGFWD_JSON_CONFIG environment variable with automatic conversion to new format (#2993)
  • Fixed missing election status metrics in DataKit, ensuring election status is reported even when not elected as leader (#2992)
  • Fixed OpenTelemetry collector parent_span_id handling when zero value, normalizing 0000000000000000 to 0 (#2987)
  • Fixed WAL infinite loop issue caused by malformed HTTP payload during data upload, now identifying and dropping dirty data (#2949)
  • Fixed sinker header value containing invalid characters (e.g., \n), now URL-encoding header values (#2947)

Improvements

  • Improved log collection multiline matching logic, removed deprecated logging_auto_multiline_detection config option, optimized multiline pattern validation flow (#2990)
  • External collectors now support cross-compilation, improving multi-platform build efficiency (#2994)
  • Oracle collector upgraded metrics to v2, supporting grouped collection with configurable intervals by metric type (tablespace/slow_query/process/system) (#2938)

1.90.0(2026/03/11)

This release is an iterative release, with the following main updates:

Breaking Changes

  • Removed legacy completion entry points datakit tool --setup-completer-script, datakit tool --completer-script, and the static script datakit-completer.sh. Use datakit completion instead.

New Features

  • APM injector added PHP application automatic injection support, including PHP interpreter detection, ddtrace extension installation, and configuration management (#2986)
  • Logstreaming input added AWS Firehose data source type support, receiving and processing logs from AWS Firehose HTTP endpoints (#2979)
  • Oracle and SQLServer collectors added DBM (Database Monitoring) functionality, including query metrics, activity monitoring, session aggregation, connection metrics, query object storage, and execution plan storage (#2904)
  • Host installer added collector configuration support during installation, passing collector configs via DK_INPUT_CONFIGS environment variable (#2967)
  • Journald input added external collector implementation (#2974)

Bug Fixes

  • Fixed logfwd storage_index configuration priority error, environment variable LOGFWD_GLOBAL_STORAGE_INDEX now takes priority over CRD configuration (#2985)
  • Fixed Helm chart DataWay token plaintext exposure issue, supporting automatic creation of Kubernetes Secrets to store tokens securely (#2981)
  • Fixed OpenTelemetry metrics missing unit and description fields, now extracting and propagating these fields from OTEL metrics (#2977)

Improvements

  • SNMP object collector exposed device information (device_type, device_vendor, device_hostname) and merged interface entries by interface name (#2978)
  • DataKit installer added collector configuration support during installation (#2967)
  • Updated APM injection documentation to include PHP support (#2986)
  • Other optimizations and bug fixes

1.89.1(2026/02/12)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • Fixed issue in DK 1.89.0 where global host tag setting host=__datakit_hostname did not correctly use k8s node name (#2971)
  • Fixed collector resume failure blocking election heartbeat, avoiding frequent election switching (#2970)
  • Fixed error triggered when accidentally collecting ECSFargate container logs (#2964)
  • Fixed election module state management to ensure metric timestamp updates accurately (#2970)

Improvements

  • Flameshot supports obtaining container resource limit information, optimizing threshold calculation accuracy in container environments (#2966)
  • DataKit supports accessing k8s Pod data through datakit-operator, providing API Server pressure relief solution for large-scale clusters (#2931)

1.89.0(2026/02/04)

This release is an iterative release, with the following main updates:

New Features

  • Added host change detection functionality, supporting user, crontab, service, and file change monitoring (#2917)
  • Flameshot now supports continuous collection mode, with default scheduled collection and threshold-triggered continuous collection (#2953)
  • Added DataKit self log collection configuration functionality (#2950)

Bug Fixes

  • Fixed Prometheus collector tags priority error (#2960)
  • Fixed issue where global host tag setting host=__datakit_ip was ineffective (#2956)
  • Fixed issue where eBPF collector caused `istio-init`` containers to not exit (#2955)
  • Fixed unnecessary operations in container log collection when using default stdout configuration (#2962)
  • Fixed WAL lock file issue using PID that prevented reuse after DataKit exit (#2948)
  • Fixed profile collector initialization timing to avoid panic due to uninitialized disk cache (#2946)
  • Fixed Statsd collector, add event/service-check collection (#2941)1

Improvements

  • Added more logs and metrics to the election module for detecting frequent election switching and collector pause failures (#2957)
  • Updated DataKit HTTP client metrics, adding URL path tags and request body transfer summary metrics (#2952)
  • SQLServer collector added sqlserver_host tag and changed instance tag to counter_instance (#2951)
  • Bug report now collects git configuration files (#2939)
  • Windows process collector added status field support (#2927)
  • DDTrace added more source_type support（#2958）

1.88.1(2026/01/16)

This release is a hotfix release, contents are as follows:

Bug Fixes

  • In version 1.87.2, the appending of global host tags for OpenTelemetry metrics was removed, which caused significant impact. By default, these tags are now appended again; if removal is required, a new flag has been added in this version for configuration (#2942)
  • Fixed trigger threshold evaluation issue in Flameshot (#2943)
  • Added IPDB configuration capability in Pipeline debugging (#2944)

1.88.0(2026/01/14)

This release is an iterative release, with the following main updates:

New Features

Bug Fixes

  • Fixed the issue of inflated Pod memory collection values (#2933)
  • Fixed the issue where KubernetesPrometheus failed to resume collection after Pod restart (#2936)
  • Fixed the issue where DDTrace NodeJS profiles could not be collected (#2937) 2
  • Fixed multi-step dial testing retry issue (#2915)
  • Fixed AWS Lambda extension collection anomaly (#2918)

Improvements

  • In DataKit log output, a separate file (default is error.log) is now provided for ERROR level logs to prevent them from being overwhelmed by other logs; meanwhile, the bug report will also include this error log (#2940)
  • Optimized the disk cache module (WAL), exposed more metrics and logs, and optimized the impact of .pos files on disk IO (#2935)
  • Added more YAML configurations for SNMP collection and fixed some legacy issues (#2923)
  • Added from_beginning_threshold_size configuration item for container log collection and logfwd (#2934)
  • Added collector_source_ip field to data collected by multiple collectors, indicating the data source (#2819) 3
  • Other optimizations (#2928/#2932/#2930)

Compatibility Adjustments

  • Removed the redundant all field from object data in SNMP collected data (#2923)

  1. We collect them into logging. ↩

  2. To fully support DDTrace NodeJS profile collection, the backend still needs to be upgraded to the latest version. ↩

  3. These collectors include zipkin/logstreaming/beats_output, etc. ↩

Feedback

Is this page helpful? ×