Changelog¶

1.89.1(2026/02/12)¶

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

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)

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)

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

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)

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

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）

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

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)

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

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)

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)

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

We collect them into logging. ↩
To fully support DDTrace NodeJS profile collection, the backend still needs to be upgraded to the latest version. ↩
These collectors include zipkin/logstreaming/beats_output, etc. ↩