发布日期: 2025-05-17
版本号: v3.4.0

本次更新主要涉及配置调整、功能新增、性能优化和错误修复。配置方面，乱序原生直方图功能现在始终启用，无需单独开关。新增了OTLP翻译和接收功能，支持可选地将OpenTelemetry显式存储桶直方图转换为自定义存储桶的原生直方图，并可直接接收OTLP增量指标。PromQL解析器允许在持续时间中使用算术运算。性能上，改进了Kubernetes服务发现的效率。错误修复涉及配置文件自动重载、原生直方图摄取、WAL重放中的元数据处理等多个关键环节。同时，增强了直方图函数、抓取配置和告警管理等功能。

更新内容 （中文）

  • [变更] 配置：将设置乱序原生直方图功能（--enable-feature=ooo-native-histograms）设为无效操作。当 out_of_order_time_window 大于零且设置了 --enable-feature=native-histograms 时，乱序原生直方图将始终启用。#16207
  • [功能] OTLP 转换：增加功能标志，可选择将 OpenTelemetry 显式存储桶直方图转换为带自定义存储桶的原生直方图。#15850
  • [功能] OTLP 转换：增加选项以接收 OTLP 指标而不转换名称或属性。#16441
  • [功能] PromQL：允许在 PromQL 解析器中对时间间隔进行算术运算。#16249
  • [功能] OTLP 接收器：初步支持按原样摄入 OTLP 差分指标。#16360
  • [改进] PromQL：为存储桶直方图添加 histogram_fraction 功能。#16095
  • [改进] TSDB：添加 prometheus_tsdb_wal_replay_unknown_refs_total 和 prometheus_tsdb_wbl_replay_unknown_refs_total 指标，用于跟踪 WAL/WBL 回放期间的未知序列引用。#16166
  • [改进] 采集：增加转义方案请求的配置选项。#16066
  • [改进] 配置：为 convert_classic_histograms_to_nhcb 添加全局配置选项。#16226
  • [改进] 告警：使批处理大小可配置（--alertmanager.notification-batch-size）。#16254
  • [性能] Kubernetes 服务发现：提高 endpointSlice 发现效率。#16433
  • [错误修复] 配置：修复规则文件和采集配置文件更改时的自动重载问题。#16340
  • [错误修复] 采集：当摄入功能禁用时跳过原生直方图序列。#16218
  • [错误修复] TSDB：处理 WAL 回放期间重复序列的元数据/墓碑/样例。#16231
  • [错误修复] TSDB：避免在 WAL 回放期间处理有效时间范围外的样例。#16242
  • [错误修复] Promtool：为 PromQL 功能添加功能标志。#16443
  • [错误修复] 规则：修正告警名称与模板数据的日志记录。#15093
  • [错误修复] PromQL：对 histogram_stddev() 和 histogram_stdvar() 使用算术平均值。#16444

更新内容 （原始）

  • [CHANGE] Config: Make setting out-of-order native histograms feature (--enable-feature=ooo-native-histograms) a no-op. Out-of-order native histograms are now always enabled when out_of_order_time_window is greater than zero and --enable-feature=native-histograms is set. #16207
  • [FEATURE] OTLP translate: Add feature flag for optionally translating OTel explicit bucket histograms into native histograms with custom buckets. #15850
  • [FEATURE] OTLP translate: Add option to receive OTLP metrics without translating names or attributes. #16441
  • [FEATURE] PromQL: allow arithmetic operations in durations in PromQL parser. #16249
  • [FEATURE] OTLP receiver: Add primitive support for ingesting OTLP delta metrics as-is. #16360
  • [ENHANCEMENT] PromQL: histogram_fraction for bucket histograms. #16095
  • [ENHANCEMENT] TSDB: add prometheus_tsdb_wal_replay_unknown_refs_total and prometheus_tsdb_wbl_replay_unknown_refs_total metrics to track unknown series references during WAL/WBL replay. #16166
  • [ENHANCEMENT] Scraping: Add config option for escaping scheme request. #16066
  • [ENHANCEMENT] Config: Add global config option for convert_classic_histograms_to_nhcb. #16226
  • [ENHANCEMENT] Alerting: make batch size configurable (--alertmanager.notification-batch-size). #16254
  • [PERF] Kubernetes SD: make endpointSlice discovery more efficient. #16433
  • [BUGFIX] Config: Fix auto-reload on changes to rule and scrape config files. #16340
  • [BUGFIX] Scraping: Skip native histogram series if ingestion is disabled. #16218
  • [BUGFIX] TSDB: Handle metadata/tombstones/exemplars for duplicate series during WAL replay. #16231
  • [BUGFIX] TSDB: Avoid processing exemplars outside the valid time range during WAL replay. #16242
  • [BUGFIX] Promtool: Add feature flags for PromQL features. #16443
  • [BUGFIX] Rules: correct logging of alert name & template data. #15093
  • [BUGFIX] PromQL: Use arithmetic mean for histogram_stddev() and histogram_stdvar() . #16444

下载链接