Don't escape UTF-8 names by default for OpenMetrics 2.0
维护者通常 1 天内回复
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 35/100
- Issue 类型
- 缺陷
- 描述清晰度
- 描述清楚
- 活跃度
- 停滞
- 技术栈
- java
- 领域
- api, backend, documentation
调研方向
Start with prometheus-metrics-config/.../EscapingScheme.java, PrometheusScrapeHandler.java, and ExpositionFormats.java, then inspect how Accept clauses are parsed and matched. Add coverage in ExpositionFormatsTest and a scrape-handler test for OM 2.0, explicit escaping, older formats, and contentNegotiation=false. Update the referenced escaping schemes documentation and verify the linked pull request before starting.
由索引模型根据 Issue 内容生成。
描述
When a scraper negotiates OpenMetrics 2.0 (Accept: application/openmetrics-text;version=2.0.0) without an escaping parameter, PrometheusScrapeHandler escapes UTF-8 metric and label names with EscapingScheme.DEFAULT (underscores). OpenMetrics 2.0 supports UTF-8 names natively and doesn't define an escaping parameter at all, so we could assume that scraper asking for 2.0.0 already supports UTF-8 names. In that case we shouldn't escape by default.
Prometheus main doesn't send escaping for OM 2.0 (https://github.com/prometheus/prometheus/blob/961c9ba40923ca4d7adf4a7a9167e56d0406df83/scrape/scrape.go#L794-L800), so UTF-8 names are currently escaped to underscores when scraping a Java target with OM 2.0. There's a Prometheus-side workaround in https://github.com/prometheus/prometheus/compare/main...bwplotka/om2-accept-escaping, but the default here should be fixed regardless.
The escaping scheme is chosen without looking at the selected format:
EscapingScheme.fromAcceptHeaderreturnsDEFAULTwhen there's noescapingterm: https://github.com/prometheus/client_java/blob/9e9deb6b9e591c2a62b12882d8a987f120661b71/prometheus-metrics-config/src/main/java/io/prometheus/metrics/config/EscapingScheme.java#L29-L67.PrometheusScrapeHandlercomputes it before, and separately from,findWriter: https://github.com/prometheus/client_java/blob/9e9deb6b9e591c2a62b12882d8a987f120661b71/prometheus-metrics-exporter-common/src/main/java/io/prometheus/metrics/exporter/common/PrometheusScrapeHandler.java#L73-L91.ExpositionFormats.findWriterpicks the OM 2.0 writer: https://github.com/prometheus/client_java/blob/9e9deb6b9e591c2a62b12882d8a987f120661b71/prometheus-metrics-exposition-textformats/src/main/java/io/prometheus/metrics/expositionformats/ExpositionFormats.java#L63-L86.
Proposed change (pseudo code):
// PrometheusScrapeHandler
String acceptHeader = request.getHeader("Accept");
ExpositionFormatWriter writer = expositionFormats.findWriter(acceptHeader);
EscapingScheme escapingScheme = EscapingScheme.fromAcceptHeader(acceptHeader, defaultFor(acceptHeader));
// Only default to UTF-8 when 2.0.0 was explicitly requested. With
// contentNegotiation=false the OM 2.0 writer also serves OM 1.0 / unversioned
// requests, e.g. from pre-3.0 Prometheus that can't parse UTF-8 names.
EscapingScheme defaultFor(String acceptHeader) {
return "2.0.0".equals(parseOpenMetricsVersion(acceptHeader))
? EscapingScheme.ALLOW_UTF8
: EscapingScheme.DEFAULT;
}
// EscapingScheme
public static EscapingScheme fromAcceptHeader(@Nullable String acceptHeader, EscapingScheme fallback) {
// Same parsing as today. An explicit escaping term always wins, also for OM 2.0
// (e.g. a Prometheus with legacy name validation asking for underscores).
// Return fallback instead of DEFAULT when the term is missing or unknown.
}
Existing fromAcceptHeader(String) can keep delegating with DEFAULT to stay API compatible. Ideally the version and escaping term should come from the same Accept clause that findWriter matched. Today both helpers scan the whole header, so a parameter from another clause can leak in.
Tests to add (e.g. ExpositionFormatsTest, plus a scrape handler test):
version=2.0.0withoutescapingwrites unescaped UTF-8 names.version=2.0.0;escaping=underscoresstill escapes.version=1.0.0, unversioned OpenMetrics andtext/plainwithoutescapingkeep the underscores default, including withcontentNegotiation=false.
The escaping schemes doc should get the same OM 2.0 exception, since its "Default Behavior" section says to use underscores whenever escaping is missing: https://github.com/prometheus/docs/blob/605cf81fefc2e8e91f8ba89bb1555ae52a43a318/docs/instrumenting/escaping_schemes.md#default-behavior.
Matching client_golang issue: https://github.com/prometheus/client_golang/issues/2149
- 主要语言
- Java
- 星标
- 2.3k
- 派生
- 832
- 平均合并
- 1 天 7 小时
- 30 天内合并 PR
- 63
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
prometheus/client_java 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 78/100
prometheus/client_java#2516 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 82/100
prometheus/client_java#2416 · 1 条评论 ·
维护者通常 1 天内回复
-
Switch Micrometer compatibility workflow to upstream once typed-descriptor path becomes default可能重新可做 关联的 PR 已关闭且未合并。 未关闭
难度 1/5 1 小时以内 新手友好度 86/100
prometheus/client_java#2182 ·
维护者通常 1 天内回复
-
难度 4/5 3-5 天 新手友好度 62/100
prometheus/client_java#2518 ·
维护者通常 1 天内回复
-
难度 5/5 一周以上 新手友好度 35/100
prometheus/client_java#2306 · 10 条评论 · 4 个 reaction ·
维护者通常 1 天内回复
查看 prometheus/client_java 的全部 Issue
相似的 Issue
-
[Bug] Logo style setting missing and `classic` style not applied across multiple platforms (v3.1.0)未关闭
难度 2/5 1-3 小时 新手友好度 75/100
Stirling-Tools/Stirling-PDF#8382 · 1 个 reaction ·
维护者通常 2 天内回复
-
难度 1/5 1 小时以内 新手友好度 70/100
维护者通常 4 天内回复
-
enhancement untriaged
难度 2/5 1-3 小时 新手友好度 78/100
opensearch-project/sql#5842 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 85/100
openhab/openhab-addons#21882 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 78/100
YunaiV/ruoyi-vue-pro#1273 ·
维护者通常 3 天内回复