[Cherry-Pick][BugFix] Seperate prometheus multiproc dir for single-server multi-dp services (#8059)

[liyonghua0910]

Motivation

💡 If this PR is a Cherry Pick, the PR title needs to follow the format by adding the [Cherry-Pick] label at the very beginning and appending the original PR ID at the end. For example, [Cherry-Pick][CI] Add check trigger and logic(#5191)

💡 如若此PR是Cherry Pick，PR标题需遵循格式，在最开始加上[Cherry-Pick]标签，以及最后面加上原PR ID，例如[Cherry-Pick][CI] Add check trigger and logic(#5191)

Modifications

Usage or Command

Accuracy Tests

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | pr_review | 2026-06-22 18:16:01

📋 Review 摘要

PR 概述：修复单 server 多 DP 场景 Prometheus multiprocess 目录共享导致的指标串扰。
变更范围：metrics helper、engine/common_engine DP 启动、OpenAI multi_api_server、相关单测。
影响面 Tag：[Engine] [APIServer]

问题

未发现阻塞性问题。PR 规范问题在下面章节报,不要在这里重复

📝 PR 规范检查

PR 描述仍保留模板说明/占位内容，Checklist 未按本次 cherry-pick 和测试变更勾选；标题格式符合 cherry-pick 要求，但建议顺手修正 Seperate 拼写。

标题建议（可直接复制）：

• [Cherry-Pick][BugFix] Separate prometheus multiproc dir for single-server multi-DP services (#8059)

PR 描述建议（点击展开，可直接复制）

## Motivation
Fix inaccurate Prometheus multiprocess metrics in single-server multi-DP services. Each DP rank should write metrics to an isolated `PROMETHEUS_MULTIPROC_DIR` subdirectory instead of sharing one directory.

## Modifications
- Add original Prometheus multiprocess dir tracking and `setup_dp_prometheus_dir()` to create per-DP `dp{rank}` directories.
- Move existing DP0 `.db` files into `dp0/` and switch DP worker launch paths to the per-DP directory when `FD_ENABLE_INTERNAL_ADAPTER` is enabled.
- Reuse the same helper in OpenAI `multi_api_server` so each API server subprocess receives an isolated metrics directory.
- Update related metrics, multi API server, cache transfer manager, and graph optimization tests.

## Usage or Command
N/A

## Accuracy Tests
N/A

## Checklist

- [x] Add at least a tag in the PR title.
  - Tag list: [`[FDConfig]`,`[APIServer]`,`[Engine]`, `[Scheduler]`, `[PD Disaggregation]`, `[Executor]`, `[Graph Optimization]`, `[Speculative Decoding]`, `[RL]`, `[Models]`, `[Quantization]`, `[Loader]`, `[OP]`, `[KVCache]`, `[DataProcessor]`, `[BugFix]`, `[Docs]`, `[CI]`, `[Optimization]`, `[Feature]`, `[Benchmark]`, `[Others]`, `[XPU]`, `[HPU]`, `[GCU]`, `[DCU]`, `[Iluvatar]`, `[Metax]`]
  - You can add new tags based on the PR content, but the semantics must be clear.
- [ ] Format your code, run `pre-commit` before commit.
- [x] Add unit tests. Please write the reason in this PR if no unit tests.
- [x] Provide accuracy results.
- [x] If the current PR is submitting to the `release` branch, make sure the PR has been submitted to the `develop` branch, then cherry-pick it to the `release` branch with the `[Cherry-Pick]` PR tag.

总体评价

本轮按风险优先审查了 Prometheus multiprocess 目录初始化、DP worker 启动顺序、OpenAI multi API server 的 env 传递，以及相关测试补丁；未发现需要阻塞合入的确定性实现问题。建议补齐 PR 描述，便于 release cherry-pick 追溯变更动机和验证范围。