[PD Disaggregation] Add timestamp for analyzing splitwise deployment

[juncaipeng]

Motivation

增加时间戳，用于分析系统耗时。

Modifications

全流程打点。

Usage or Command

请求加上 collect_metrics=True 字段

Accuracy Tests

单侧覆盖

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.

[Copilot]

Pull request overview

This PR refactors the request timing metrics infrastructure to support comprehensive performance analysis of splitwise (prefill/decode disaggregation) deployments. The changes centralize timing attributes from the Request class into a dedicated RequestMetrics dataclass, adding numerous timestamp fields to track the full lifecycle of requests across prefill and decode nodes.

Key Changes

• Introduced a comprehensive RequestMetrics dataclass with 20+ timestamp fields tracking request flow through the system
• Migrated timing attributes from Request class to the new metrics object
• Added timestamp recording at critical points: scheduler receipt, resource allocation, prefill/decode handoff, and token generation
• Changed default value of FD_ENABLE_CACHE_TASK from enabled to disabled

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
fastdeploy/engine/request.py	Added RequestMetrics dataclass with comprehensive timestamp fields and helper methods; removed timing attributes from Request class
fastdeploy/output/token_processor.py	Refactored to use metrics object instead of direct timing attributes; updated token processing to record timing via metrics methods
fastdeploy/engine/engine.py	Updated to record preprocess timing in metrics object
fastdeploy/engine/async_llm.py	Updated request preprocessing to use metrics.preprocess_start_time
fastdeploy/engine/common_engine.py	Added timestamp recording at key lifecycle points (engine_get_req_time, ask_decode_resource times, inference_start_time); updated comment to correctly indicate v0 scheduler usage
fastdeploy/engine/sched/resource_manager_v1.py	Removed direct inference_start_time assignments; added metrics propagation for decode scenarios
fastdeploy/engine/resource_manager.py	Removed direct inference_start_time assignment during resource allocation
fastdeploy/entrypoints/openai/protocol.py	Added metrics field to ChatCompletionStreamResponse
fastdeploy/entrypoints/openai/serving_chat.py	Updated to include metrics in streaming response chunks
fastdeploy/envs.py	Changed FD_ENABLE_CACHE_TASK default from "1" to "0"

[paddle-bot]

Thanks for your contribution!

[codecov-commenter]

Codecov Report

❌ Patch coverage is 38.29787% with 87 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@69e003a). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/engine/request.py	53.12%	28 Missing and 2 partials ⚠️
fastdeploy/engine/common_engine.py	0.00%	28 Missing ⚠️
fastdeploy/output/token_processor.py	27.58%	21 Missing ⚠️
fastdeploy/engine/sched/resource_manager_v1.py	0.00%	4 Missing ⚠️
fastdeploy/entrypoints/openai/serving_chat.py	60.00%	1 Missing and 1 partial ⚠️
fastdeploy/scheduler/splitwise_scheduler.py	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5317   +/-   ##
==========================================
  Coverage           ?   58.90%           
==========================================
  Files              ?      324           
  Lines              ?    40144           
  Branches           ?     6062           
==========================================
  Hits               ?    23648           
  Misses             ?    14623           
  Partials           ?     1873           

Flag	Coverage Δ
GPU	58.90% <38.29%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.