[Feature]Report FD statistical information by luukunn · Pull Request #5646 · PaddlePaddle/FastDeploy

luukunn · 2025-12-18T13:44:53Z

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

Add an information reporting function; it will automatically report when the service is launched. For specific details

Usage or Command

Automatic reporting is enabled by default. To disable this feature, set DO_NOT_TRACK="1".

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.

paddle-bot · 2025-12-18T13:45:00Z

Thanks for your contribution!

codecov-commenter · 2025-12-19T04:21:17Z

Codecov Report

❌ Patch coverage is 77.55102% with 55 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@2c17acd). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/usage/usage_lib.py	77.50%	39 Missing and 15 partials ⚠️
fastdeploy/platforms/base.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5646   +/-   ##
==========================================
  Coverage           ?   67.29%           
==========================================
  Files              ?      349           
  Lines              ?    45017           
  Branches           ?     6934           
==========================================
  Hits               ?    30294           
  Misses             ?    12485           
  Partials           ?     2238

Flag	Coverage Δ
GPU	`67.29% <77.55%> (?)`

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.

…into statistics

Copilot

Pull request overview

This PR introduces a telemetry/usage statistics collection feature for FastDeploy. The system automatically reports platform, hardware, and configuration information to a remote server when the service launches, with the ability to opt-out via the DO_NOT_TRACK=1 environment variable.

Key Changes:

Implements comprehensive usage statistics collection including CPU, GPU, environment, and configuration data
Adds automatic reporting mechanism that triggers on worker initialization and continues every 10 minutes
Integrates reporting into all worker types (GPU, XPU, HPU, GCU)

Reviewed changes

Copilot reviewed 12 out of 13 changed files in this pull request and generated 27 comments.

Show a summary per file

File	Description
fastdeploy/usage/usage_lib.py	Core implementation of usage statistics collection, serialization, and reporting
tests/usage/test_usage_lib.py	Unit tests for usage statistics functionality
fastdeploy/usage/init.py	Package initialization file with copyright header
fastdeploy/worker/gpu_worker.py	Integrates usage reporting into GPU worker initialization
fastdeploy/worker/xpu_worker.py	Integrates usage reporting into XPU worker initialization
fastdeploy/worker/hpu_worker.py	Integrates usage reporting into HPU worker initialization
fastdeploy/worker/gcu_worker.py	Integrates usage reporting into GCU worker initialization
fastdeploy/envs.py	Adds environment variables for controlling usage statistics (DO_NOT_TRACK, server URL, source, config path)
fastdeploy/platforms/base.py	Adds is_cuda_alike() helper method to detect CUDA or ROCm platforms
requirements.txt	Adds py-cpuinfo dependency for CPU information collection
requirements_dcu.txt	Adds py-cpuinfo dependency
requirements_iluvatar.txt	Adds py-cpuinfo dependency
requirements_metaxgpu.txt	Adds py-cpuinfo dependency

Copilot · 2025-12-26T02:45:40Z

+        except:
+            return str(obj)


Bare except clause used without specifying exception type. This catches all exceptions including SystemExit and KeyboardInterrupt, which could hide serious bugs. Specify the expected exception type or at minimum use "except Exception:".

Copilot · 2025-12-26T02:45:41Z

+
+    def _send_to_server(self, data: dict[str, Any]) -> None:
+        try:
+            requests.post(url=_USAGE_STATS_SERVER, json=data, timeout=10)


The HTTP POST request has a 10-second timeout, which could block the reporting thread during network issues. While exceptions are caught, this could still cause delays in the continuous reporting loop. Consider reducing the timeout or making it configurable.

Copilot · 2025-12-26T02:45:41Z

+            return tuple([None] * len(names))
+
+    # Run in subprocess to avoid initializing CUDA as a side effect.
+    mp_ctx = multiprocessing.get_context("fork")


The function uses multiprocessing with "fork" context which can be problematic on some platforms (especially macOS). Consider using "spawn" context or making the context configurable, especially since this is for a one-time property query.

Suggested change

mp_ctx = multiprocessing.get_context("fork")

# Prefer "spawn" for better portability (e.g., macOS); fall back to default if unavailable.

try:

mp_ctx = multiprocessing.get_context("spawn")

except ValueError:

mp_ctx = multiprocessing.get_context()

Copilot · 2025-12-26T02:45:41Z

+    # Try detecting through vendor file
+    vendor_files = [
+        "/sys/class/dmi/id/product_version",
+        "/sys/class/dmi/id/bios_vendor",
+        "/sys/class/dmi/id/product_name",
+        "/sys/class/dmi/id/chassis_asset_tag",
+        "/sys/class/dmi/id/sys_vendor",
+    ]
+    # Mapping of identifiable strings to cloud providers
+    cloud_identifiers = {
+        "amazon": "AWS",
+        "microsoft corporation": "AZURE",
+        "google": "GCP",
+        "oraclecloud": "OCI",
+    }
+
+    for vendor_file in vendor_files:
+        path = Path(vendor_file)
+        if path.is_file():
+            file_content = path.read_text().lower()


The file path /sys/class/dmi/id/ files are Linux-specific and will fail on other operating systems. Consider wrapping this detection logic in a try-except block or checking the OS first to avoid errors on non-Linux systems.

Suggested change

# Try detecting through vendor file

vendor_files = [

"/sys/class/dmi/id/product_version",

"/sys/class/dmi/id/bios_vendor",

"/sys/class/dmi/id/product_name",

"/sys/class/dmi/id/chassis_asset_tag",

"/sys/class/dmi/id/sys_vendor",

]

# Mapping of identifiable strings to cloud providers

cloud_identifiers = {

"amazon": "AWS",

"microsoft corporation": "AZURE",

"google": "GCP",

"oraclecloud": "OCI",

}

for vendor_file in vendor_files:

path = Path(vendor_file)

if path.is_file():

file_content = path.read_text().lower()

# Try detecting through vendor file (Linux-specific paths)

if platform.system().lower() == "linux":

vendor_files = [

"/sys/class/dmi/id/product_version",

"/sys/class/dmi/id/bios_vendor",

"/sys/class/dmi/id/product_name",

"/sys/class/dmi/id/chassis_asset_tag",

"/sys/class/dmi/id/sys_vendor",

]

# Mapping of identifiable strings to cloud providers

cloud_identifiers = {

"amazon": "AWS",

"microsoft corporation": "AZURE",

"google": "GCP",

"oraclecloud": "OCI",

}

for vendor_file in vendor_files:

path = Path(vendor_file)

try:

if path.is_file():

file_content = path.read_text().lower()

else:

continue

except (OSError, IOError):

# Ignore unreadable or missing vendor files

continue

Copilot · 2025-12-26T02:45:42Z

+
+def get_xpu_model():
+    try:
+        result = subprocess.run(["xpu-smi"], capture_output=True, text=True)


The subprocess.run call for xpu-smi doesn't specify a timeout parameter. If the xpu-smi command hangs, it could block the usage reporting indefinitely. Add a timeout parameter (e.g., timeout=5) to prevent hanging.

Suggested change

result = subprocess.run(["xpu-smi"], capture_output=True, text=True)

result = subprocess.run(["xpu-smi"], capture_output=True, text=True, timeout=5)

Copilot · 2025-12-26T02:45:47Z

+        data = {"test": "data"}
+        # Should not raise exception, just log debug message
+        self.usage_message._send_to_server(data)
+


Missing test coverage for the complete flow of _report_usage_once() which collects CPU info, environment variables, and config serialization. These are critical paths that should be tested to ensure data is collected and formatted correctly.

Copilot · 2025-12-26T02:45:47Z

+    return "Unknown"
+
+
+def simple_convert(obj):


Missing docstring for the simple_convert() function. This is a complex function that handles various object types and conversion strategies. Add a comprehensive docstring explaining:

What the function does

Parameter types and meanings

Return type and format

Examples of conversions

Suggested change

def simple_convert(obj):

def simple_convert(obj):

"""

Recursively convert an arbitrary Python object into a JSON-serializable

structure (composed of primitives, lists and dictionaries) or a string.

The conversion strategy is:

1. Primitive values (`None`, `str`, `int`, `float`, `bool`) are returned

as-is.

2. Mappings (`dict`) are converted by recursively converting each value.

3. Iterables (`list`, `tuple`, `set`) are converted into lists with each

element recursively converted.

4. If the object is (or becomes) a string that looks like JSON, an

attempt is made to parse it with :func:`json.loads`. On failure, the

original string is returned.

5. For custom objects that expose one of ``to_dict()``, ``to_json()``,

``__getstate__()`` or ``as_dict()``, the corresponding method is

called and its result is processed recursively if it is a ``dict`` or

parsed as JSON if it is a ``str``.

6. If no conversion method is available, the object's ``__dict__`` (via

:func:`vars`) is converted recursively, excluding private attributes

whose names start with an underscore.

7. As a last resort, :func:`str` is used to obtain a string representation.

This helper is designed to prepare data for logging, telemetry and other

serialization scenarios where complex objects need to be reduced to basic

JSON-friendly structures.

Parameters

----------

obj : Any

The object to convert. This can be a primitive value, container

(``dict``, ``list``, ``tuple``, ``set``), string (including JSON

strings), or an arbitrary user-defined object.

Returns

-------

Any

A value that is typically JSON-serializable, composed of nested

primitives, lists and dictionaries. If an object cannot be reduced to

such a structure, its string representation is returned instead.

Examples

--------

Basic types are returned unchanged::

simple_convert(1) # -> 1

simple_convert("text") # -> "text"

simple_convert(True) # -> True

Containers are converted recursively::

simple_convert({"a": 1, "b": [2, 3]})

# -> {"a": 1, "b": [2, 3]}

JSON strings are parsed when possible::

simple_convert('{"x": 10}')

# -> {"x": 10}

User-defined objects with ``to_dict`` are expanded::

class Config:

def __init__(self):

self.a = 1

def to_dict(self):

return {"a": self.a}

simple_convert(Config())

# -> {"a": 1}

"""

Copilot · 2025-12-26T02:45:48Z

 einops
 setproctitle
 aistudio_sdk
+py-cpuinfo


The new dependency py-cpuinfo is added without a fixed version, so deployments will always pull the latest release from PyPI, which increases exposure to supply-chain attacks. If the package or its distribution channel is compromised in the future, new deployments could automatically execute malicious code with access to your runtime environment and secrets. Pin this dependency to a vetted version (or vendor it) and update it explicitly after review instead of relying on a floating latest version.

Copilot · 2025-12-26T02:45:48Z

 partial_json_parser
 msgspec
 safetensors>=0.7.0
+py-cpuinfo


The new dependency py-cpuinfo is added without a fixed version, so deployments will always pull the latest release from PyPI, which increases exposure to supply-chain attacks. If the package or its distribution channel is compromised in the future, new deployments could automatically execute malicious code with access to your runtime environment and secrets. Pin this dependency to a vetted version (or vendor it) and update it explicitly after review instead of relying on a floating latest version.

Copilot · 2025-12-26T02:45:48Z

 opentelemetry-instrumentation-fastapi
 opentelemetry-instrumentation-logging>=0.57b0
 partial_json_parser
+py-cpuinfo


The new dependency py-cpuinfo is added without a fixed version, so deployments will always pull the latest release from PyPI, which increases exposure to supply-chain attacks. If the package or its distribution channel is compromised in the future, new deployments could automatically execute malicious code with access to your runtime environment and secrets. Pin this dependency to a vetted version (or vendor it) and update it explicitly after review instead of relying on a floating latest version.

…into statistics

…to statistics

* add usage commit * update envs and xpu * add requirements * fix quantization value * add unit test * add unit test * fix unit test * add unit test * add unit test * add unit test * add unit test * add unit test * add unit test * fix FD_USAGE_STATS_SERVER * fix * fix * add doc * add doc * add doc * add doc * add doc * fix file name

luukunn added 2 commits December 16, 2025 16:16

add usage commit

3648384

update envs and xpu

059b10b

luukunn added 2 commits December 18, 2025 21:58

add requirements

41e10a9

add requirements

e2ef272

luukunn force-pushed the statistics branch from 9bf651f to e2ef272 Compare December 18, 2025 13:59

fix quantization value

7554ef5

luukunn added 3 commits December 25, 2025 22:36

add unit test

135528a

add unit test

949170a

Merge branch 'develop' of https://github.com/PaddlePaddle/FastDeploy …

ae73cd0

…into statistics

luukunn force-pushed the statistics branch from bb6ec1e to ae73cd0 Compare December 25, 2025 14:39

luukunn temporarily deployed to Metax_ci December 25, 2025 14:39 — with GitHub Actions Inactive

fix unit test

b83c8e9

luukunn temporarily deployed to Metax_ci December 25, 2025 18:09 — with GitHub Actions Inactive

Jiang-Jia-Jun requested a review from Copilot December 26, 2025 02:40

Copilot started reviewing on behalf of Jiang-Jia-Jun December 26, 2025 02:40 View session

Copilot AI reviewed Dec 26, 2025

View reviewed changes

add unit test

e22007a

luukunn temporarily deployed to Metax_ci December 26, 2025 10:46 — with GitHub Actions Inactive

add unit test

50fed73

luukunn temporarily deployed to Metax_ci December 29, 2025 01:58 — with GitHub Actions Inactive

luukunn temporarily deployed to Metax_ci December 29, 2025 02:31 — with GitHub Actions Inactive

luukunn added 2 commits December 29, 2025 10:46

add unit test

1bf953f

Merge branch 'develop' of https://github.com/PaddlePaddle/FastDeploy …

1ee33f1

…into statistics

luukunn force-pushed the statistics branch from f2e3d93 to 1ee33f1 Compare December 29, 2025 02:47

luukunn had a problem deploying to Metax_ci December 29, 2025 02:47 — with GitHub Actions Error

add unit test

253245c

luukunn temporarily deployed to Metax_ci December 29, 2025 02:57 — with GitHub Actions Inactive

add unit test

233b5a2

luukunn added 3 commits January 12, 2026 18:35

fix FD_USAGE_STATS_SERVER

f2f1993

fix FD_USAGE_STATS_SERVER

c8769e9

Merge branch 'statistics' of https://github.com/luukunn/FastDeploy in…

4954ada

…to statistics

luukunn had a problem deploying to Metax_ci January 12, 2026 10:37 — with GitHub Actions Failure

fix

c8d5e21

luukunn temporarily deployed to Metax_ci January 12, 2026 12:41 — with GitHub Actions Inactive

fix

512640c

luukunn temporarily deployed to Metax_ci January 12, 2026 13:48 — with GitHub Actions Inactive

add doc

e4904c4

luukunn had a problem deploying to Metax_ci January 13, 2026 12:04 — with GitHub Actions Error

add doc

05d0bcd

luukunn had a problem deploying to Metax_ci January 13, 2026 12:06 — with GitHub Actions Error

add doc

799dfb4

luukunn had a problem deploying to Metax_ci January 13, 2026 12:10 — with GitHub Actions Error

add doc

d2415b4

luukunn temporarily deployed to Metax_ci January 13, 2026 12:16 — with GitHub Actions Inactive

add doc

d81fa77

luukunn temporarily deployed to Metax_ci January 13, 2026 12:37 — with GitHub Actions Inactive

fix file name

fa6eec7

luukunn temporarily deployed to Metax_ci January 14, 2026 03:16 — with GitHub Actions Inactive

Merge branch 'develop' into statistics

c4a2818

luukunn had a problem deploying to Metax_ci January 14, 2026 06:55 — with GitHub Actions Failure

Merge branch 'develop' into statistics

527ddd3

luukunn temporarily deployed to Metax_ci January 14, 2026 07:54 — with GitHub Actions Inactive

Jiang-Jia-Jun approved these changes Jan 14, 2026

View reviewed changes

Jiang-Jia-Jun merged commit 93b7675 into PaddlePaddle:develop Jan 14, 2026
16 of 20 checks passed

luukunn deleted the statistics branch March 24, 2026 12:01

Navigation Menu

[Feature]Report FD statistical information - #5646

[Feature]Report FD statistical information#5646
Jiang-Jia-Jun merged 31 commits into
PaddlePaddle:developfrom
luukunn:statistics

luukunn commented Dec 18, 2025 •

edited

Loading

Uh oh!

paddle-bot Bot commented Dec 18, 2025

Uh oh!

codecov-commenter commented Dec 19, 2025 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Copilot AI Dec 26, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

-    mp_ctx = multiprocessing.get_context("fork")
+    # Prefer "spawn" for better portability (e.g., macOS); fall back to default if unavailable.
+    try:
+        mp_ctx = multiprocessing.get_context("spawn")
+    except ValueError:
+        mp_ctx = multiprocessing.get_context()

	result = subprocess.run(["xpu-smi"], capture_output=True, text=True)
	result = subprocess.run(["xpu-smi"], capture_output=True, text=True, timeout=5)

-def simple_convert(obj):
+def simple_convert(obj):
+    """
+    Recursively convert an arbitrary Python object into a JSON-serializable
+    structure (composed of primitives, lists and dictionaries) or a string.
+    The conversion strategy is:
+. Primitive values (`None`, `str`, `int`, `float`, `bool`) are returned
+       as-is.
+. Mappings (`dict`) are converted by recursively converting each value.
+. Iterables (`list`, `tuple`, `set`) are converted into lists with each
+       element recursively converted.
+. If the object is (or becomes) a string that looks like JSON, an
+       attempt is made to parse it with :func:`json.loads`. On failure, the
+       original string is returned.
+. For custom objects that expose one of ``to_dict()``, ``to_json()``,
+       ``__getstate__()`` or ``as_dict()``, the corresponding method is
+       called and its result is processed recursively if it is a ``dict`` or
+       parsed as JSON if it is a ``str``.
+. If no conversion method is available, the object's ``__dict__`` (via
+       :func:`vars`) is converted recursively, excluding private attributes
+       whose names start with an underscore.
+. As a last resort, :func:`str` is used to obtain a string representation.
+    This helper is designed to prepare data for logging, telemetry and other
+    serialization scenarios where complex objects need to be reduced to basic
+    JSON-friendly structures.
+    Parameters
+    ----------
+    obj : Any
+        The object to convert. This can be a primitive value, container
+        (``dict``, ``list``, ``tuple``, ``set``), string (including JSON
+        strings), or an arbitrary user-defined object.
+    Returns
+    -------
+    Any
+        A value that is typically JSON-serializable, composed of nested
+        primitives, lists and dictionaries. If an object cannot be reduced to
+        such a structure, its string representation is returned instead.
+    Examples
+    --------
+    Basic types are returned unchanged::
+        simple_convert(1)                    # -> 1
+        simple_convert("text")               # -> "text"
+        simple_convert(True)                 # -> True
+    Containers are converted recursively::
+        simple_convert({"a": 1, "b": [2, 3]})
+        # -> {"a": 1, "b": [2, 3]}
+    JSON strings are parsed when possible::
+        simple_convert('{"x": 10}')
+        # -> {"x": 10}
+    User-defined objects with ``to_dict`` are expanded::
+        class Config:
+            def __init__(self):
+                self.a = 1
+            def to_dict(self):
+                return {"a": self.a}
+        simple_convert(Config())
+        # -> {"a": 1}
+    """

Conversation

luukunn commented Dec 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Modifications

Usage or Command

Accuracy Tests

Checklist

Uh oh!

paddle-bot Bot commented Dec 18, 2025

Uh oh!

codecov-commenter commented Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 26, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

luukunn commented Dec 18, 2025 •

edited

Loading

codecov-commenter commented Dec 19, 2025 •

edited

Loading