Merge branch 'dev' into 'main'

added way to include driver version in api

See merge request oscar.krause/fastapi-dls!35
This commit is contained in:
Oscar Krause 2024-10-24 13:28:08 +02:00
commit a7fe8b867e
7 changed files with 311 additions and 83 deletions

View File

@ -1,4 +1,4 @@
FROM python:3.11-alpine
FROM python:3.12-alpine
ARG VERSION
ARG COMMIT=""

View File

@ -2,7 +2,8 @@
Minimal Delegated License Service (DLS).
Compatibility tested with official NLS 2.0.1, 2.1.0, 3.1.0. For Driver compatibility see [here](#setup-client).
Compatibility tested with official NLS 2.0.1, 2.1.0, 3.1.0, 3.3.1. For Driver compatibility
see [compatibility matrix](#vgpu-software-compatibility-matrix).
This service can be used without internet connection.
Only the clients need a connection to this service on configured port.
@ -26,7 +27,7 @@ Only the clients need a connection to this service on configured port.
---
[[_TOC_]]
[TOC]
# Setup (Service)
@ -42,6 +43,9 @@ Tested with Ubuntu 22.10 (EOL!) (from Proxmox templates), actually its consuming
- Make sure your timezone is set correct on you fastapi-dls server and your client
This guide does not show how to install vGPU host drivers! Look at the official documentation packed with the driver
releases.
## Docker
Docker-Images are available here for Intel (x86), AMD (amd64) and ARM (arm64):
@ -434,32 +438,8 @@ client has 19.2 hours in which to re-establish connectivity before its license e
**The token file has to be copied! It's not enough to C&P file contents, because there can be special characters.**
Successfully tested with this package versions:
| vGPU Suftware | Driver Branch | Linux vGPU Manager | Linux Driver | Windows Driver | Release Date | EOL Date |
|:-------------:|:-------------:|--------------------|--------------|----------------|--------------:|--------------:|
| `17.2` | R550 | `550.90.05` | `550.90.07` | `552.55` | June 2024 | February 2025 |
| `17.1` | R550 | `550.54.16` | `550.54.15` | `551.78` | March 2024 | |
| `17.0` | R550 | `550.54.10` | `550.54.14` | `551.61` | February 2024 | |
| `16.6` | R535 | `535.183.04` | `535.183.01` | `538.67` | June 2024 | July 2026 |
| `16.5` | R535 | `535.161.05` | `535.161.08` | `538.46` | February 2024 | |
| `16.4` | R535 | `535.161.05` | `535.161.07` | `538.33` | February 2024 | |
| `16.3` | R535 | `535.154.02` | `535.154.05` | `538.15` | January 2024 | |
| `16.2` | R535 | `535.129.03` | `535.129.03` | `537.70` | October 2023 | |
| `16.1` | R535 | `535.104.06` | `535.104.05` | `537.13` | August 2023 | |
| `16.0` | R535 | `535.54.06` | `535.54.03` | `536.22` | July 2023 | |
| `15.4` | R525 | `525.147.01` | `525.147.05` | `529.19` | June 2023 | October 2023 |
| `15.3` | R525 | `525.125.03` | `525.125.06` | `529.11` | June 2023 | |
| `15.2` | R525 | `525.105.14` | `525.105.17` | `528.89` | March 2023 | |
| `15.1` | R525 | `525.85.07` | `525.85.05` | `528.24` | January 2023 | |
| `15.0` | R525 | `525.60.12` | `525.60.13` | `527.41` | December 2022 | |
| `14.4` | R510 | `510.108.03` | `510.108.03` | `514.08` | December 2022 | February 2023 |
| `14.3` | R510 | `510.108.03` | `510.108.03` | `513.91` | November 2022 | |
- https://docs.nvidia.com/grid/index.html
- https://docs.nvidia.com/grid/gpus-supported-by-vgpu.html
*To get the latest drivers, visit Nvidia or search in Discord-Channel `GPU Unlocking` (Server-ID: `829786927829745685`) on channel `licensing` `biggerthanshit`
This guide does not show how to install vGPU guest drivers! Look at the official documentation packed with the driver
releases.
## Linux
@ -535,33 +515,32 @@ Done. For more information check [troubleshoot section](#troubleshoot).
8. Set schedule to `At First Array Start Only`
9. Click on Apply
# Endpoints
# API Endpoints
<details>
<summary>show</summary>
### `GET /`
**`GET /`**
Redirect to `/-/readme`.
### `GET /-/health`
**`GET /-/health`**
Status endpoint, used for *healthcheck*.
### `GET /-/config`
**`GET /-/config`**
Shows current runtime environment variables and their values.
### `GET /-/readme`
**`GET /-/readme`**
HTML rendered README.md.
### `GET /-/manage`
**`GET /-/manage`**
Shows a very basic UI to delete origins or leases.
### `GET /-/origins?leases=false`
**`GET /-/origins?leases=false`**
List registered origins.
@ -569,11 +548,11 @@ List registered origins.
|-----------------|---------|--------------------------------------|
| `leases` | `false` | Include referenced leases per origin |
### `DELETE /-/origins`
**`DELETE /-/origins`**
Deletes all origins and their leases.
### `GET /-/leases?origin=false`
**`GET /-/leases?origin=false`**
List current leases.
@ -581,15 +560,15 @@ List current leases.
|-----------------|---------|-------------------------------------|
| `origin` | `false` | Include referenced origin per lease |
### `DELETE /-/lease/{lease_ref}`
**`DELETE /-/lease/{lease_ref}`**
Deletes an lease.
### `GET /-/client-token`
**`GET /-/client-token`**
Generate client token, (see [installation](#installation)).
### Others
**Others**
There are many other internal api endpoints for handling authentication and lease process.
</details>
@ -616,9 +595,9 @@ Logs are available in `C:\Users\Public\Documents\Nvidia\LoggingLog.NVDisplay.Con
## Linux
### `uvicorn.error:Invalid HTTP request received.`
### Invalid HTTP request
This message can be ignored.
This error message: `uvicorn.error:Invalid HTTP request received.` can be ignored.
- Ref. https://github.com/encode/uvicorn/issues/441
@ -744,11 +723,40 @@ The error message can safely be ignored (since we have no license limitation :P)
</details>
# vGPU Software Compatibility Matrix
Successfully tested with this package versions.
| vGPU Suftware | Driver Branch | Linux vGPU Manager | Linux Driver | Windows Driver | Release Date | EOL Date |
|:-------------:|:-------------:|--------------------|--------------|----------------|--------------:|--------------:|
| `17.4` | R550 | `550.127.06` | `550.127.05` | `553.24` | October 2024 | February 2025 |
| `17.3` | R550 | `550.90.05` | `550.90.07` | `552.74` | July 2024 | |
| `17.2` | R550 | `550.90.05` | `550.90.07` | `552.55` | June 2024 | |
| `17.1` | R550 | `550.54.16` | `550.54.15` | `551.78` | March 2024 | |
| `17.0` | R550 | `550.54.10` | `550.54.14` | `551.61` | February 2024 | |
| `16.8` | R535 | `535.216.01` | `535.216.01` | `538.95` | October 2024 | July 2026 |
| `16.7` | R535 | `535.183.04` | `535.183.06` | `538.78` | July 2024 | |
| `16.6` | R535 | `535.183.04` | `535.183.01` | `538.67` | June 2024 | |
| `16.5` | R535 | `535.161.05` | `535.161.08` | `538.46` | February 2024 | |
| `16.4` | R535 | `535.161.05` | `535.161.07` | `538.33` | February 2024 | |
| `16.3` | R535 | `535.154.02` | `535.154.05` | `538.15` | January 2024 | |
| `16.2` | R535 | `535.129.03` | `535.129.03` | `537.70` | October 2023 | |
| `16.1` | R535 | `535.104.06` | `535.104.05` | `537.13` | August 2023 | |
| `16.0` | R535 | `535.54.06` | `535.54.03` | `536.22` | July 2023 | |
| `15.4` | R525 | `525.147.01` | `525.147.05` | `529.19` | June 2023 | December 2023 |
| `14.4` | R510 | `510.108.03` | `510.108.03` | `514.08` | December 2022 | February 2023 |
- https://docs.nvidia.com/grid/index.html
- https://docs.nvidia.com/grid/gpus-supported-by-vgpu.html
*To get the latest drivers, visit Nvidia or search in Discord-Channel `GPU Unlocking` (Server-ID: `829786927829745685`)
on channel `licensing`
# Credits
Thanks to vGPU community and all who uses this project and report bugs.
Special thanks to
Special thanks to:
- @samicrusader who created build file for **ArchLinux**
- @cyrus who wrote the section for **openSUSE**

View File

@ -1,39 +1,42 @@
import logging
from base64 import b64encode as b64enc
from calendar import timegm
from contextlib import asynccontextmanager
from datetime import datetime, timedelta
from hashlib import sha256
from uuid import uuid4
from os.path import join, dirname
from json import loads as json_loads
from os import getenv as env
from os.path import join, dirname
from uuid import uuid4
from dateutil.relativedelta import relativedelta
from dotenv import load_dotenv
from fastapi import FastAPI
from fastapi.requests import Request
from json import loads as json_loads
from datetime import datetime, timedelta
from dateutil.relativedelta import relativedelta
from calendar import timegm
from jose import jws, jwk, jwt, JWTError
from jose.constants import ALGORITHMS
from starlette.middleware.cors import CORSMiddleware
from starlette.responses import StreamingResponse, JSONResponse as JSONr, HTMLResponse as HTMLr, Response, RedirectResponse
from sqlalchemy import create_engine
from sqlalchemy.orm import sessionmaker
from starlette.middleware.cors import CORSMiddleware
from starlette.responses import StreamingResponse, JSONResponse as JSONr, HTMLResponse as HTMLr, Response, RedirectResponse
from util import load_key, load_file
from orm import Origin, Lease, init as db_init, migrate
from util import load_key, load_file
# Load variables
load_dotenv('../version.env')
# Get current timezone
TZ = datetime.now().astimezone().tzinfo
# Load basic variables
VERSION, COMMIT, DEBUG = env('VERSION', 'unknown'), env('COMMIT', 'unknown'), bool(env('DEBUG', False))
config = dict(openapi_url=None, docs_url=None, redoc_url=None) # dict(openapi_url='/-/openapi.json', docs_url='/-/docs', redoc_url='/-/redoc')
app = FastAPI(title='FastAPI-DLS', description='Minimal Delegated License Service (DLS).', version=VERSION, **config)
# Database connection
db = create_engine(str(env('DATABASE', 'sqlite:///db.sqlite')))
db_init(db), migrate(db)
# everything prefixed with "INSTANCE_*" is used as "SERVICE_INSTANCE_*" or "SI_*" in official dls service
# Load DLS variables (all prefixed with "INSTANCE_*" is used as "SERVICE_INSTANCE_*" or "SI_*" in official dls service)
DLS_URL = str(env('DLS_URL', 'localhost'))
DLS_PORT = int(env('DLS_PORT', '443'))
SITE_KEY_XID = str(env('SITE_KEY_XID', '00000000-0000-0000-0000-000000000000'))
@ -51,6 +54,40 @@ CORS_ORIGINS = str(env('CORS_ORIGINS', '')).split(',') if (env('CORS_ORIGINS'))
jwt_encode_key = jwk.construct(INSTANCE_KEY_RSA.export_key().decode('utf-8'), algorithm=ALGORITHMS.RS256)
jwt_decode_key = jwk.construct(INSTANCE_KEY_PUB.export_key().decode('utf-8'), algorithm=ALGORITHMS.RS256)
# Logging
LOG_LEVEL = logging.DEBUG if DEBUG else logging.INFO
logging.basicConfig(format='[{levelname:^7}] [{module:^15}] {message}', style='{')
logger = logging.getLogger(__name__)
logger.setLevel(LOG_LEVEL)
logging.getLogger('util').setLevel(LOG_LEVEL)
logging.getLogger('NV').setLevel(LOG_LEVEL)
# FastAPI
@asynccontextmanager
async def lifespan(_: FastAPI):
# on startup
logger.info(f'''
Using timezone: {str(TZ)}. Make sure this is correct and match your clients!
Your clients renew their license every {str(Lease.calculate_renewal(LEASE_RENEWAL_PERIOD, LEASE_RENEWAL_DELTA))}.
If the renewal fails, the license is {str(LEASE_RENEWAL_DELTA)} valid.
Your client-token file (.tok) is valid for {str(CLIENT_TOKEN_EXPIRE_DELTA)}.
''')
logger.info(f'Debug is {"enabled" if DEBUG else "disabled"}.')
yield
# on shutdown
logger.info(f'Shutting down ...')
config = dict(openapi_url=None, docs_url=None, redoc_url=None) # dict(openapi_url='/-/openapi.json', docs_url='/-/docs', redoc_url='/-/redoc')
app = FastAPI(title='FastAPI-DLS', description='Minimal Delegated License Service (DLS).', version=VERSION, lifespan=lifespan, **config)
app.debug = DEBUG
app.add_middleware(
CORSMiddleware,
@ -60,17 +97,16 @@ app.add_middleware(
allow_headers=['*'],
)
logging.basicConfig()
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG if DEBUG else logging.INFO)
# Helper
def __get_token(request: Request) -> dict:
authorization_header = request.headers.get('authorization')
token = authorization_header.split(' ')[1]
return jwt.decode(token=token, key=jwt_decode_key, algorithms=ALGORITHMS.RS256, options={'verify_aud': False})
# Endpoints
@app.get('/', summary='Index')
async def index():
return RedirectResponse('/-/readme')
@ -108,7 +144,7 @@ async def _config():
@app.get('/-/readme', summary='* Readme')
async def _readme():
from markdown import markdown
content = load_file('../README.md').decode('utf-8')
content = load_file(join(dirname(__file__), '../README.md')).decode('utf-8')
return HTMLr(markdown(text=content, extensions=['tables', 'fenced_code', 'md_in_html', 'nl2br', 'toc']))
@ -540,18 +576,6 @@ async def leasing_v1_lessor_shutdown(request: Request):
return JSONr(response)
@app.on_event('startup')
async def app_on_startup():
logger.info(f'''
Using timezone: {str(TZ)}. Make sure this is correct and match your clients!
Your clients renew their license every {str(Lease.calculate_renewal(LEASE_RENEWAL_PERIOD, LEASE_RENEWAL_DELTA))}.
If the renewal fails, the license is {str(LEASE_RENEWAL_DELTA)} valid.
Your client-token file (.tok) is valid for {str(CLIENT_TOKEN_EXPIRE_DELTA)}.
''')
if __name__ == '__main__':
import uvicorn

View File

@ -1,10 +1,12 @@
from datetime import datetime, timedelta
from dateutil.relativedelta import relativedelta
from dateutil.relativedelta import relativedelta
from sqlalchemy import Column, VARCHAR, CHAR, ForeignKey, DATETIME, update, and_, inspect, text
from sqlalchemy.engine import Engine
from sqlalchemy.orm import sessionmaker, declarative_base
from util import NV
Base = declarative_base()
@ -23,6 +25,8 @@ class Origin(Base):
return f'Origin(origin_ref={self.origin_ref}, hostname={self.hostname})'
def serialize(self) -> dict:
_ = NV().find(self.guest_driver_version)
return {
'origin_ref': self.origin_ref,
# 'service_instance_xid': self.service_instance_xid,
@ -30,6 +34,7 @@ class Origin(Base):
'guest_driver_version': self.guest_driver_version,
'os_platform': self.os_platform,
'os_version': self.os_version,
'$driver': _ if _ is not None else None,
}
@staticmethod

View File

@ -1,10 +1,17 @@
def load_file(filename) -> bytes:
import logging
logging.basicConfig()
def load_file(filename: str) -> bytes:
log = logging.getLogger(f'{__name__}')
log.debug(f'Loading contents of file "{filename}')
with open(filename, 'rb') as file:
content = file.read()
return content
def load_key(filename) -> "RsaKey":
def load_key(filename: str) -> "RsaKey":
try:
# Crypto | Cryptodome on Debian
from Crypto.PublicKey import RSA
@ -13,6 +20,8 @@ def load_key(filename) -> "RsaKey":
from Cryptodome.PublicKey import RSA
from Cryptodome.PublicKey.RSA import RsaKey
log = logging.getLogger(__name__)
log.debug(f'Importing RSA-Key from "{filename}"')
return RSA.import_key(extern_key=load_file(filename), passphrase=None)
@ -24,5 +33,50 @@ def generate_key() -> "RsaKey":
except ModuleNotFoundError:
from Cryptodome.PublicKey import RSA
from Cryptodome.PublicKey.RSA import RsaKey
log = logging.getLogger(__name__)
log.debug(f'Generating RSA-Key')
return RSA.generate(bits=2048)
class NV:
__DRIVER_MATRIX_FILENAME = 'static/driver_matrix.json'
__DRIVER_MATRIX: None | dict = None # https://docs.nvidia.com/grid/ => "Driver Versions"
def __init__(self):
self.log = logging.getLogger(self.__class__.__name__)
if NV.__DRIVER_MATRIX is None:
from json import load as json_load
try:
file = open(NV.__DRIVER_MATRIX_FILENAME)
NV.__DRIVER_MATRIX = json_load(file)
file.close()
self.log.debug(f'Successfully loaded "{NV.__DRIVER_MATRIX_FILENAME}".')
except Exception as e:
NV.__DRIVER_MATRIX = {} # init empty dict to not try open file everytime, just when restarting app
# self.log.warning(f'Failed to load "{NV.__DRIVER_MATRIX_FILENAME}": {e}')
@staticmethod
def find(version: str) -> dict | None:
if NV.__DRIVER_MATRIX is None:
return None
for idx, (key, branch) in enumerate(NV.__DRIVER_MATRIX.items()):
for release in branch.get('$releases'):
linux_driver = release.get('Linux Driver')
windows_driver = release.get('Windows Driver')
if version == linux_driver or version == windows_driver:
tmp = branch.copy()
tmp.pop('$releases')
is_latest = release.get('vGPU Software') == branch.get('Latest Release in Branch')
return {
'software_branch': branch.get('vGPU Software Branch'),
'branch_version': release.get('vGPU Software'),
'driver_branch': branch.get('Driver Branch'),
'branch_status': branch.get('vGPU Branch Status'),
'release_date': release.get('Release Date'),
'eol': branch.get('EOL Date') if is_latest else None,
'is_latest': is_latest,
}
return None

View File

@ -1,8 +1,8 @@
fastapi==0.111.0
uvicorn[standard]==0.29.0
fastapi==0.115.3
uvicorn[standard]==0.32.0
python-jose==3.3.0
pycryptodome==3.20.0
pycryptodome==3.21.0
python-dateutil==2.8.2
sqlalchemy==2.0.30
markdown==3.6
sqlalchemy==2.0.36
markdown==3.7
python-dotenv==1.0.1

View File

@ -0,0 +1,137 @@
import logging
logging.basicConfig()
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)
URL = 'https://docs.nvidia.com/vgpu/index.html'
BRANCH_STATUS_KEY, SOFTWARE_BRANCH_KEY, = 'vGPU Branch Status', 'vGPU Software Branch'
VGPU_KEY, GRID_KEY, DRIVER_BRANCH_KEY = 'vGPU Software', 'vGPU Software', 'Driver Branch'
LINUX_VGPU_MANAGER_KEY, LINUX_DRIVER_KEY = 'Linux vGPU Manager', 'Linux Driver'
WINDOWS_VGPU_MANAGER_KEY, WINDOWS_DRIVER_KEY = 'Windows vGPU Manager', 'Windows Driver'
ALT_VGPU_MANAGER_KEY = 'vGPU Manager'
RELEASE_DATE_KEY, LATEST_KEY, EOL_KEY = 'Release Date', 'Latest Release in Branch', 'EOL Date'
JSON_RELEASES_KEY = '$releases'
def __driver_versions(html: 'BeautifulSoup'):
def __strip(_: str) -> str:
# removes content after linebreak (e.g. "Hello\n World" to "Hello")
_ = _.strip()
tmp = _.split('\n')
if len(tmp) > 0:
return tmp[0]
return _
# find wrapper for "DriverVersions" and find tables
data = html.find('div', {'id': 'driver-versions'})
items = data.findAll('bsp-accordion', {'class': 'Accordion-items-item'})
for item in items:
software_branch = item.find('div', {'class': 'Accordion-items-item-title'}).text.strip()
software_branch = software_branch.replace(' Releases', '')
matrix_key = software_branch.lower()
# driver version info from table-heads (ths) and table-rows (trs)
table = item.find('table')
ths, trs = table.find_all('th'), table.find_all('tr')
headers, releases = [header.text.strip() for header in ths], []
for trs in trs:
tds = trs.find_all('td')
if len(tds) == 0: # skip empty
continue
# create dict with table-heads as key and cell content as value
x = {headers[i]: __strip(cell.text) for i, cell in enumerate(tds)}
releases.append(x)
# add to matrix
MATRIX.update({matrix_key: {JSON_RELEASES_KEY: releases}})
def __release_branches(html: 'BeautifulSoup'):
# find wrapper for "AllReleaseBranches" and find table
data = html.find('div', {'id': 'all-release-branches'})
table = data.find('table')
# branch releases info from table-heads (ths) and table-rows (trs)
ths, trs = table.find_all('th'), table.find_all('tr')
headers = [header.text.strip() for header in ths]
for trs in trs:
tds = trs.find_all('td')
if len(tds) == 0: # skip empty
continue
# create dict with table-heads as key and cell content as value
x = {headers[i]: cell.text.strip() for i, cell in enumerate(tds)}
# get matrix_key
software_branch = x.get(SOFTWARE_BRANCH_KEY)
matrix_key = software_branch.lower()
# add to matrix
MATRIX.update({matrix_key: MATRIX.get(matrix_key) | x})
def __debug():
# print table head
s = f'{SOFTWARE_BRANCH_KEY:^21} | {BRANCH_STATUS_KEY:^21} | {VGPU_KEY:^13} | {LINUX_VGPU_MANAGER_KEY:^21} | {LINUX_DRIVER_KEY:^21} | {WINDOWS_VGPU_MANAGER_KEY:^21} | {WINDOWS_DRIVER_KEY:^21} | {RELEASE_DATE_KEY:>21} | {EOL_KEY:>21}'
print(s)
# iterate over dict & format some variables to not overload table
for idx, (key, branch) in enumerate(MATRIX.items()):
branch_status = branch.get(BRANCH_STATUS_KEY)
branch_status = branch_status.replace('Branch ', '')
branch_status = branch_status.replace('Long-Term Support', 'LTS')
branch_status = branch_status.replace('Production', 'Prod.')
software_branch = branch.get(SOFTWARE_BRANCH_KEY).replace('NVIDIA ', '')
for release in branch.get(JSON_RELEASES_KEY):
version = release.get(VGPU_KEY, release.get(GRID_KEY, ''))
linux_manager = release.get(LINUX_VGPU_MANAGER_KEY, release.get(ALT_VGPU_MANAGER_KEY, ''))
linux_driver = release.get(LINUX_DRIVER_KEY)
windows_manager = release.get(WINDOWS_VGPU_MANAGER_KEY, release.get(ALT_VGPU_MANAGER_KEY, ''))
windows_driver = release.get(WINDOWS_DRIVER_KEY)
release_date = release.get(RELEASE_DATE_KEY)
is_latest = release.get(VGPU_KEY) == branch.get(LATEST_KEY)
version = f'{version} *' if is_latest else version
eol = branch.get(EOL_KEY) if is_latest else ''
s = f'{software_branch:^21} | {branch_status:^21} | {version:<13} | {linux_manager:<21} | {linux_driver:<21} | {windows_manager:<21} | {windows_driver:<21} | {release_date:>21} | {eol:>21}'
print(s)
def __dump(filename: str):
import json
file = open(filename, 'w')
json.dump(MATRIX, file)
file.close()
if __name__ == '__main__':
MATRIX = {}
try:
import httpx
from bs4 import BeautifulSoup
except Exception as e:
logger.error(f'Failed to import module: {e}')
logger.info('Run "pip install beautifulsoup4 httpx"')
exit(1)
r = httpx.get(URL)
if r.status_code != 200:
logger.error(f'Error loading "{URL}" with status code {r.status_code}.')
exit(2)
# parse html
soup = BeautifulSoup(r.text, features='html.parser')
# build matrix
__driver_versions(soup)
__release_branches(soup)
# debug output
__debug()
# dump data to file
__dump('../app/static/driver_matrix.json')