Files
chia-blockchain/chia/data_layer/s3_plugin_service.py
Matt HauffandGitHub 40db4635a8 [LABS-245] Enable PEP604 Ruff rules (#20269)
* Enable PEP604 Ruff rules

* Fix harcoded signature in test

* Hack CLVMStreamable test with note to fast follow
2025-11-18 12:34:00 -08:00

464 lines
19 KiB
Python

from __future__ import annotations
import asyncio
import concurrent.futures
import functools
import json
import logging
import os
import shutil
import sys
import tempfile
from dataclasses import dataclass
from pathlib import Path
from typing import Any, overload
from urllib.parse import urlparse
import boto3
import yaml
from aiohttp import web
from botocore.exceptions import ClientError
from chia_rs.sized_bytes import bytes32
from chia.data_layer.download_data import is_filename_valid
log = logging.getLogger(__name__)
plugin_name = "Chia S3 Datalayer plugin"
plugin_version = "0.1.0"
@dataclass(frozen=True)
class StoreConfig:
id: bytes32
bucket: str | None
urls: set[str]
@classmethod
def unmarshal(cls, d: dict[str, Any]) -> StoreConfig:
upload_bucket = d.get("upload_bucket", None)
if upload_bucket and len(upload_bucket) == 0:
upload_bucket = None
return StoreConfig(bytes32.from_hexstr(d["store_id"]), upload_bucket, d.get("download_urls", set()))
def marshal(self) -> dict[str, Any]:
return {"store_id": self.id.hex(), "upload_bucket": self.bucket, "download_urls": self.urls}
class S3Plugin:
boto_resource: boto3.resource
port: int
region: str
aws_access_key_id: str
aws_secret_access_key: str
server_files_path: Path
stores: list[StoreConfig]
instance_name: str
def __init__(
self,
region: str,
aws_access_key_id: str,
aws_secret_access_key: str,
server_files_path: Path,
stores: list[StoreConfig],
instance_name: str,
):
self.boto_resource = boto3.resource(
"s3",
region_name=region,
aws_access_key_id=aws_access_key_id,
aws_secret_access_key=aws_secret_access_key,
)
self.stores = stores
self.instance_name = instance_name
self.server_files_path = server_files_path
async def add_store_id(self, request: web.Request) -> web.Response:
"""Add a store id to the config file. Returns False for store ids that are already in the config."""
self.update_instance_from_config()
try:
data = await request.json()
store_id = bytes32.from_hexstr(data["store_id"])
except Exception as e:
log.error(f"failed parsing request {request} {type(e).__name__} {e}")
return web.json_response({"success": False})
bucket = data.get("bucket", None)
urls = data.get("urls", [])
if not bucket and not urls:
return web.json_response({"success": False, "reason": "bucket or urls must be provided"})
for stores in self.stores:
if store_id == stores.id:
return web.json_response({"success": False, "reason": f"store {store_id.hex()} already exists"})
new_store = StoreConfig(store_id, bucket, urls)
self.stores.append(new_store)
self.update_config()
return web.json_response({"success": True, "id": store_id.hex()})
async def remove_store_id(self, request: web.Request) -> web.Response:
"""Remove a store id from the config file. Returns True for store ids that are not in the config."""
self.update_instance_from_config()
try:
data = await request.json()
store_id = bytes32.from_hexstr(data["store_id"])
except Exception as e:
log.error(f"failed parsing request {request} {e}")
return web.json_response({"success": False})
dirty = False
for i, store in enumerate(self.stores):
if store.id == store_id:
del self.stores[i]
dirty = True
break
if dirty:
self.update_config()
return web.json_response({"success": True, "store_id": store_id.hex()})
async def handle_upload(self, request: web.Request) -> web.Response:
self.update_instance_from_config()
try:
data = await request.json()
except Exception as e:
log.error(f"failed parsing request {request} {type(e).__name__} {e}")
return web.json_response({"handle_upload": False})
store_id = bytes32.from_hexstr(data["store_id"])
for store in self.stores:
if store.id == store_id and store.bucket:
return web.json_response({"handle_upload": True, "bucket": store.bucket})
return web.json_response({"handle_upload": False})
@overload
def get_path_for_filename(self, store_id: bytes32, filename: str, group_files_by_store: bool) -> Path: ...
@overload
def get_path_for_filename(self, store_id: bytes32, filename: None, group_files_by_store: bool) -> None: ...
def get_path_for_filename(self, store_id: bytes32, filename: str | None, group_files_by_store: bool) -> Path | None:
if filename is None:
return None
if group_files_by_store:
return self.server_files_path.joinpath(f"{store_id}").joinpath(filename)
return self.server_files_path.joinpath(filename)
@overload
def get_s3_target_from_path(self, store_id: bytes32, path: Path, group_files_by_store: bool) -> str: ...
@overload
def get_s3_target_from_path(self, store_id: bytes32, path: None, group_files_by_store: bool) -> None: ...
def get_s3_target_from_path(self, store_id: bytes32, path: Path | None, group_files_by_store: bool) -> str | None:
if path is None:
return None
if group_files_by_store:
return f"{store_id}/{path.name}"
return path.name
async def upload(self, request: web.Request) -> web.Response:
try:
data = await request.json()
store_id = bytes32.from_hexstr(data["store_id"])
bucket_str = self.get_bucket(store_id)
my_bucket = self.boto_resource.Bucket(bucket_str)
full_tree_name: str | None = data.get("full_tree_filename", None)
diff_name: str = data["diff_filename"]
group_files_by_store: bool = data.get("group_files_by_store", False)
# filenames must follow the DataLayer naming convention
if full_tree_name is not None:
full_tree_name_to_check = f"{store_id}-{full_tree_name}" if group_files_by_store else full_tree_name
else:
full_tree_name_to_check = None
delta_name_to_check = f"{store_id}-{diff_name}" if group_files_by_store else diff_name
if full_tree_name_to_check is not None and not is_filename_valid(full_tree_name_to_check):
return web.json_response({"uploaded": False})
if not is_filename_valid(delta_name_to_check):
return web.json_response({"uploaded": False})
if not group_files_by_store:
# Pull the store_id from the filename to make sure we only upload for configured stores
full_store_id = None if full_tree_name is None else bytes32.fromhex(full_tree_name[:64])
diff_store_id = bytes32.fromhex(diff_name[:64])
if full_store_id is not None and not (full_store_id == diff_store_id == store_id):
return web.json_response({"uploaded": False})
if full_store_id is None and diff_store_id != store_id:
return web.json_response({"uploaded": False})
full_tree_path = self.get_path_for_filename(store_id, full_tree_name, group_files_by_store)
diff_path = self.get_path_for_filename(store_id, diff_name, group_files_by_store)
target_full_tree_path = self.get_s3_target_from_path(store_id, full_tree_path, group_files_by_store)
target_diff_path = self.get_s3_target_from_path(store_id, diff_path, group_files_by_store)
try:
with concurrent.futures.ThreadPoolExecutor(thread_name_prefix="s3-upload-") as pool:
if full_tree_path is not None:
await asyncio.get_running_loop().run_in_executor(
pool,
functools.partial(
my_bucket.upload_file,
full_tree_path,
target_full_tree_path,
),
)
await asyncio.get_running_loop().run_in_executor(
pool, functools.partial(my_bucket.upload_file, diff_path, target_diff_path)
)
except ClientError as e:
log.error(f"failed uploading file to aws {type(e).__name__} {e}")
return web.json_response({"uploaded": False})
except Exception as e:
log.error(f"failed handling request {request} {type(e).__name__} {e}")
return web.json_response({"uploaded": False})
return web.json_response({"uploaded": True})
async def healthz(self, request: web.Request) -> web.Response:
return web.json_response({"success": True})
async def plugin_info(self, request: web.Request) -> web.Response:
return web.json_response(
{
"name": plugin_name,
"version": plugin_version,
"instance": self.instance_name,
}
)
async def handle_download(self, request: web.Request) -> web.Response:
self.update_instance_from_config()
try:
data = await request.json()
except Exception as e:
log.error(f"failed parsing request {request} {type(e).__name__} {e}")
return web.json_response({"handle_download": False})
store_id = bytes32.from_hexstr(data["store_id"])
parse_result = urlparse(data["url"])
for store in self.stores:
if store.id == store_id and parse_result.scheme == "s3" and data["url"] in store.urls:
return web.json_response({"handle_download": True, "urls": list(store.urls)})
return web.json_response({"handle_download": False})
async def download(self, request: web.Request) -> web.Response:
try:
data = await request.json()
url = data["url"]
filename = data["filename"]
group_files_by_store = data.get("group_files_by_store", False)
# filename must follow the DataLayer naming convention
if not is_filename_valid(filename, group_files_by_store):
return web.json_response({"downloaded": False})
# Pull the store_id from the filename to make sure we only download for configured stores
filename_store_id = bytes32.fromhex(filename[:64])
parse_result = urlparse(url)
should_download = False
for store in self.stores:
if store.id == filename_store_id and parse_result.scheme == "s3" and url in store.urls:
should_download = True
break
if not should_download:
return web.json_response({"downloaded": False})
bucket_str = parse_result.netloc
my_bucket = self.boto_resource.Bucket(bucket_str)
trimmed_filename = filename[65:] if group_files_by_store else filename
target_filename = self.get_path_for_filename(filename_store_id, trimmed_filename, group_files_by_store)
# Create folder for parent directory
target_filename.parent.mkdir(parents=True, exist_ok=True)
log.info(f"downloading {url} to {target_filename}...")
with concurrent.futures.ThreadPoolExecutor(thread_name_prefix="s3-download-") as pool:
await asyncio.get_running_loop().run_in_executor(
pool, functools.partial(my_bucket.download_file, filename, str(target_filename))
)
except Exception as e:
log.error(f"failed parsing request {request} {type(e).__name__} {e}")
return web.json_response({"downloaded": False})
return web.json_response({"downloaded": True})
async def add_missing_files(self, request: web.Request) -> web.Response:
try:
data = await request.json()
store_id = bytes32.from_hexstr(data["store_id"])
bucket_str = self.get_bucket(store_id)
files = json.loads(data["files"])
group_files_by_store: bool = data.get("group_files_by_store", False)
my_bucket = self.boto_resource.Bucket(bucket_str)
existing_file_list = []
for my_bucket_object in my_bucket.objects.all():
existing_file_list.append(my_bucket_object.key)
try:
for file_name in files:
# filenames must follow the DataLayer naming convention
if group_files_by_store:
if not is_filename_valid(f"{store_id}-{file_name}"):
log.error(f"failed uploading file {store_id}-{file_name}, invalid file name")
continue
else:
if not is_filename_valid(file_name):
log.error(f"failed uploading file {file_name}, invalid file name")
continue
if not (bytes32.fromhex(file_name[:64]) == store_id):
log.error(f"failed uploading file {file_name}, store id mismatch")
file_path = self.get_path_for_filename(store_id, file_name, group_files_by_store)
target_file_name = self.get_s3_target_from_path(store_id, file_path, group_files_by_store)
if not os.path.isfile(file_path):
log.error(f"failed uploading file to aws, file {file_path} does not exist")
continue
if file_name in existing_file_list:
log.debug(f"skip {file_name} already in bucket")
continue
with concurrent.futures.ThreadPoolExecutor(thread_name_prefix="s3-missing-") as pool:
await asyncio.get_running_loop().run_in_executor(
pool,
functools.partial(my_bucket.upload_file, file_path, target_file_name),
)
except ClientError as e:
log.error(f"failed uploading file to aws {e}")
return web.json_response({"uploaded": False})
except Exception as e:
log.error(f"failed handling request {request} {e}")
return web.json_response({"uploaded": False})
return web.json_response({"uploaded": True})
def get_bucket(self, store_id: bytes32) -> str:
for store in self.stores:
if store.id == store_id and store.bucket:
return store.bucket
raise Exception(f"bucket not found for store id {store_id.hex()}")
def update_instance_from_config(self) -> None:
config = load_config(self.instance_name)
self.stores = read_store_ids_from_config(config)
def update_config(self) -> None:
with open("s3_plugin_config.yml") as file:
full_config = yaml.safe_load(file)
full_config[self.instance_name]["stores"] = [store.marshal() for store in self.stores]
self.save_config("s3_plugin_config.yml", full_config)
def save_config(self, filename: str, config_data: Any) -> None:
path: Path = Path(filename)
with tempfile.TemporaryDirectory(dir=path.parent) as tmp_dir:
tmp_path: Path = Path(tmp_dir) / Path(filename)
with open(tmp_path, "w") as f:
yaml.safe_dump(config_data, f)
try:
os.replace(str(tmp_path), path)
except PermissionError:
shutil.move(str(tmp_path), str(path))
def read_store_ids_from_config(config: dict[str, Any]) -> list[StoreConfig]:
stores = []
for store in config.get("stores", []):
try:
stores.append(StoreConfig.unmarshal(store))
except Exception as e:
if "store_id" in store:
bad_store_id = f"{store['store_id']!r}"
else:
bad_store_id = "<missing>"
log.info(f"Ignoring invalid store id: {bad_store_id}: {type(e).__name__} {e}")
return stores
def make_app(config: dict[str, Any], instance_name: str) -> web.Application:
try:
region = config["aws_credentials"]["region"]
aws_access_key_id = config["aws_credentials"]["access_key_id"]
aws_secret_access_key = config["aws_credentials"]["secret_access_key"]
server_files_location = config["server_files_location"]
server_files_path = Path(server_files_location).resolve()
except KeyError as e:
sys.exit(
"config file must have server_files_location, aws_credentials with region, access_key_id. "
f", and secret_access_key. Missing config key: {e.args[0]!r}"
)
log_level = config.get("log_level", "INFO")
log.setLevel(log_level)
fh = logging.FileHandler(config.get("log_filename", "s3_plugin.log"))
fh.setLevel(log_level)
# create formatter and add it to the handlers
file_log_formatter = logging.Formatter(
fmt="%(asctime)s.%(msecs)03d %(name)s %(levelname)s %(message)s", datefmt="%Y-%m-%dT%H:%M:%S"
)
fh.setFormatter(file_log_formatter)
# add the handlers to logger
log.addHandler(fh)
stores = read_store_ids_from_config(config)
s3_client = S3Plugin(
region=region,
aws_access_key_id=aws_access_key_id,
aws_secret_access_key=aws_secret_access_key,
server_files_path=server_files_path,
stores=stores,
instance_name=instance_name,
)
app = web.Application()
app.add_routes([web.post("/handle_upload", s3_client.handle_upload)])
app.add_routes([web.post("/upload", s3_client.upload)])
app.add_routes([web.post("/handle_download", s3_client.handle_download)])
app.add_routes([web.post("/download", s3_client.download)])
app.add_routes([web.post("/add_store_id", s3_client.add_store_id)])
app.add_routes([web.post("/remove_store_id", s3_client.remove_store_id)])
app.add_routes([web.post("/add_missing_files", s3_client.add_missing_files)])
app.add_routes([web.post("/plugin_info", s3_client.plugin_info)])
app.add_routes([web.post("/healthz", s3_client.healthz)])
log.info(f"Starting s3 plugin {instance_name} on port {config['port']}")
return app
def load_config(instance: str) -> Any:
with open("s3_plugin_config.yml") as f:
full_config = yaml.safe_load(f)
return full_config[instance]
def run_server() -> None:
instance_name = sys.argv[1]
try:
config = load_config(instance_name)
except KeyError:
sys.exit(f"Config for instance {instance_name} not found.")
if not config:
sys.exit(f"Config for instance {instance_name} is empty.")
try:
port = config["port"]
except KeyError:
sys.exit("Missing port in config file.")
web.run_app(make_app(config, instance_name), port=port, host="localhost")
log.info(f"Stopped s3 plugin {instance_name}")
if __name__ == "__main__":
run_server()