-
Notifications
You must be signed in to change notification settings - Fork 757
/
app.py
513 lines (431 loc) · 19.3 KB
/
app.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
from __future__ import annotations
import asyncio
import functools
import inspect
import math
import typing as t
from http import HTTPStatus
from pathlib import Path
import anyio
import anyio.to_thread
from simple_di import Provide
from simple_di import inject
from starlette.middleware import Middleware
from starlette.responses import Response
from starlette.staticfiles import StaticFiles
from _bentoml_sdk import Service
from _bentoml_sdk.service import set_current_service
from bentoml._internal.container import BentoMLContainer
from bentoml._internal.marshal.dispatcher import CorkDispatcher
from bentoml._internal.resource import system_resources
from bentoml._internal.server.base_app import BaseAppFactory
from bentoml._internal.server.http_app import log_exception
from bentoml._internal.utils.metrics import exponential_buckets
from bentoml.exceptions import BentoMLException
from bentoml.exceptions import ServiceUnavailable
if t.TYPE_CHECKING:
from opentelemetry.sdk.trace import Span
from starlette.applications import Starlette
from starlette.requests import Request
from starlette.routing import BaseRoute
from bentoml._internal import external_typing as ext
from bentoml._internal.context import ServiceContext
from bentoml._internal.types import LifecycleHook
from bentoml.metrics import Histogram
R = t.TypeVar("R")
class ContextMiddleware:
def __init__(self, app: ext.ASGIApp, context: ServiceContext) -> None:
self.app = app
self.context = context
async def __call__(
self, scope: ext.ASGIScope, receive: ext.ASGIReceive, send: ext.ASGISend
) -> None:
from starlette.requests import Request
if scope["type"] not in ("http",):
return await self.app(scope, receive, send)
req = Request(scope, receive, send)
with self.context.in_request(req):
await self.app(scope, receive, send)
class ServiceAppFactory(BaseAppFactory):
@inject
def __init__(
self,
service: Service[t.Any],
is_main: bool = False,
enable_metrics: bool = Provide[
BentoMLContainer.api_server_config.metrics.enabled
],
services: dict[str, t.Any] = Provide[BentoMLContainer.config.services],
# traffic: dict[str, t.Any] = Provide[BentoMLContainer.api_server_config.traffic],
enable_access_control: bool = Provide[BentoMLContainer.http.cors.enabled],
access_control_options: dict[str, list[str] | str | int] = Provide[
BentoMLContainer.access_control_options
],
) -> None:
from bentoml._internal.runner.container import AutoContainer
self.service = service
self.enable_metrics = enable_metrics
self.is_main = is_main
config = services[service.name]
traffic = config.get("traffic")
workers = config.get("workers")
timeout = traffic.get("timeout")
max_concurrency = traffic.get("max_concurrency")
self.enable_access_control = enable_access_control
self.access_control_options = access_control_options
# max_concurrency per worker is the max_concurrency per service divided by the number of workers
num_workers = 1
if workers:
if (workers := config["workers"]) == "cpu_count":
srs = system_resources()
num_workers = int(srs["cpu"])
else: # workers is a number
num_workers = workers
super().__init__(
timeout=timeout,
max_concurrency=max_concurrency
if not max_concurrency
else math.ceil(max_concurrency / num_workers),
)
self.dispatchers: dict[str, CorkDispatcher[t.Any, t.Any]] = {}
self._service_instance: t.Any | None = None
self._limiter: anyio.CapacityLimiter | None = None
def fallback() -> t.NoReturn:
raise ServiceUnavailable("process is overloaded")
for name, method in service.apis.items():
if not method.batchable:
continue
self.dispatchers[name] = CorkDispatcher(
max_latency_in_ms=method.max_latency_ms,
max_batch_size=method.max_batch_size,
fallback=fallback,
get_batch_size=functools.partial(
AutoContainer.get_batch_size, batch_dim=method.batch_dim[0]
),
)
@functools.cached_property
def adaptive_batch_size_hist(self) -> Histogram:
metrics_client = BentoMLContainer.metrics_client.get()
max_max_batch_size = max(
(
method.max_batch_size
for method in self.service.apis.values()
if method.batchable
),
default=100,
)
return metrics_client.Histogram(
namespace="bentoml_service",
name="adaptive_batch_size",
documentation="Service adaptive batch size",
labelnames=[
"runner_name",
"worker_index",
"method_name",
"service_version",
"service_name",
],
buckets=exponential_buckets(1, 2, max_max_batch_size),
)
async def index_page(self, _: Request) -> Response:
from starlette.responses import FileResponse
if BentoMLContainer.new_index:
filename = "main-ui.html"
else:
filename = "main-openapi.html"
return FileResponse(Path(__file__).parent / filename)
async def openapi_spec_view(self, req: Request) -> Response:
from starlette.responses import JSONResponse
try:
return JSONResponse(self.service.openapi_spec.asdict())
except Exception:
log_exception(req)
return JSONResponse(
{"error": "Failed to generate OpenAPI spec"}, status_code=500
)
def __call__(self) -> Starlette:
app = super().__call__()
app.add_route("/schema.json", self.schema_view, name="schema")
for mount_app, path, name in self.service.mount_apps:
app.mount(app=mount_app, path=path, name=name)
if self.is_main:
if BentoMLContainer.new_index:
assets = Path(__file__).parent / "assets"
app.mount("/assets", StaticFiles(directory=assets), name="assets")
else:
from bentoml._internal import server
assets = Path(server.__file__).parent / "static_content"
app.mount(
"/static_content",
StaticFiles(directory=assets),
name="static_content",
)
app.add_route("/docs.json", self.openapi_spec_view, name="openapi-spec")
app.add_route("/", self.index_page, name="index")
return app
@property
def name(self) -> str:
return self.service.name
@property
def middlewares(self) -> list[Middleware]:
from opentelemetry.instrumentation.asgi import OpenTelemetryMiddleware
from bentoml._internal.container import BentoMLContainer
middlewares = super().middlewares + [
Middleware(ContextMiddleware, context=self.service.context)
]
for middleware_cls, options in self.service.middlewares:
middlewares.append(Middleware(middleware_cls, **options))
if self.enable_access_control:
assert (
self.access_control_options.get("allow_origins") is not None
), "To enable cors, access_control_allow_origin must be set"
from starlette.middleware.cors import CORSMiddleware
middlewares.append(
Middleware(CORSMiddleware, **self.access_control_options)
)
def client_request_hook(span: Span | None, _scope: dict[str, t.Any]) -> None:
from bentoml._internal.context import trace_context
if span is not None:
trace_context.request_id = span.context.span_id
middlewares.append(
Middleware(
OpenTelemetryMiddleware,
excluded_urls=BentoMLContainer.tracing_excluded_urls.get(),
default_span_details=None,
server_request_hook=None,
client_request_hook=client_request_hook,
tracer_provider=BentoMLContainer.tracer_provider.get(),
)
)
if self.enable_metrics:
from bentoml._internal.server.http.instruments import (
RunnerTrafficMetricsMiddleware,
)
middlewares.append(
Middleware(RunnerTrafficMetricsMiddleware, namespace="bentoml_service")
)
access_log_config = BentoMLContainer.api_server_config.logging.access
if access_log_config.enabled.get():
from bentoml._internal.server.http.access import AccessLogMiddleware
middlewares.append(
Middleware(
AccessLogMiddleware,
has_request_content_length=access_log_config.request_content_length.get(),
has_request_content_type=access_log_config.request_content_type.get(),
has_response_content_length=access_log_config.response_content_length.get(),
has_response_content_type=access_log_config.response_content_type.get(),
skip_paths=access_log_config.skip_paths.get(),
)
)
return middlewares
def create_instance(self) -> None:
self._service_instance = self.service()
set_current_service(self._service_instance)
def _add_response_headers(self, resp: Response) -> None:
from bentoml._internal.context import trace_context
if trace_context.request_id is not None:
resp.headers["X-BentoML-Request-ID"] = str(trace_context.request_id)
if (
BentoMLContainer.http.response.trace_id.get()
and trace_context.trace_id is not None
):
resp.headers["X-BentoML-Trace-ID"] = str(trace_context.trace_id)
async def destroy_instance(self) -> None:
from _bentoml_sdk.service.dependency import cleanup
# Call on_shutdown hook with optional ctx or context parameter
for name, member in vars(self.service.inner).items():
if callable(member) and getattr(member, "__bentoml_shutdown_hook__", False):
result = getattr(
self._service_instance, name
)() # call the bound method
if inspect.isawaitable(result):
await result
await cleanup()
self._service_instance = None
set_current_service(None)
async def readyz(self, _: Request) -> Response:
from starlette.exceptions import HTTPException
from starlette.responses import PlainTextResponse
from ..client import RemoteProxy
if BentoMLContainer.api_server_config.runner_probe.enabled.get():
dependency_statuses: list[t.Coroutine[None, None, bool]] = []
for dependency in self.service.dependencies.values():
real = dependency.get()
if isinstance(real, RemoteProxy):
dependency_statuses.append(real.is_ready())
runners_ready = all(await asyncio.gather(*dependency_statuses))
if not runners_ready:
raise HTTPException(status_code=503, detail="Runners are not ready.")
return PlainTextResponse("\n", status_code=200)
@property
def on_startup(self) -> list[LifecycleHook]:
return [*super().on_startup, self.create_instance]
@property
def on_shutdown(self) -> list[LifecycleHook]:
return [*super().on_shutdown, self.destroy_instance]
async def schema_view(self, request: Request) -> Response:
from starlette.responses import JSONResponse
schema = self.service.schema()
return JSONResponse(schema)
@property
def routes(self) -> list[BaseRoute]:
from starlette.routing import Route
routes = super().routes
for name, method in self.service.apis.items():
api_endpoint = functools.partial(self.api_endpoint_wrapper, name)
route_path = method.route
if not route_path.startswith("/"):
route_path = "/" + route_path
routes.append(Route(route_path, api_endpoint, methods=["POST"], name=name))
return routes
async def _to_thread(
self,
func: t.Callable[..., R],
*args: t.Any,
**kwargs: t.Any,
) -> R:
if self._limiter is None:
threads = self.service.config.get("threads", 1)
self._limiter = anyio.CapacityLimiter(threads)
func = functools.partial(func, *args, **kwargs)
output = await anyio.to_thread.run_sync(func, limiter=self._limiter)
return output
async def batch_infer(
self, name: str, input_args: tuple[t.Any, ...], input_kwargs: dict[str, t.Any]
) -> t.Any:
method = self.service.apis[name]
func = getattr(self._service_instance, name)
async def inner_infer(
batches: t.Sequence[t.Any], **kwargs: t.Any
) -> t.Sequence[t.Any]:
from bentoml._internal.context import server_context
from bentoml._internal.runner.container import AutoContainer
from bentoml._internal.utils import is_async_callable
if self.enable_metrics:
self.adaptive_batch_size_hist.labels( # type: ignore
runner_name=self.service.name,
worker_index=server_context.worker_index,
method_name=name,
service_version=server_context.bento_version,
service_name=server_context.bento_name,
).observe(len(batches))
if len(batches) == 0:
return []
batch, indices = AutoContainer.batches_to_batch(
batches, method.batch_dim[0]
)
if is_async_callable(func):
result = await func(batch, **kwargs)
else:
result = await self._to_thread(func, batch, **kwargs)
return AutoContainer.batch_to_batches(result, indices, method.batch_dim[1])
arg_names = [k for k in input_kwargs if k not in ("ctx", "context")]
if input_args:
if len(input_args) > 1 or len(arg_names) > 0:
raise TypeError("Batch inference function only accept one argument")
value = input_args[0]
else:
if len(arg_names) != 1:
raise TypeError("Batch inference function only accept one argument")
value = input_kwargs.pop(arg_names[0])
return await self.dispatchers[name](
functools.partial(inner_infer, **input_kwargs)
)(value)
async def api_endpoint_wrapper(self, name: str, request: Request) -> Response:
from pydantic import ValidationError
from starlette.responses import JSONResponse
try:
resp = await self.api_endpoint(name, request)
except ValidationError as exc:
log_exception(request)
data = {
"error": f"{exc.error_count()} validation error for {exc.title}",
"detail": exc.errors(include_context=False),
}
resp = JSONResponse(data, status_code=400)
except BentoMLException as exc:
log_exception(request)
status = exc.error_code.value
if status in (401, 403):
detail = {
"error": "Authorization error",
}
elif status >= 500:
detail = {
"error": "An unexpected error has occurred, please check the server log."
}
else:
detail = ({"error": str(exc)},)
resp = JSONResponse(detail, status_code=status)
except Exception:
log_exception(request)
resp = JSONResponse(
{
"error": "An unexpected error has occurred, please check the server log."
},
status_code=500,
)
self._add_response_headers(resp)
return resp
async def api_endpoint(self, name: str, request: Request) -> Response:
from starlette.background import BackgroundTask
from _bentoml_sdk.io_models import ARGS
from _bentoml_sdk.io_models import KWARGS
from bentoml._internal.utils import get_original_func
from bentoml._internal.utils.http import set_cookies
from ..serde import ALL_SERDE
media_type = request.headers.get("Content-Type", "application/json")
media_type = media_type.split(";")[0].strip()
if self.is_main and media_type == "application/vnd.bentoml+pickle":
# Disallow pickle media type for main service for security reasons
raise BentoMLException(
"Pickle media type is not allowed for main service",
error_code=HTTPStatus.UNSUPPORTED_MEDIA_TYPE,
)
method = self.service.apis[name]
func = getattr(self._service_instance, name)
ctx = self.service.context
serde = ALL_SERDE[media_type]()
input_data = await method.input_spec.from_http_request(request, serde)
input_args: tuple[t.Any, ...] = ()
input_params = {k: getattr(input_data, k) for k in input_data.model_fields}
if method.ctx_param is not None:
input_params[method.ctx_param] = ctx
if ARGS in input_params:
input_args = tuple(input_params.pop(ARGS))
if KWARGS in input_params:
input_params.update(input_params.pop(KWARGS))
original_func = get_original_func(func)
if method.batchable:
output = await self.batch_infer(name, input_args, input_params)
elif inspect.iscoroutinefunction(original_func):
output = await func(*input_args, **input_params)
elif inspect.isasyncgenfunction(original_func):
output = func(*input_args, **input_params)
elif inspect.isgeneratorfunction(original_func):
async def inner() -> t.AsyncGenerator[t.Any, None]:
gen = func(*input_args, **input_params)
while True:
try:
yield await self._to_thread(next, gen)
except StopIteration:
break
except RuntimeError as e:
if "StopIteration" in str(e):
break
raise
output = inner()
else:
output = await self._to_thread(func, *input_args, **input_params)
if isinstance(output, Response):
response = output
else:
response = await method.output_spec.to_http_response(output, serde)
response.headers.update({"Server": f"BentoML Service/{self.service.name}"})
if method.ctx_param is not None:
response.status_code = ctx.response.status_code
response.headers.update(ctx.response.metadata)
set_cookies(response, ctx.response.cookies)
# clean the request resources after the response is consumed.
response.background = BackgroundTask(request.close)
return response