def rest_prediction_server():
options = {
"bind": "%s:%s" % ("0.0.0.0", http_port),
"accesslog": accesslog(args.access_log),
"loglevel": args.log_level.lower(),
"timeout": 5000,
"threads": threads(args.threads, args.single_threaded),
"workers": args.workers,
"max_requests": args.max_requests,
"max_requests_jitter": args.max_requests_jitter,
"post_worker_init": post_worker_init,
"worker_exit": partial(worker_exit,
seldon_metrics=seldon_metrics),
}
if args.pidfile is not None:
options["pidfile"] = args.pidfile
app = seldon_microservice.get_rest_microservice(
user_object, seldon_metrics)
UserModelApplication(
app,
user_object,
jaeger_extra_tags,
args.interface_name,
options=options,
).run()
def rest_prediction_server():
rest_timeout = DEFAULT_ANNOTATION_REST_TIMEOUT
if ANNOTATION_REST_TIMEOUT in annotations:
# Gunicorn timeout is in seconds so convert as annotation is in miliseconds
rest_timeout = int(annotations[ANNOTATION_REST_TIMEOUT]) / 1000
# Converting timeout from float to int and set to 1 if is 0
rest_timeout = int(rest_timeout) or 1
options = {
"bind": "%s:%s" % ("0.0.0.0", http_port),
"accesslog": accesslog(args.access_log),
"loglevel": args.log_level.lower(),
"timeout": rest_timeout,
"threads": threads(args.threads, args.single_threaded),
"workers": args.workers,
"max_requests": args.max_requests,
"max_requests_jitter": args.max_requests_jitter,
"post_worker_init": post_worker_init,
"worker_exit": partial(worker_exit,
seldon_metrics=seldon_metrics),
"keepalive": args.keepalive,
}
logger.info(f"Gunicorn Config: {options}")
if args.pidfile is not None:
options["pidfile"] = args.pidfile
app = seldon_microservice.get_rest_microservice(
user_object, seldon_metrics)
UserModelApplication(
app,
user_object,
args.tracing,
jaeger_extra_tags,
args.interface_name,
options=options,
).run()