def echo_v1(_):
return "v1"
def echo_v2(_):
return "v2"
serve.init(blocking=True)
serve.create_endpoint("my_endpoint", "/echo")
serve.create_backend(echo_v1, "echo:v1")
serve.link("my_endpoint", "echo:v1")
for _ in range(3):
resp = requests.get("http://127.0.0.1:8000/echo").json()
print(pformat_color_json(resp))
print("...Sleeping for 2 seconds...")
time.sleep(2)
serve.create_backend(echo_v2, "echo:v2")
serve.split("my_endpoint", {"echo:v1": 0.5, "echo:v2": 0.5})
while True:
resp = requests.get("http://127.0.0.1:8000/echo").json()
print(pformat_color_json(resp))
print("...Sleeping for 2 seconds...")
time.sleep(2)
def split(endpoint, traffic):
ray.init(address="auto")
serve.init()
serve.split(endpoint, json.loads(traffic))
from ray.serve.constants import DEFAULT_HTTP_ADDRESS
import requests
import time
import pandas as pd
from tqdm import tqdm
serve.init(blocking=True)
def noop(_):
return ""
serve.create_endpoint("noop", "/noop")
serve.create_backend(noop, "noop")
serve.split("noop", {"noop": 1.0})
url = "{}/noop".format(DEFAULT_HTTP_ADDRESS)
while requests.get(url).status_code == 404:
time.sleep(1)
print("Waiting for noop route to showup.")
latency = []
for _ in tqdm(range(5200)):
start = time.perf_counter()
resp = requests.get(url)
end = time.perf_counter()
latency.append(end - start)
# Remove initial samples
latency = latency[200:]
from ray import serve
import requests
serve.init()
class Counter:
def __init__(self):
self.count = 0
def __call__(self, flask_request):
return {"current_counter": self.count}
serve.create_endpoint("counter", "/counter")
serve.create_backend(Counter, "counter")
serve.split("counter", {"counter": 1.0})
requests.get("http://127.0.0.1:8000/counter").json()
# > {"current_counter": self.count}
