def call_ray_start_with_external_redis(request):
ports = getattr(request, "param", "6379")
port_list = ports.split(",")
for port in port_list:
temp_dir = ray._private.utils.get_ray_temp_dir()
_start_redis_instance(REDIS_EXECUTABLE,
temp_dir,
int(port),
password="******")
address_str = ",".join(map(lambda x: "localhost:" + x, port_list))
cmd = f"ray start --head --address={address_str} --redis-password=123"
subprocess.call(cmd.split(" "))
yield address_str.split(",")[0]
# Disconnect from the Ray cluster.
ray.shutdown()
# Kill the Ray cluster.
subprocess.check_call(["ray", "stop"])
def _setup_redis(request):
# Setup external Redis and env var for initialization.
param = getattr(request, "param", {})
external_redis_ports = param.get("external_redis_ports")
if external_redis_ports is None:
with socket.socket() as s:
s.bind(("", 0))
port = s.getsockname()[1]
external_redis_ports = [port]
else:
del param["external_redis_ports"]
processes = []
for port in external_redis_ports:
temp_dir = ray._private.utils.get_ray_temp_dir()
port, proc = _start_redis_instance(
REDIS_EXECUTABLE,
temp_dir,
port,
password=ray_constants.REDIS_DEFAULT_PASSWORD,
)
processes.append(proc)
wait_for_redis_to_start("127.0.0.1", port,
ray_constants.REDIS_DEFAULT_PASSWORD)
address_str = ",".join(
map(lambda x: f"127.0.0.1:{x}", external_redis_ports))
import os
old_addr = os.environ.get("RAY_REDIS_ADDRESS")
os.environ["RAY_REDIS_ADDRESS"] = address_str
yield
if old_addr is not None:
os.environ["RAY_REDIS_ADDRESS"] = old_addr
else:
del os.environ["RAY_REDIS_ADDRESS"]
for proc in processes:
proc.process.terminate()
def test_calling_start_ray_head(call_ray_stop_only):
# Test that we can call ray start with various command line
# parameters.
# Test starting Ray with a redis port specified.
check_call_ray(["start", "--head", "--port", "0"])
check_call_ray(["stop"])
# Test starting Ray with a node IP address specified.
check_call_ray(
["start", "--head", "--node-ip-address", "127.0.0.1", "--port", "0"])
check_call_ray(["stop"])
# Test starting Ray with a system config parameter set.
check_call_ray([
"start",
"--head",
"--system-config",
'{"metrics_report_interval_ms":100}',
"--port",
"0",
])
check_call_ray(["stop"])
# Test starting Ray with the object manager and node manager ports
# specified.
check_call_ray([
"start",
"--head",
"--object-manager-port",
"22345",
"--node-manager-port",
"54321",
"--port",
"0",
])
check_call_ray(["stop"])
# Test starting Ray with the worker port range specified.
check_call_ray([
"start",
"--head",
"--min-worker-port",
"51000",
"--max-worker-port",
"51050",
"--port",
"0",
])
check_call_ray(["stop"])
# Test starting Ray with a worker port list.
check_call_ray(["start", "--head", "--worker-port-list", "10002,10003"])
check_call_ray(["stop"])
# Test starting Ray with a non-int in the worker port list.
with pytest.raises(subprocess.CalledProcessError):
check_call_ray(["start", "--head", "--worker-port-list", "10002,a"])
check_call_ray(["stop"])
# Test starting Ray with an invalid port in the worker port list.
with pytest.raises(subprocess.CalledProcessError):
check_call_ray(["start", "--head", "--worker-port-list", "100"])
check_call_ray(["stop"])
# Test starting Ray with the number of CPUs specified.
check_call_ray(["start", "--head", "--num-cpus", "2", "--port", "0"])
check_call_ray(["stop"])
# Test starting Ray with the number of GPUs specified.
check_call_ray(["start", "--head", "--num-gpus", "100", "--port", "0"])
check_call_ray(["stop"])
# Test starting Ray with redis shard ports specified.
check_call_ray([
"start", "--head", "--redis-shard-ports", "6380,6381,6382", "--port",
"0"
])
check_call_ray(["stop"])
# Test starting Ray with all arguments specified.
check_call_ray([
"start",
"--head",
"--redis-shard-ports",
"6380,6381,6382",
"--object-manager-port",
"22345",
"--num-cpus",
"2",
"--num-gpus",
"0",
"--resources",
'{"Custom": 1}',
"--port",
"0",
])
check_call_ray(["stop"])
temp_dir = ray._private.utils.get_ray_temp_dir()
# Test starting Ray with RAY_REDIS_ADDRESS env.
_, proc = _start_redis_instance(
REDIS_EXECUTABLE,
temp_dir,
8888,
password=ray_constants.REDIS_DEFAULT_PASSWORD)
os.environ["RAY_REDIS_ADDRESS"] = "127.0.0.1:8888"
check_call_ray(["start", "--head"])
check_call_ray(["stop"])
proc.process.terminate()
del os.environ["RAY_REDIS_ADDRESS"]
# Test --block. Killing a child process should cause the command to exit.
blocked = subprocess.Popen(
["ray", "start", "--head", "--block", "--port", "0"])
blocked.poll()
assert blocked.returncode is None
# Make sure ray cluster is up
run_string_as_driver("""
import ray
from time import sleep
for i in range(0, 5):
try:
ray.init(address='auto')
break
except:
sleep(1)
""")
# Make sure ray cluster is up
run_string_as_driver("""
import ray
from time import sleep
for i in range(0, 5):
try:
ray.init(address='auto')
break
except:
sleep(1)
""")
kill_process_by_name("raylet", SIGKILL=True)
wait_for_children_of_pid_to_exit(blocked.pid, timeout=30)
blocked.wait()
assert blocked.returncode != 0, "ray start shouldn't return 0 on bad exit"
# Test --block. Killing the command should clean up all child processes.
blocked = subprocess.Popen(
["ray", "start", "--head", "--block", "--port", "0"])
blocked.poll()
assert blocked.returncode is None
# Include GCS, autoscaler monitor, client server, dashboard, raylet and
# log_monitor.py
num_children = 6
if not detect_fate_sharing_support():
# Account for ray_process_reaper.py
num_children += 1
# Check a set of child process commands & scripts instead?
wait_for_children_of_pid(blocked.pid,
num_children=num_children,
timeout=30)
blocked.terminate()
wait_for_children_of_pid_to_exit(blocked.pid, timeout=30)
blocked.wait()
assert blocked.returncode != 0, "ray start shouldn't return 0 on bad exit"
