hiro/ray
1
0
Fork 0
mirror of https://github.com/vale981/ray synced 2025-03-06 10:31:39 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

[Serve] Bump test_cluster from small to medium (#20942)

Browse source
This commit is contained in:
architkulkarni 2021-12-08 09:58:27 -08:00 • committed by GitHub
parent 5b168a1515
commit 78cd377775
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 13 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
python/ray/serve/BUILD
View file

@ -224,7 +224,7 @@ py_test(
py_test( py_test(
name = "test_cluster", name = "test_cluster",
size = "small", size = "medium",
srcs = serve_tests_srcs, srcs = serve_tests_srcs,
tags = ["exclusive", "team:serve"], tags = ["exclusive", "team:serve"],
deps = [":serve_lib"], deps = [":serve_lib"],

17
python/ray/serve/tests/test_cluster.py
View file

@ -126,16 +126,16 @@ def test_replica_startup_status_transitions(ray_cluster):
signal = SignalActor.remote() signal = SignalActor.remote()
@serve.deployment(version="1", ray_actor_options={"num_cpus": 2}) @serve.deployment(version="1", ray_actor_options={"num_cpus": 2})
class D: class E:
def __init__(self): def __init__(self):
ray.get(signal.wait.remote()) ray.get(signal.wait.remote())
D.deploy(_blocking=False) E.deploy(_blocking=False)
def get_replicas(replica_state): def get_replicas(replica_state):
controller = serve_instance._controller controller = serve_instance._controller
replicas = ray.get( replicas = ray.get(
controller._dump_replica_states_for_testing.remote(D.name)) controller._dump_replica_states_for_testing.remote(E.name))
return replicas.get([replica_state]) return replicas.get([replica_state])
# wait for serve to start the replica, and catch a reference to it. # wait for serve to start the replica, and catch a reference to it.
@ -152,8 +152,15 @@ def test_replica_startup_status_transitions(ray_cluster):
# add the necessary resources to allocate the replica # add the necessary resources to allocate the replica
cluster.add_node(num_cpus=4) cluster.add_node(num_cpus=4)
wait_for_condition( wait_for_condition(lambda: (ray.cluster_resources().get("CPU", 0) >= 4))
lambda: (replica.check_started() == PENDING_INITIALIZATION)) wait_for_condition(lambda: (ray.available_resources().get("CPU", 0) >= 2))
def is_replica_pending_initialization():
status = replica.check_started()
print(status)
return status == PENDING_INITIALIZATION
wait_for_condition(is_replica_pending_initialization)
# send signal to complete replica intialization # send signal to complete replica intialization
signal.send.remote() signal.send.remote()