mirror of
https://github.com/vale981/ray
synced 2025-03-09 12:56:46 -04:00
0b1608a546
* Factor out starting Ray processes. * Detect flags through environment variables. * Return ProcessInfo from start_ray_process. * Print valgrind errors at exit. * Test valgrind in travis. * Some valgrind fixes. * Undo raylet monitor change. * Only test plasma store in valgrind.
121 lines
3.4 KiB
Python
121 lines
3.4 KiB
Python
from __future__ import absolute_import
|
|
from __future__ import division
|
|
from __future__ import print_function
|
|
|
|
import json
|
|
import logging
|
|
import pytest
|
|
import time
|
|
|
|
import ray
|
|
import ray.ray_constants as ray_constants
|
|
from ray.test.cluster_utils import Cluster
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
@pytest.fixture
|
|
def start_connected_cluster():
|
|
# Start the Ray processes.
|
|
g = Cluster(
|
|
initialize_head=True,
|
|
connect=True,
|
|
head_node_args={
|
|
"num_cpus": 1,
|
|
"_internal_config": json.dumps({
|
|
"num_heartbeats_timeout": 10
|
|
})
|
|
})
|
|
yield g
|
|
# The code after the yield will run as teardown code.
|
|
ray.shutdown()
|
|
g.shutdown()
|
|
|
|
|
|
@pytest.fixture
|
|
def start_connected_longer_cluster():
|
|
"""Creates a cluster with a longer timeout."""
|
|
g = Cluster(
|
|
initialize_head=True,
|
|
connect=True,
|
|
head_node_args={
|
|
"num_cpus": 1,
|
|
"_internal_config": json.dumps({
|
|
"num_heartbeats_timeout": 20
|
|
})
|
|
})
|
|
yield g
|
|
# The code after the yield will run as teardown code.
|
|
ray.shutdown()
|
|
g.shutdown()
|
|
|
|
|
|
def test_cluster():
|
|
"""Basic test for adding and removing nodes in cluster."""
|
|
g = Cluster(initialize_head=False)
|
|
node = g.add_node()
|
|
node2 = g.add_node()
|
|
assert node.remaining_processes_alive()
|
|
assert node2.remaining_processes_alive()
|
|
g.remove_node(node2)
|
|
g.remove_node(node)
|
|
assert not any(n.any_processes_alive() for n in [node, node2])
|
|
|
|
|
|
def test_shutdown():
|
|
g = Cluster(initialize_head=False)
|
|
node = g.add_node()
|
|
node2 = g.add_node()
|
|
g.shutdown()
|
|
assert not any(n.any_processes_alive() for n in [node, node2])
|
|
|
|
|
|
def test_internal_config(start_connected_longer_cluster):
|
|
"""Checks that the internal configuration setting works.
|
|
|
|
We set the cluster to timeout nodes after 2 seconds of no timeouts. We
|
|
then remove a node, wait for 1 second to check that the cluster is out
|
|
of sync, then wait another 2 seconds (giving 1 second of leeway) to check
|
|
that the client has timed out.
|
|
"""
|
|
cluster = start_connected_longer_cluster
|
|
worker = cluster.add_node()
|
|
cluster.wait_for_nodes()
|
|
|
|
cluster.remove_node(worker)
|
|
time.sleep(1)
|
|
assert ray.global_state.cluster_resources()["CPU"] == 2
|
|
|
|
time.sleep(2)
|
|
assert ray.global_state.cluster_resources()["CPU"] == 1
|
|
|
|
|
|
def test_wait_for_nodes(start_connected_cluster):
|
|
"""Unit test for `Cluster.wait_for_nodes`.
|
|
|
|
Adds 4 workers, waits, then removes 4 workers, waits,
|
|
then adds 1 worker, waits, and removes 1 worker, waits.
|
|
"""
|
|
cluster = start_connected_cluster
|
|
workers = [cluster.add_node() for i in range(4)]
|
|
cluster.wait_for_nodes()
|
|
[cluster.remove_node(w) for w in workers]
|
|
cluster.wait_for_nodes()
|
|
|
|
assert ray.global_state.cluster_resources()["CPU"] == 1
|
|
worker2 = cluster.add_node()
|
|
cluster.wait_for_nodes()
|
|
cluster.remove_node(worker2)
|
|
cluster.wait_for_nodes()
|
|
assert ray.global_state.cluster_resources()["CPU"] == 1
|
|
|
|
|
|
def test_worker_plasma_store_failure(start_connected_cluster):
|
|
cluster = start_connected_cluster
|
|
worker = cluster.add_node()
|
|
cluster.wait_for_nodes()
|
|
# Log monitor doesn't die for some reason
|
|
worker.kill_log_monitor()
|
|
worker.kill_plasma_store()
|
|
worker.all_processes[ray_constants.PROCESS_TYPE_RAYLET][0].process.wait()
|
|
assert not worker.any_processes_alive(), worker.live_processes()
|