wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-13 15:03:44 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Core] Run Plasma Store as a Raylet thread (with a feature flag) (#8897)

Browse Source 
* integrate plasma store as a thread (C++)

* integrate plasma store as a thread (Python)

* fix config issues

* remove plasma component fail tests

* without forcefully kill the plasma store thread
This commit is contained in:
Siyuan (Ryans) Zhuang
2020-06-11 22:54:08 -07:00
committed by GitHub
parent dfa4768fc6
commit 4b31b383f3
11 changed files with 123 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/tests/test_component_failures_2.py
-23
View File
@@ -153,29 +153,6 @@ def test_raylet_failed(ray_start_cluster):
True)
@pytest.mark.skipif(
os.environ.get("RAY_USE_NEW_GCS") == "on",
reason="Hanging with new GCS API.")
@pytest.mark.parametrize(
"ray_start_cluster", [{
"num_cpus": 8,
"num_nodes": 2,
"_internal_config": json.dumps({
"num_heartbeats_timeout": 100
}),
}],
indirect=True)
def test_plasma_store_failed(ray_start_cluster):
cluster = ray_start_cluster
# Kill all plasma stores on worker nodes.
_test_component_failed(cluster, ray_constants.PROCESS_TYPE_PLASMA_STORE)
# No processes should be left alive on the worker nodes.
check_components_alive(cluster, ray_constants.PROCESS_TYPE_PLASMA_STORE,
False)
check_components_alive(cluster, ray_constants.PROCESS_TYPE_RAYLET, False)
if __name__ == "__main__":
import pytest
sys.exit(pytest.main(["-v", __file__]))
python/ray/tests/test_multi_node_2.py
-12
View File
@@ -184,18 +184,6 @@ def test_wait_for_nodes(ray_start_cluster_head):
assert ray.cluster_resources()["CPU"] == 1
def test_worker_plasma_store_failure(ray_start_cluster_head):
cluster = ray_start_cluster_head
worker = cluster.add_node()
cluster.wait_for_nodes()
worker.kill_reporter()
worker.kill_plasma_store()
if ray_constants.PROCESS_TYPE_REAPER in worker.all_processes:
worker.kill_reaper()
worker.all_processes[ray_constants.PROCESS_TYPE_RAYLET][0].process.wait()
assert not worker.any_processes_alive(), worker.live_processes()
if __name__ == "__main__":
import pytest
import sys
python/ray/tests/test_multinode_failures.py
-25
View File
@@ -161,31 +161,6 @@ def test_raylet_failed(ray_start_cluster):
True)
@pytest.mark.skipif(
os.environ.get("RAY_USE_NEW_GCS") == "on",
reason="Hanging with new GCS API.")
@pytest.mark.parametrize(
"ray_start_cluster",
[{
"num_cpus": 8,
"num_nodes": 2,
"_internal_config": json.dumps({
# Raylet codepath is not stable with a shorter timeout.
"num_heartbeats_timeout": 10
}),
}],
indirect=True)
def test_plasma_store_failed(ray_start_cluster):
cluster = ray_start_cluster
# Kill all plasma stores on worker nodes.
_test_component_failed(cluster, ray_constants.PROCESS_TYPE_PLASMA_STORE)
# No processes should be left alive on the worker nodes.
check_components_alive(cluster, ray_constants.PROCESS_TYPE_PLASMA_STORE,
False)
check_components_alive(cluster, ray_constants.PROCESS_TYPE_RAYLET, False)
if __name__ == "__main__":
import pytest
sys.exit(pytest.main(["-v", __file__]))