chore: import upstream snapshot with attribution

This commit is contained in:
wehub-resource-sync
2026-07-13 13:17:40 +08:00
commit f1825c8ceb
10096 changed files with 2364182 additions and 0 deletions
+46
View File
@@ -0,0 +1,46 @@
import os
import pathlib
import pytest
import ray.dashboard.modules # noqa
from ray.tests.conftest import * # noqa
@pytest.fixture
def enable_test_module():
os.environ["RAY_DASHBOARD_MODULE_TEST"] = "true"
import ray.dashboard.tests
p = pathlib.Path(ray.dashboard.modules.__path__[0]) / "tests" / "__init__.py"
p.touch()
yield
os.environ.pop("RAY_DASHBOARD_MODULE_TEST", None)
p.unlink()
@pytest.fixture
def disable_aiohttp_cache():
os.environ["RAY_DASHBOARD_NO_CACHE"] = "true"
yield
os.environ.pop("RAY_DASHBOARD_NO_CACHE", None)
@pytest.fixture
def small_event_line_limit():
os.environ["EVENT_READ_LINE_LENGTH_LIMIT"] = "1024"
yield 1024
os.environ.pop("EVENT_READ_LINE_LENGTH_LIMIT", None)
@pytest.fixture
def fast_gcs_failure_detection(monkeypatch):
monkeypatch.setenv("RAY_gcs_rpc_server_reconnect_timeout_s", "2")
monkeypatch.setenv("GCS_CHECK_ALIVE_INTERVAL_SECONDS", "1")
@pytest.fixture
def reduce_actor_cache():
os.environ["RAY_maximum_gcs_destroyed_actor_cached_count"] = "3"
yield
os.environ.pop("RAY_maximum_gcs_destroyed_actor_cached_count", None)
@@ -0,0 +1,10 @@
module.exports = {
video: false,
screenshotOnRunFailure: false,
retries: {
runMode: 2,
},
e2e: {
setupNodeEvents(on, config) {},
},
}
@@ -0,0 +1,3 @@
videos
plugins
screenshots
@@ -0,0 +1,9 @@
describe("Ray Dashboard Test", () => {
it("opens a new Ray dashboard", () => {
cy.visit("localhost:8653");
cy.contains("Overview");
cy.contains("Jobs");
cy.contains("Cluster");
cy.contains("Logs");
});
});
@@ -0,0 +1,7 @@
Cypress.on(`window:before:load`, win => {
cy.stub( win.console, `error`, msg => {
// Abort the test on any error in console,
// this is a hack to show full stack trace.
throw new Error( msg.stack );
});
});
+41
View File
@@ -0,0 +1,41 @@
#!/usr/bin/env bash
set -x
set -euo pipefail
ROOT_DIR=$(cd "$(dirname "${BASH_SOURCE:-$0}")"; pwd)
cd "$ROOT_DIR"
clean_up() {
ray stop --force
}
trap clean_up EXIT
CYPRESS_VERSION=14.2.1
(
cd ../client
npm ci
)
echo "Installing cypress"
if [[ -n "$BUILDKITE" ]]; then
apt-get update -qq
apt install -y libgtk2.0-0 libgtk-3-0 libgbm-dev libnotify-dev libgconf-2-4 libnss3 libxss1 libasound2 libxtst6 xauth xvfb
sudo npm install "cypress@$CYPRESS_VERSION"
else
which cypress || npm install "cypress@$CYPRESS_VERSION" -g
fi
ray stop --force
ray start --head --dashboard-port=8653 --dashboard-host=0.0.0.0
sleep 5 # Wait for Ray dashboard to become ready
curl localhost:8653 || cat /tmp/ray/session_latest/logs/dashboard.log
node_modules/.bin/cypress run --project . --headless
# Run frontend UI tests
(
cd ../client
CI=true npm test
)
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,155 @@
"""Tests for dashboard token authentication."""
import sys
import pytest
import requests
def test_dashboard_request_requires_auth_with_valid_token(
setup_cluster_with_token_auth,
):
"""Test that requests succeed with valid token when auth is enabled."""
cluster_info = setup_cluster_with_token_auth
headers = {"Authorization": f"Bearer {cluster_info['token']}"}
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
headers=headers,
)
assert response.status_code == 200
def test_dashboard_request_requires_auth_missing_token(setup_cluster_with_token_auth):
"""Test that requests fail without token when auth is enabled."""
cluster_info = setup_cluster_with_token_auth
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
json={"test": "data"},
)
assert response.status_code == 401
def test_dashboard_request_requires_auth_invalid_token(setup_cluster_with_token_auth):
"""Test that requests fail with invalid token when auth is enabled."""
cluster_info = setup_cluster_with_token_auth
headers = {"Authorization": "Bearer wrong_token_00000000000000000000000000000000"}
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
json={"test": "data"},
headers=headers,
)
assert response.status_code == 403
def test_dashboard_request_with_ray_auth_header(setup_cluster_with_token_auth):
"""Test that requests succeed with valid token in X-Ray-Authorization header."""
cluster_info = setup_cluster_with_token_auth
headers = {"X-Ray-Authorization": f"Bearer {cluster_info['token']}"}
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
headers=headers,
)
assert response.status_code == 200
def test_authorization_header_takes_precedence(setup_cluster_with_token_auth):
"""Test that standard Authorization header takes precedence over X-Ray-Authorization."""
cluster_info = setup_cluster_with_token_auth
# Provide both headers: valid token in Authorization, invalid in X-Ray-Authorization
headers = {
"Authorization": f"Bearer {cluster_info['token']}",
"X-Ray-Authorization": "Bearer invalid_token_000000000000000000000000",
}
# Should succeed because Authorization header takes precedence
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
headers=headers,
)
assert response.status_code == 200
# Now test with invalid Authorization but valid X-Ray-Authorization
headers = {
"Authorization": "Bearer invalid_token_000000000000000000000000",
"X-Ray-Authorization": f"Bearer {cluster_info['token']}",
}
# Should fail because Authorization header takes precedence (even though it's invalid)
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
headers=headers,
)
assert response.status_code == 403
def test_dashboard_auth_disabled(setup_cluster_without_token_auth):
"""Test that auth is not enforced when AUTH_MODE is disabled."""
cluster_info = setup_cluster_without_token_auth
response = requests.get(
f"{cluster_info['dashboard_url']}/api/component_activities",
json={"test": "data"},
)
assert response.status_code == 200
def test_authentication_mode_endpoint_with_token_auth(setup_cluster_with_token_auth):
"""Test authentication_mode endpoint returns 'token' when auth is enabled."""
cluster_info = setup_cluster_with_token_auth
# This endpoint should be accessible WITHOUT authentication
response = requests.get(f"{cluster_info['dashboard_url']}/api/authentication_mode")
assert response.status_code == 200
assert response.json() == {"authentication_mode": "token"}
def test_authentication_mode_endpoint_without_auth(setup_cluster_without_token_auth):
"""Test authentication_mode endpoint returns 'disabled' when auth is off."""
cluster_info = setup_cluster_without_token_auth
response = requests.get(f"{cluster_info['dashboard_url']}/api/authentication_mode")
assert response.status_code == 200
assert response.json() == {"authentication_mode": "disabled"}
def test_authentication_mode_endpoint_is_public(setup_cluster_with_token_auth):
"""Test authentication_mode endpoint works without Authorization header."""
cluster_info = setup_cluster_with_token_auth
# Call WITHOUT any authorization header - should still succeed
response = requests.get(
f"{cluster_info['dashboard_url']}/api/authentication_mode",
headers={}, # Explicitly no auth
)
# Should succeed even with token auth enabled
assert response.status_code == 200
assert response.json() == {"authentication_mode": "token"}
if __name__ == "__main__":
sys.exit(pytest.main(["-vv", __file__]))
@@ -0,0 +1,273 @@
import ray
from ray.dashboard.memory_utils import (
MemoryTable,
MemoryTableEntry,
ReferenceType,
SortingType,
decode_object_ref_if_needed,
)
"""Memory Table Unit Test"""
NODE_ADDRESS = "127.0.0.1"
IS_DRIVER = True
PID = 1
OBJECT_ID = "ZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZg=="
ACTOR_ID = "fffffffffffffffffffffffffffffffff66d17ba010000c801000000"
DECODED_ID = decode_object_ref_if_needed(OBJECT_ID)
OBJECT_SIZE = 100
def build_memory_entry(
*,
local_ref_count,
pinned_in_memory,
submitted_task_reference_count,
contained_in_owned,
object_size,
pid,
object_id=OBJECT_ID,
node_address=NODE_ADDRESS
):
object_ref = {
"objectId": object_id,
"callSite": "(task call) /Users:458",
"objectSize": object_size,
"localRefCount": local_ref_count,
"pinnedInMemory": pinned_in_memory,
"submittedTaskRefCount": submitted_task_reference_count,
"containedInOwned": contained_in_owned,
}
return MemoryTableEntry(
object_ref=object_ref, node_address=node_address, is_driver=IS_DRIVER, pid=pid
)
def build_local_reference_entry(
object_size=OBJECT_SIZE, pid=PID, node_address=NODE_ADDRESS
):
return build_memory_entry(
local_ref_count=1,
pinned_in_memory=False,
submitted_task_reference_count=0,
contained_in_owned=[],
object_size=object_size,
pid=pid,
node_address=node_address,
)
def build_used_by_pending_task_entry(
object_size=OBJECT_SIZE, pid=PID, node_address=NODE_ADDRESS
):
return build_memory_entry(
local_ref_count=0,
pinned_in_memory=False,
submitted_task_reference_count=2,
contained_in_owned=[],
object_size=object_size,
pid=pid,
node_address=node_address,
)
def build_captured_in_object_entry(
object_size=OBJECT_SIZE, pid=PID, node_address=NODE_ADDRESS
):
return build_memory_entry(
local_ref_count=0,
pinned_in_memory=False,
submitted_task_reference_count=0,
contained_in_owned=[OBJECT_ID],
object_size=object_size,
pid=pid,
node_address=node_address,
)
def build_actor_handle_entry(
object_size=OBJECT_SIZE, pid=PID, node_address=NODE_ADDRESS
):
return build_memory_entry(
local_ref_count=1,
pinned_in_memory=False,
submitted_task_reference_count=0,
contained_in_owned=[],
object_size=object_size,
pid=pid,
node_address=node_address,
object_id=ACTOR_ID,
)
def build_pinned_in_memory_entry(
object_size=OBJECT_SIZE, pid=PID, node_address=NODE_ADDRESS
):
return build_memory_entry(
local_ref_count=0,
pinned_in_memory=True,
submitted_task_reference_count=0,
contained_in_owned=[],
object_size=object_size,
pid=pid,
node_address=node_address,
)
def build_entry(
object_size=OBJECT_SIZE,
pid=PID,
node_address=NODE_ADDRESS,
reference_type=ReferenceType.PINNED_IN_MEMORY,
):
if reference_type == ReferenceType.USED_BY_PENDING_TASK:
return build_used_by_pending_task_entry(
pid=pid, object_size=object_size, node_address=node_address
)
elif reference_type == ReferenceType.LOCAL_REFERENCE:
return build_local_reference_entry(
pid=pid, object_size=object_size, node_address=node_address
)
elif reference_type == ReferenceType.PINNED_IN_MEMORY:
return build_pinned_in_memory_entry(
pid=pid, object_size=object_size, node_address=node_address
)
elif reference_type == ReferenceType.ACTOR_HANDLE:
return build_actor_handle_entry(
pid=pid, object_size=object_size, node_address=node_address
)
elif reference_type == ReferenceType.CAPTURED_IN_OBJECT:
return build_captured_in_object_entry(
pid=pid, object_size=object_size, node_address=node_address
)
def test_invalid_memory_entry():
memory_entry = build_memory_entry(
local_ref_count=0,
pinned_in_memory=False,
submitted_task_reference_count=0,
contained_in_owned=[],
object_size=OBJECT_SIZE,
pid=PID,
)
assert memory_entry.is_valid() is False
memory_entry = build_memory_entry(
local_ref_count=0,
pinned_in_memory=False,
submitted_task_reference_count=0,
contained_in_owned=[],
object_size=-1,
pid=PID,
)
assert memory_entry.is_valid() is False
def test_valid_reference_memory_entry():
memory_entry = build_local_reference_entry()
assert memory_entry.reference_type == ReferenceType.LOCAL_REFERENCE.value
assert memory_entry.object_ref == ray.ObjectRef(
decode_object_ref_if_needed(OBJECT_ID)
)
assert memory_entry.is_valid() is True
def test_reference_type():
# pinned in memory
memory_entry = build_pinned_in_memory_entry()
assert memory_entry.reference_type == ReferenceType.PINNED_IN_MEMORY.value
# used by pending task
memory_entry = build_used_by_pending_task_entry()
assert memory_entry.reference_type == ReferenceType.USED_BY_PENDING_TASK.value
# captued in object
memory_entry = build_captured_in_object_entry()
assert memory_entry.reference_type == ReferenceType.CAPTURED_IN_OBJECT.value
# actor handle
memory_entry = build_actor_handle_entry()
assert memory_entry.reference_type == ReferenceType.ACTOR_HANDLE.value
def test_memory_table_summary():
entries = [
build_pinned_in_memory_entry(),
build_used_by_pending_task_entry(),
build_captured_in_object_entry(),
build_actor_handle_entry(),
build_local_reference_entry(),
build_local_reference_entry(),
]
memory_table = MemoryTable(entries)
assert len(memory_table.group) == 1
assert memory_table.summary["total_actor_handles"] == 1
assert memory_table.summary["total_captured_in_objects"] == 1
assert memory_table.summary["total_local_ref_count"] == 2
assert memory_table.summary["total_object_size"] == len(entries) * OBJECT_SIZE
assert memory_table.summary["total_pinned_in_memory"] == 1
assert memory_table.summary["total_used_by_pending_task"] == 1
def test_memory_table_sort_by_pid():
unsort = [1, 3, 2]
entries = [build_entry(pid=pid) for pid in unsort]
memory_table = MemoryTable(entries, sort_by_type=SortingType.PID)
sort = sorted(unsort)
for pid, entry in zip(sort, memory_table.table):
assert pid == entry.pid
def test_memory_table_sort_by_reference_type():
unsort = [
ReferenceType.USED_BY_PENDING_TASK,
ReferenceType.LOCAL_REFERENCE,
ReferenceType.LOCAL_REFERENCE,
ReferenceType.PINNED_IN_MEMORY,
]
entries = [build_entry(reference_type=reference_type) for reference_type in unsort]
memory_table = MemoryTable(entries, sort_by_type=SortingType.REFERENCE_TYPE)
sort = sorted([entry.value for entry in unsort])
for reference_type, entry in zip(sort, memory_table.table):
assert reference_type == entry.reference_type
def test_memory_table_sort_by_object_size():
unsort = [312, 214, -1, 1244, 642]
entries = [build_entry(object_size=object_size) for object_size in unsort]
memory_table = MemoryTable(entries, sort_by_type=SortingType.OBJECT_SIZE)
sort = sorted(unsort)
for object_size, entry in zip(sort, memory_table.table):
assert object_size == entry.object_size
def test_group_by():
node_second = "127.0.0.2"
node_first = "127.0.0.1"
entries = [
build_entry(node_address=node_second, pid=2),
build_entry(node_address=node_second, pid=1),
build_entry(node_address=node_first, pid=2),
build_entry(node_address=node_first, pid=1),
]
memory_table = MemoryTable(entries)
# Make sure it is correctly grouped
assert node_first in memory_table.group
assert node_second in memory_table.group
# make sure pid is sorted in the right order.
for group_key, group_memory_table in memory_table.group.items():
pid = 1
for entry in group_memory_table.table:
assert pid == entry.pid
pid += 1
if __name__ == "__main__":
import sys
import pytest
sys.exit(pytest.main(["-v", __file__]))
@@ -0,0 +1,94 @@
import asyncio
import sys
import pytest
from ray.dashboard.modules.state.state_head import RateLimitedModule
class FailedCallError(Exception):
pass
class A(RateLimitedModule):
def __init__(self, max_num_call: int):
import logging
super().__init__(max_num_call, logging.getLogger(__name__))
@RateLimitedModule.enforce_max_concurrent_calls
async def fn1(self, err: bool = False):
if err:
raise FailedCallError
await asyncio.sleep(3)
return True
@RateLimitedModule.enforce_max_concurrent_calls
async def fn2(self):
await asyncio.sleep(3)
return True
async def limit_handler_(self):
return False
@pytest.mark.asyncio
@pytest.mark.parametrize("extra_req_num", [-5, -3, -1, 0, 1, 3, 5])
async def test_max_concurrent_in_progress_functions(extra_req_num):
"""Test rate limiting for concurrent in-progress requests on StateHead"""
max_req = 10
a = A(max_num_call=max_req)
# Run more than allowed concurrent async functions should trigger rate limiting
res_arr = await asyncio.gather(
*[a.fn1() if i % 2 == 0 else a.fn2() for i in range(max_req + extra_req_num)]
)
fail_cnt = 0
for ok in res_arr:
fail_cnt += 0 if ok else 1
expected_fail_cnt = max(0, extra_req_num)
assert fail_cnt == expected_fail_cnt, (
f"{expected_fail_cnt} out of {max_req + extra_req_num} "
f"concurrent runs should fail with max={max_req} but {fail_cnt}."
)
assert a.num_call_ == 0, "All requests should be done"
@pytest.mark.asyncio
@pytest.mark.parametrize(
"failures",
[
[True, True, True, True, True],
[False, False, False, False, False],
[False, True, False, True, False],
[False, False, False, True, True],
[True, True, False, False, False],
],
)
async def test_max_concurrent_with_exceptions(failures):
max_req = 10
a = A(max_num_call=max_req)
# Run more than allowed concurrent async functions should trigger rate limiting
res_arr = await asyncio.gather(
*[a.fn1(err=should_throw_err) for should_throw_err in failures],
return_exceptions=True,
)
expected_num_failure = sum(failures)
actual_num_failure = 0
for res in res_arr:
if isinstance(res, FailedCallError):
actual_num_failure += 1
assert expected_num_failure == actual_num_failure, "All failures should be captured"
assert a.num_call_ == 0, "Failure should decrement the counter correctly"
if __name__ == "__main__":
sys.exit(pytest.main(["-v", __file__]))
+25
View File
@@ -0,0 +1,25 @@
import logging
import sys
import pytest
from ray.dashboard.utils import close_logger_file_descriptor
def test_close_logger_file_descriptor():
logger_format = "%(message)s"
logger = logging.getLogger("test_job_id")
job_driver_log_path = "/tmp/ray.log"
job_driver_handler = logging.FileHandler(job_driver_log_path)
job_driver_formatter = logging.Formatter(logger_format)
job_driver_handler.setFormatter(job_driver_formatter)
logger.addHandler(job_driver_handler)
assert job_driver_handler.stream.closed is False
close_logger_file_descriptor(logger)
assert job_driver_handler.stream is None
if __name__ == "__main__":
sys.exit(pytest.main(["-v", __file__]))