feat(chv-muse-p1): 常驻 VM + WarmPool 扶正 + 气球 + vsock，对齐 Muse
每用户一台常驻 VM（暂停态，resume 到 SSH 5 秒内为待验证目标）， 池空/失败回落冷启动；CHV 配置加 virtio-balloon + free-page-reporting 与 vsock notify 最小实现（SSH 仍是命令通道与就绪判定）。 新模块 resident_pool.py / vsock.py + 客内 vsock_notify.py； _config 增加 CHV_RESIDENT_IDLE_MINUTES / CHV_BALLOON / CHV_VSOCK； worker_adapter 接常驻孤儿清理 facade；ADR-17 差距表见 docs/chv-muse-p1-gap.md。 恢复收尾（分离后回归，见 gap.md §4）：VmDeathTest 死亡模拟改杀 实际跑命令的 VM；两处冷路径旧测试显式关常驻；_balloon_vsock_payload 与 listen_notify 拆到 50 行内；resident_pool.py:208 mypy 修掉。 executor.py out 未注解与 marker_tools.py:142 是基线旧错，未动。 测试：全量 cloud 套件 1934 OK（skipped=8）；mypy 在改动文件上 干净（仅基线旧错 4 处）。真机验证待第三棒让出 WSL 测试机。
This commit is contained in:
host committed
1 parent
06c3e172fe
commit
78098c5594
13 files changed
+2527
-36
No files matched your search
@@ -0,0 +1,423 @@
|
||||
"""P1-B（feat-chv-muse-p1）：气球/vsock 配置、payload 与 vsock 编解码测试。
|
||||
|
||||
覆盖三件：
|
||||
|
||||
- config：CHV_BALLOON/CHV_VSOCK 开关默认开、只认 0/1，
|
||||
CHV_VSOCK_CID 默认 3 且拒绝保留 CID（0/1/2）与非法值；
|
||||
- executor：build_vm_config 默认带 balloon/vsock 设备，开关
|
||||
关时对应键整个不出现，vsock socket 路径按任务盘派生；
|
||||
- vsock（宿主侧）与客内脚本：JSON 行编解码、监听/超时、
|
||||
客内脚本静默失败（连不上宿主即 exit 0，不影响 sshd/任务）。
|
||||
|
||||
全部用假执行器与本地 Unix socket，绝不连真实网络、不起
|
||||
真实 VM。CHV 54 的 balloon size=0 与 vsock 端到端收发尚未
|
||||
真机验证（本机无 /dev/kvm，第三棒占着 WSL 测试机），见
|
||||
汇报的待验证项——这里只钉住配置层 + payload 形态。
|
||||
"""
|
||||
|
||||
import ast
|
||||
import importlib.util
|
||||
import json
|
||||
import os
|
||||
import socket
|
||||
import sys
|
||||
import tempfile
|
||||
import threading
|
||||
import time
|
||||
import unittest
|
||||
from pathlib import Path
|
||||
|
||||
from cloud.chv import vsock as chv_vsock
|
||||
from cloud.chv.config import _balloon_vsock_fragment
|
||||
from cloud.chv.config import _base_fragment
|
||||
from cloud.chv.config import _disk_fragment
|
||||
from cloud.chv.config import _network_fragment
|
||||
from cloud.chv.config import _quota_fragment
|
||||
from cloud.chv.config import _ssh_fragment
|
||||
from cloud.chv.config import _timeout_fragment
|
||||
from cloud.chv.config import build_chv_config
|
||||
from cloud.chv.executor import CloudHypervisorExecutor
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[2]
|
||||
GUEST_SCRIPT = ROOT / "cloud" / "chv" / "guest" / "vsock_notify.py"
|
||||
|
||||
|
||||
def _load_guest_module():
|
||||
spec = importlib.util.spec_from_file_location(
|
||||
"vsock_notify_guest", str(GUEST_SCRIPT))
|
||||
module = importlib.util.module_from_spec(spec)
|
||||
spec.loader.exec_module(module)
|
||||
return module
|
||||
|
||||
|
||||
def _stdlib_only(path):
|
||||
tree = ast.parse(Path(path).read_text())
|
||||
mods = set()
|
||||
for node in ast.walk(tree):
|
||||
if isinstance(node, ast.Import):
|
||||
mods.update(a.name.split(".")[0] for a in node.names)
|
||||
elif isinstance(node, ast.ImportFrom) and node.module:
|
||||
mods.add(node.module.split(".")[0])
|
||||
return mods <= set(sys.stdlib_module_names)
|
||||
|
||||
|
||||
class BalloonVsockConfigTest(unittest.TestCase):
|
||||
def test_defaults_on(self):
|
||||
cfg = build_chv_config({}, {})
|
||||
self.assertTrue(cfg.balloon)
|
||||
self.assertTrue(cfg.vsock)
|
||||
self.assertEqual(cfg.vsock_cid, 3)
|
||||
|
||||
def test_switches_off(self):
|
||||
cfg = build_chv_config(
|
||||
{}, {"CHV_BALLOON": "0", "CHV_VSOCK": "0"})
|
||||
self.assertFalse(cfg.balloon)
|
||||
self.assertFalse(cfg.vsock)
|
||||
self.assertEqual(cfg.vsock_cid, 3)
|
||||
|
||||
def test_switches_on_explicit(self):
|
||||
cfg = build_chv_config(
|
||||
{}, {"CHV_BALLOON": "1", "CHV_VSOCK": "1"})
|
||||
self.assertTrue(cfg.balloon)
|
||||
self.assertTrue(cfg.vsock)
|
||||
|
||||
def test_illegal_switch_values_raise(self):
|
||||
for env in ({"CHV_BALLOON": "2"}, {"CHV_BALLOON": "yes"},
|
||||
{"CHV_BALLOON": "true"}, {"CHV_VSOCK": "on"},
|
||||
{"CHV_VSOCK": ""}):
|
||||
with self.assertRaises(ValueError):
|
||||
build_chv_config({}, env)
|
||||
|
||||
def test_vsock_cid_override(self):
|
||||
cfg = build_chv_config({}, {"CHV_VSOCK_CID": "7"})
|
||||
self.assertEqual(cfg.vsock_cid, 7)
|
||||
cfg2 = build_chv_config({}, {"CHV_VSOCK_CID": "4294967295"})
|
||||
self.assertEqual(cfg2.vsock_cid, 4294967295)
|
||||
|
||||
def test_vsock_cid_reserved_and_illegal_raise(self):
|
||||
# 0/1/2 是 vsock 保留 CID（Hypervisor/Loopback/Host），
|
||||
# -1 随机 CID 也被 CHV vm.create 400 拒绝。
|
||||
for cid in ("0", "1", "2", "-1", "4294967296", "abc", ""):
|
||||
with self.assertRaises(ValueError, msg=cid):
|
||||
build_chv_config({}, {"CHV_VSOCK_CID": cid})
|
||||
|
||||
def test_fragment_keys_disjoint(self):
|
||||
frag = _balloon_vsock_fragment({})
|
||||
self.assertEqual(set(frag), {"balloon", "vsock", "vsock_cid"})
|
||||
others = [_base_fragment({}, {}), _ssh_fragment({}, {}),
|
||||
_network_fragment({}, {}), _quota_fragment({}),
|
||||
_disk_fragment({}), _timeout_fragment({}, {})]
|
||||
seen = set()
|
||||
for other in others:
|
||||
seen |= set(other)
|
||||
self.assertFalse(seen & set(frag), "新片段键与旧片段冲突")
|
||||
|
||||
|
||||
class _PayloadCase(unittest.TestCase):
|
||||
def setUp(self):
|
||||
self.dir = tempfile.mkdtemp(prefix="chv-p1b-")
|
||||
self.state_dir = os.path.join(self.dir, "state")
|
||||
os.makedirs(self.state_dir)
|
||||
|
||||
def tearDown(self):
|
||||
import shutil
|
||||
shutil.rmtree(self.dir, ignore_errors=True)
|
||||
|
||||
def make_exec(self, env_extra=None):
|
||||
env = dict(env_extra or {})
|
||||
return CloudHypervisorExecutor(
|
||||
kernel=os.path.join(self.dir, "vmlinux"),
|
||||
rootfs=os.path.join(self.dir, "root.raw"),
|
||||
ssh_key=os.path.join(self.dir, "id_ed25519"),
|
||||
state_dir=self.state_dir,
|
||||
runner=lambda cmd, timeout=None: None,
|
||||
launcher=lambda cmd: None,
|
||||
api_client=lambda *a: (200, {}),
|
||||
sleeper=lambda s: None, env=env)
|
||||
|
||||
|
||||
class BalloonVsockPayloadTest(_PayloadCase):
|
||||
def test_default_payload_has_balloon_and_vsock(self):
|
||||
ex = self.make_exec()
|
||||
cfg = ex.build_vm_config()
|
||||
# CHV 54 BalloonConfig：size 是目标气球尺寸（字节），
|
||||
# 0 = 启动不预扣，只靠 free_page_reporting 被动回收。
|
||||
self.assertEqual(cfg["balloon"], {
|
||||
"size": 0, "deflate_on_oom": True,
|
||||
"free_page_reporting": True})
|
||||
self.assertEqual(cfg["vsock"], {
|
||||
"cid": 3,
|
||||
"socket": os.path.join(self.state_dir, "vsock.sock")})
|
||||
|
||||
def test_switches_off_drop_keys(self):
|
||||
ex = self.make_exec({"CHV_BALLOON": "0", "CHV_VSOCK": "0"})
|
||||
cfg = ex.build_vm_config()
|
||||
self.assertNotIn("balloon", cfg)
|
||||
self.assertNotIn("vsock", cfg)
|
||||
# 旧键原样保留。
|
||||
self.assertIn("cpus", cfg)
|
||||
self.assertIn("memory", cfg)
|
||||
self.assertIn("payload", cfg)
|
||||
self.assertIn("disks", cfg)
|
||||
self.assertIn("rng", cfg)
|
||||
|
||||
def test_balloon_off_only(self):
|
||||
ex = self.make_exec({"CHV_BALLOON": "0"})
|
||||
cfg = ex.build_vm_config()
|
||||
self.assertNotIn("balloon", cfg)
|
||||
self.assertIn("vsock", cfg)
|
||||
|
||||
def test_vsock_cid_honored(self):
|
||||
ex = self.make_exec({"CHV_VSOCK_CID": "9"})
|
||||
cfg = ex.build_vm_config()
|
||||
self.assertEqual(cfg["vsock"]["cid"], 9)
|
||||
|
||||
def test_vsock_socket_derived_from_task_disk_dir(self):
|
||||
ex = self.make_exec()
|
||||
disk = os.path.join(self.state_dir, "chv-t1", "disk.raw")
|
||||
cfg = ex.build_vm_config(disk)
|
||||
self.assertEqual(
|
||||
cfg["vsock"]["socket"],
|
||||
os.path.join(self.state_dir, "chv-t1", "vsock.sock"))
|
||||
|
||||
def test_vsock_socket_falls_back_when_disk_outside_state(self):
|
||||
# reuse/lab：任务盘即基础镜像（state_dir 外），socket
|
||||
# 落 state_dir/vsock.sock（单 VM 可用，见 helper 注释）。
|
||||
ex = self.make_exec()
|
||||
cfg = ex.build_vm_config()
|
||||
self.assertEqual(
|
||||
cfg["vsock"]["socket"],
|
||||
os.path.join(self.state_dir, "vsock.sock"))
|
||||
|
||||
def test_explicit_vsock_socket_wins(self):
|
||||
ex = self.make_exec()
|
||||
cfg = ex.build_vm_config(
|
||||
vsock_socket="/tmp/per-vm/v.sock")
|
||||
self.assertEqual(cfg["vsock"]["socket"], "/tmp/per-vm/v.sock")
|
||||
|
||||
def test_helper_direct(self):
|
||||
ex = self.make_exec({"CHV_BALLOON": "0"})
|
||||
extra = ex._balloon_vsock_payload(
|
||||
vsock_socket="/tmp/s.sock")
|
||||
self.assertEqual(extra, {"vsock": {"cid": 3,
|
||||
"socket": "/tmp/s.sock"}})
|
||||
|
||||
|
||||
class VsockCodecTest(unittest.TestCase):
|
||||
def test_encode_decode_round_trip(self):
|
||||
raw = chv_vsock.encode_notify("ready", "chv-t1", ts=1234.5)
|
||||
self.assertTrue(raw.endswith(b"\n"))
|
||||
msg = chv_vsock.decode_notify(raw)
|
||||
self.assertEqual(msg, {"type": "ready", "vm": "chv-t1",
|
||||
"ts": 1234.5})
|
||||
|
||||
def test_encode_rejects_bad_type_or_vm(self):
|
||||
with self.assertRaises(ValueError):
|
||||
chv_vsock.encode_notify("bogus", "chv-t1")
|
||||
with self.assertRaises(ValueError):
|
||||
chv_vsock.encode_notify("ready", "")
|
||||
|
||||
def test_decode_rejects_garbage(self):
|
||||
cases = [b"not json\n", b"[1, 2]\n", b"123\n", b"\n",
|
||||
b'{"type": "bogus", "vm": "v", "ts": 1}\n',
|
||||
b'{"type": "ready", "vm": "", "ts": 1}\n',
|
||||
b'{"type": "ready", "vm": 7, "ts": 1}\n',
|
||||
b'{"type": "ready", "vm": "v"}\n',
|
||||
b'{"type": "ready", "vm": "v", "ts": "x"}\n',
|
||||
b'{"type": "ready", "vm": "v", "ts": true}\n',
|
||||
b'\xff\xfe bad utf8\n']
|
||||
for line in cases:
|
||||
self.assertIsNone(chv_vsock.decode_notify(line), line)
|
||||
|
||||
def test_decode_accepts_int_ts(self):
|
||||
msg = chv_vsock.decode_notify(
|
||||
b'{"type": "heartbeat", "vm": "v", "ts": 7}')
|
||||
self.assertEqual(msg["ts"], 7)
|
||||
|
||||
def test_paths(self):
|
||||
self.assertEqual(
|
||||
chv_vsock.vsock_socket_path("/s", "chv-t1"),
|
||||
"/s/chv-t1/vsock.sock")
|
||||
self.assertEqual(
|
||||
chv_vsock.notify_socket_path("/s/chv-t1/vsock.sock"),
|
||||
"/s/chv-t1/vsock.sock_1024")
|
||||
self.assertEqual(chv_vsock.HOST_CID, 2)
|
||||
self.assertEqual(chv_vsock.VSOCK_NOTIFY_PORT, 1024)
|
||||
|
||||
|
||||
def _wait_for_socket(path, timeout=2.0):
|
||||
deadline = time.monotonic() + timeout
|
||||
while time.monotonic() < deadline:
|
||||
if os.path.exists(path):
|
||||
return True
|
||||
time.sleep(0.02)
|
||||
return False
|
||||
|
||||
|
||||
class VsockListenTest(unittest.TestCase):
|
||||
def setUp(self):
|
||||
self.dir = tempfile.mkdtemp(prefix="chv-p1b-vsock-")
|
||||
|
||||
def tearDown(self):
|
||||
import shutil
|
||||
shutil.rmtree(self.dir, ignore_errors=True)
|
||||
|
||||
def _client(self, path, lines, delay=0.05):
|
||||
"""后台线程：等 socket 出现，连上发 lines 后关闭。
|
||||
|
||||
connect 带重试：监听方 bind 与 listen 之间、以及删
|
||||
残留文件重 bind 的窗口里，单次 connect 可能被拒；
|
||||
须重试到成功或超时，否则测试会偶发失败。"""
|
||||
def run():
|
||||
if not _wait_for_socket(path):
|
||||
return
|
||||
deadline = time.monotonic() + 5
|
||||
sock = None
|
||||
while time.monotonic() < deadline:
|
||||
sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
|
||||
try:
|
||||
sock.connect(path)
|
||||
break
|
||||
except OSError:
|
||||
try:
|
||||
sock.close()
|
||||
except OSError:
|
||||
pass
|
||||
sock = None
|
||||
time.sleep(0.02)
|
||||
if sock is None:
|
||||
return
|
||||
try:
|
||||
time.sleep(delay)
|
||||
for line in lines:
|
||||
sock.sendall(line)
|
||||
except OSError:
|
||||
pass
|
||||
finally:
|
||||
try:
|
||||
sock.close()
|
||||
except OSError:
|
||||
pass
|
||||
thread = threading.Thread(target=run, daemon=True)
|
||||
thread.start()
|
||||
return thread
|
||||
|
||||
def test_listen_returns_ready(self):
|
||||
path = os.path.join(self.dir, "n.sock_1024")
|
||||
thread = self._client(
|
||||
path, [chv_vsock.encode_notify("ready", "chv-t1")])
|
||||
msg = chv_vsock.listen_notify(path, timeout=5)
|
||||
thread.join(timeout=5)
|
||||
self.assertIsNotNone(msg)
|
||||
self.assertEqual(msg["type"], "ready")
|
||||
self.assertEqual(msg["vm"], "chv-t1")
|
||||
self.assertFalse(os.path.exists(path), "监听完须清 socket 文件")
|
||||
|
||||
def test_listen_timeout_returns_none(self):
|
||||
path = os.path.join(self.dir, "t.sock_1024")
|
||||
start = time.monotonic()
|
||||
self.assertIsNone(chv_vsock.listen_notify(path, timeout=0.3))
|
||||
self.assertLess(time.monotonic() - start, 2)
|
||||
self.assertFalse(os.path.exists(path))
|
||||
|
||||
def test_listen_skips_invalid_lines_and_wrong_type(self):
|
||||
path = os.path.join(self.dir, "w.sock_1024")
|
||||
thread = self._client(path, [
|
||||
b"not json\n",
|
||||
chv_vsock.encode_notify("heartbeat", "chv-t1"),
|
||||
chv_vsock.encode_notify("ready", "chv-t1"),
|
||||
])
|
||||
msg = chv_vsock.listen_notify(path, timeout=5,
|
||||
want_type="ready")
|
||||
thread.join(timeout=5)
|
||||
self.assertIsNotNone(msg)
|
||||
self.assertEqual(msg["type"], "ready")
|
||||
|
||||
def test_listen_survives_stale_socket_file(self):
|
||||
path = os.path.join(self.dir, "s.sock_1024")
|
||||
with open(path, "w") as fh:
|
||||
fh.write("stale")
|
||||
thread = self._client(
|
||||
path, [chv_vsock.encode_notify("ready", "chv-t1")])
|
||||
msg = chv_vsock.listen_notify(path, timeout=5)
|
||||
thread.join(timeout=5)
|
||||
self.assertIsNotNone(msg)
|
||||
self.assertEqual(msg["type"], "ready")
|
||||
|
||||
def test_wait_ready_end_to_end(self):
|
||||
vm_dir = os.path.join(self.dir, "chv-t1")
|
||||
os.makedirs(vm_dir)
|
||||
path = chv_vsock.notify_socket_path(
|
||||
chv_vsock.vsock_socket_path(self.dir, "chv-t1"))
|
||||
thread = self._client(
|
||||
path, [chv_vsock.encode_notify("ready", "chv-t1")])
|
||||
msg = chv_vsock.wait_ready(self.dir, "chv-t1", timeout=5)
|
||||
thread.join(timeout=5)
|
||||
self.assertIsNotNone(msg)
|
||||
self.assertEqual(msg["type"], "ready")
|
||||
self.assertEqual(msg["vm"], "chv-t1")
|
||||
|
||||
def test_wait_ready_timeout_returns_none(self):
|
||||
vm_dir = os.path.join(self.dir, "chv-t2")
|
||||
os.makedirs(vm_dir)
|
||||
self.assertIsNone(
|
||||
chv_vsock.wait_ready(self.dir, "chv-t2", timeout=0.3))
|
||||
|
||||
|
||||
class GuestNotifyScriptTest(unittest.TestCase):
|
||||
def test_script_exists_and_stdlib_only(self):
|
||||
self.assertTrue(GUEST_SCRIPT.is_file())
|
||||
self.assertTrue(_stdlib_only(GUEST_SCRIPT))
|
||||
self.assertTrue(
|
||||
_stdlib_only(ROOT / "cloud" / "chv" / "vsock.py"))
|
||||
|
||||
def test_vm_name_sources(self):
|
||||
mod = _load_guest_module()
|
||||
self.assertEqual(mod._vm_name(["prog", "chv-t1"]), "chv-t1")
|
||||
old = os.environ.pop("CHV_VM_NAME", None)
|
||||
try:
|
||||
os.environ["CHV_VM_NAME"] = "chv-env"
|
||||
self.assertEqual(mod._vm_name(["prog"]), "chv-env")
|
||||
finally:
|
||||
if old is None:
|
||||
os.environ.pop("CHV_VM_NAME", None)
|
||||
else:
|
||||
os.environ["CHV_VM_NAME"] = old
|
||||
name = mod._vm_name(["prog"])
|
||||
self.assertIsInstance(name, str)
|
||||
self.assertTrue(name)
|
||||
|
||||
def test_send_writes_json_line(self):
|
||||
mod = _load_guest_module()
|
||||
client, server = socket.socketpair()
|
||||
try:
|
||||
mod._send(client, "heartbeat", "chv-t1")
|
||||
server.settimeout(2)
|
||||
data = server.recv(4096)
|
||||
msg = json.loads(data.decode("utf-8"))
|
||||
self.assertEqual(msg["type"], "heartbeat")
|
||||
self.assertEqual(msg["vm"], "chv-t1")
|
||||
self.assertIsInstance(msg["ts"], float)
|
||||
finally:
|
||||
client.close()
|
||||
server.close()
|
||||
|
||||
def test_main_exits_silently_when_host_unreachable(self):
|
||||
# 本机无 vsock 监听：main 应在重试窗口内静默返回 0，
|
||||
# 绝不抛异常、不影响调用方。
|
||||
mod = _load_guest_module()
|
||||
old = mod.CONNECT_RETRY_SECONDS
|
||||
mod.CONNECT_RETRY_SECONDS = 0
|
||||
try:
|
||||
self.assertEqual(mod.main(["prog", "chv-t1"]), 0)
|
||||
finally:
|
||||
mod.CONNECT_RETRY_SECONDS = old
|
||||
|
||||
def test_ports_agree(self):
|
||||
mod = _load_guest_module()
|
||||
self.assertEqual(mod.NOTIFY_PORT, chv_vsock.VSOCK_NOTIFY_PORT)
|
||||
self.assertEqual(mod.HOST_CID, chv_vsock.HOST_CID)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
Reference in new issue
Block a user