fix(prepare-win): detect reboot by boot-time, not TCP-down
A fast reboot never showed TCP/5986 "down" between polls, so the old down→up wait could hang on "waiting for VM to go down" until timeout. Replace it: record the guest LastBootUpTime (FileTime ticks) before the reboot, then poll until the port is back AND the boot time is newer — a signal immune to reboot speed. Applies to both the 3010 loop and the final reboot. Drops _wait_tcp_down in favour of _tcp_open + _guest_boot_ticks.
This commit is contained in:
@@ -292,42 +292,61 @@ def _wait_tcp(host: str, port: int, timeout: float) -> bool:
|
||||
return False
|
||||
|
||||
|
||||
def _wait_tcp_down(host: str, port: int, timeout: float, interval: float = 5.0) -> bool:
|
||||
"""Return True once TCP *port* on *host* stops accepting connections (host rebooting).
|
||||
|
||||
Used after issuing a reboot: a pending Windows Update / CBS install can delay
|
||||
the actual restart for minutes while the WinRM session is still alive, so we
|
||||
must observe the port go *down* before waiting for it to come back — otherwise
|
||||
we'd treat the still-up pre-reboot session as "ready".
|
||||
"""
|
||||
deadline = time.monotonic() + timeout
|
||||
while time.monotonic() < deadline:
|
||||
try:
|
||||
with socket.create_connection((host, port), timeout=5.0):
|
||||
time.sleep(interval)
|
||||
except OSError:
|
||||
def _tcp_open(host: str, port: int, timeout: float = 3.0) -> bool:
|
||||
"""True if *port* on *host* currently accepts a TCP connection."""
|
||||
try:
|
||||
with socket.create_connection((host, port), timeout=timeout):
|
||||
return True
|
||||
return False
|
||||
except OSError:
|
||||
return False
|
||||
|
||||
|
||||
def _guest_boot_ticks(transport: WinRmTransport) -> int | None:
|
||||
"""Read the guest's last boot time as FileTime ticks; None if unreachable.
|
||||
|
||||
Used as a reboot signal that is immune to how fast the box restarts —
|
||||
comparing boot time before/after a reboot is reliable even when the WinRM
|
||||
port never appears "down" between polls.
|
||||
"""
|
||||
try:
|
||||
result = transport.run(
|
||||
"(Get-CimInstance Win32_OperatingSystem).LastBootUpTime.ToFileTimeUtc()",
|
||||
check=False,
|
||||
)
|
||||
except TransportError:
|
||||
return None
|
||||
try:
|
||||
return int(result.stdout.strip())
|
||||
except (ValueError, AttributeError):
|
||||
return None
|
||||
|
||||
|
||||
def _reboot_and_wait(
|
||||
transport: WinRmTransport, host: str, port: int, winrm_timeout: int
|
||||
) -> None:
|
||||
"""Reboot the guest and wait for it to actually go down then come back ready."""
|
||||
"""Reboot the guest and wait until it comes back with a *newer* boot time.
|
||||
|
||||
A TCP "down" probe can miss a fast reboot entirely, so instead we record the
|
||||
boot time before the reboot and poll until WinRM answers with a later one.
|
||||
"""
|
||||
boot_before = _guest_boot_ticks(transport)
|
||||
with contextlib.suppress(TransportError):
|
||||
transport.run("Restart-Computer -Force", check=False)
|
||||
transport.close()
|
||||
click.echo("[prepare-win] waiting for VM to go down ...")
|
||||
if _wait_tcp_down(host, port, timeout=300.0):
|
||||
click.echo("[prepare-win] VM down — waiting for it to come back ...")
|
||||
else:
|
||||
click.echo(
|
||||
"[prepare-win] [WARN] WinRM still reachable after reboot request; "
|
||||
"proceeding anyway."
|
||||
)
|
||||
if not _wait_tcp(host, port, float(winrm_timeout)):
|
||||
raise click.ClickException(f"TCP/{port} on {host} not reachable after reboot.")
|
||||
_wait_winrm(transport, winrm_timeout)
|
||||
click.echo("[prepare-win] waiting for guest to reboot (boot-time change) ...")
|
||||
deadline = time.monotonic() + float(winrm_timeout) + 300.0
|
||||
while time.monotonic() < deadline:
|
||||
time.sleep(5)
|
||||
if not _tcp_open(host, port):
|
||||
continue # still down / not back yet
|
||||
ticks = _guest_boot_ticks(transport)
|
||||
if ticks is not None and (boot_before is None or ticks > boot_before):
|
||||
click.echo("[prepare-win] guest rebooted — WinRM back.")
|
||||
_wait_winrm(transport, winrm_timeout)
|
||||
return
|
||||
raise click.ClickException(
|
||||
f"Guest did not reboot / WinRM not back on {host}:{port} within timeout."
|
||||
)
|
||||
|
||||
|
||||
def _parse_exit_marker(stdout: str) -> int:
|
||||
|
||||
Reference in New Issue
Block a user