fix(prepare-win): wait for reboot to actually happen (down→up)

On exit 3010 the old path issued Restart-Computer, slept 30s, then probed
WinRM. With a pending WU/CBS install the restart is delayed minutes while
the WinRM session stays alive, so the probe saw the still-up pre-reboot
session as "ready" and re-ran the toolchain — which then died mid-run when
the real reboot kicked in (looping "treating as 3010").

Add `_reboot_and_wait`: issue the reboot, then wait for TCP/5986 to go
*down* (`_wait_tcp_down`, up to 300s) before waiting for it to come back
and re-probing WinRM. Use it for both the 3010 loop and the final reboot.
This commit is contained in:
2026-06-09 20:53:19 +02:00
parent 6a57c1d815
commit 107bff8c8a
2 changed files with 83 additions and 12 deletions
+40 -12
View File
@@ -292,6 +292,44 @@ 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:
return True
return False
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."""
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)
def _parse_exit_marker(stdout: str) -> int:
"""Extract ``CI_EXITCODE:N`` written by the invoke wrapper; default 0."""
for line in reversed(stdout.splitlines()):
@@ -582,12 +620,7 @@ def template_prepare_win(
)
click.echo("[prepare-win] Windows Update requires reboot — rebooting ...")
with contextlib.suppress(TransportError):
transport.run("Restart-Computer -Force", check=False)
transport.close()
time.sleep(30)
click.echo(f"[prepare-win] Waiting for WinRM after reboot ({winrm_timeout}s) ...")
_wait_winrm(transport, winrm_timeout)
_reboot_and_wait(transport, guest_ip, 5986, winrm_timeout)
click.echo("[prepare-win] WinRM ready after reboot.")
# Install CI-StaticIp scheduled task
@@ -653,12 +686,7 @@ Register-ScheduledTask -TaskName 'CI-StaticIp' `
# Final reboot to flush installation state
click.echo("[prepare-win] Final reboot to flush installation state ...")
with contextlib.suppress(TransportError):
transport.run("Restart-Computer -Force", check=False)
transport.close()
time.sleep(30)
click.echo("[prepare-win] Waiting for WinRM after final reboot ...")
_wait_winrm(transport, winrm_timeout)
_reboot_and_wait(transport, guest_ip, 5986, winrm_timeout)
# Disk cleanup — runs after final reboot on fully-installed system
if not skip_cleanup: