fix(wol-sleeper): wait for the internal address after a wakeup

Remote clients reach the backup server over wireguard only on its
internal address; waiting for the wakeonlan network address never
ended there, so htz.mails and the mseibert nodes had no backup since
2026-01-07. The wait now gives up after ten minutes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RywMtsL8qXG7GDY3MHq9Uh
This commit is contained in:
CroneKorkN 2026-10-23 14:30:21 +02:00
parent 81732f05ca
commit b8d281348e
No known key found for this signature in database

View file

@ -53,12 +53,18 @@ def wake_command(metadata):
mac = metadata.get(f"network/{metadata.get('wol-sleeper/network')}/mac")
network = ip_interface(metadata.get(f"network/{metadata.get('wol-sleeper/network')}/ipv4"))
waker_command = f"/usr/bin/wakeonlan -i {network.network.broadcast_address} {mac}"
# wait on the internal address: the wakeonlan network is not routed to remote clients (wireguard);
# bounded, so a failed wakeup cannot block the next runs forever
internal_ip = ip_interface(metadata.get('network/internal/ipv4')).ip
return {
'wol-sleeper': {
'mac': mac,
'waker_command': waker_command,
'wake_command': f"ssh -o StrictHostKeyChecking=no wol@{waker_hostname} '{waker_command}' && while ! ping {network.ip} -c1 -W3; do true; done",
'wake_command': (
f"ssh -o StrictHostKeyChecking=no wol@{waker_hostname} '{waker_command}'"
f" && timeout 600 sh -c 'until ping -c1 -W3 {internal_ip} > /dev/null; do sleep 2; done'"
),
},
}