You could use the networkd-broker tool (or its predecessor networkd-dispatcher) to react to networkd's configuration events, or a netlink-based tool (similar to `ip mon addr`) to directly watch IP address changes.
To be honest though, a bond0 that's connected to two completely different networks kind of makes no sense to me at all. How is it better than just having two network interfaces?
On Thu, Sep 29, 2022 at 12:22 PM m1027 <m1027@xxxxxxxxxx> wrote:
Hi,
With a working wlan/lan bond0 failover, how can we trigger other
local services whenever a failover happens?
Why: In our case, a local openvpn client service needs SIGUSR1
whenever the own IP address changes, after a failover. See man
openvpn(8), --ipchange.
The problem: After a failover, the openvpn client keeps sitting on
the wrong IP address/interface. Sending SIGUSR1 manually helps
indeed: openvpn then reconfigures itself and uses the correct IP
address. The man page recommends to write a hook script for the
local dhcpcd to send SIGUSR1 to the openvpn client, however we are
not using dhcpcd but systemd.
Some more information:
- Failover works in general here; pulling out the cable from lan1
activates wlan0 and vice versa, and this triggers external DHCP
servers to issue a new IP address. The external servers are
triggered because of FailOverMACPolicy=active for bond0.
- Our bond0 is on top of wlan0 + lan1 and configured like this:
[NetDev]
Name=bond0
Kind=bond
[Bond]
Mode=active-backup
AdSelect=bandwidth
FailOverMACPolicy=active
- We know of networkmanager's dispatcher scripts but are looking for
a solution within systemd. (It's also unclear whether our network
setup could be configured entirely by networkmanager.)
- There is a workaround: There is the "inactive <sec>" option for
the openvpn client to shutdown in case of no action. And we can
additionally set the client to Restart=always. However, this
interrupts even working openvpn client sessions every 60 sec of
inactivity.
Thanks for any pointers...
--
Mantas Mikulėnas
