ndppd does not wait for docker interfaces

[fansari]

In my scenario I use ndppd for docker interfaces (which is quite a typical scenario I guess).

But on startup the daemon complains that it cannot bind to the interfaces (because there are not available yet).

With "top" I see that ndppd is taking 100% of the CPU time after boot.

Mar 14 12:36:10 myhost.mydomain systemd[1]: Starting NDP Proxy Daemon...
Mar 14 12:36:10 myhost.mydomain ndppd[748]: (notice) ndppd (NDP Proxy Daemon) version 0.2.4
Mar 14 12:36:10 myhost.mydomain ndppd[748]: (notice) Using configuration file '/etc/ndppd.conf'
Mar 14 12:36:10 myhost.mydomain systemd[1]: ndppd.service: Can't open PID file /run/ndppd/ndppd.pid (yet?) after start: Operation not permitted
Mar 14 12:36:10 myhost.mydomain ndppd[757]: (error) Failed to bind to interface 'docker0'
Mar 14 12:36:10 myhost.mydomain ndppd[757]: (error) Failed to bind to interface 'br_nginx'
Mar 14 12:36:10 myhost.mydomain systemd[1]: Started NDP Proxy Daemon.
Mar 14 12:36:12 myhost.mydomain ndppd[757]: (error) Failed to read from interface '%s'
Mar 14 12:37:30 myhost.mydomain systemd[1]: Stopping NDP Proxy Daemon...
Mar 14 12:37:30 myhost.mydomain ndppd[757]: (error) Shutting down...
Mar 14 12:37:30 myhost.mydomain ndppd[757]: (notice) Bye
Mar 14 12:37:30 myhost.mydomain systemd[1]: ndppd.service: Deactivated successfully.
Mar 14 12:37:30 myhost.mydomain systemd[1]: Stopped NDP Proxy Daemon.
Mar 14 12:37:30 myhost.mydomain systemd[1]: ndppd.service: Consumed 2.069s CPU time.
Mar 14 12:37:30 myhost.mydomain systemd[1]: Starting NDP Proxy Daemon...
Mar 14 12:37:30 myhost.mydomain ndppd[1804]: (notice) ndppd (NDP Proxy Daemon) version 0.2.4
Mar 14 12:37:30 myhost.mydomain ndppd[1804]: (notice) Using configuration file '/etc/ndppd.conf'
Mar 14 12:37:30 myhost.mydomain systemd[1]: ndppd.service: Can't open PID file /run/ndppd/ndppd.pid (yet?) after start: Operation not permitted
Mar 14 12:37:30 myhost.mydomain systemd[1]: Started NDP Proxy Daemon.

My first try was to startup my container with a systemd service and then configure the npppd.service with a "Require" and "After" statement for this service. But this did not help (probably because systemd does not wait until interface and container are up).

I have configured a workaround now with a systemd timer which simply restarts the ndppd one minute after boot.

What is the proper way to configure this?

# /etc/ndppd.conf

proxy ens18 {
  rule 2xxx::xxxx:xxxx:xxxx::/80 {
    iface docker0
  }
  rule 2yyy::yyyy:yyyy:yyyy::/80 {
    iface br_nginx
  }
}

Tested with ndppd-0.2.5-6.el9.x86_64 on CentOS Stream 9.