rsyslog attached to OpenVPN-managed tun0 fails to bind to tcp socket on restart

[Christian Reis]

My rsyslog configuration listens on TCP port 514 and is bound to an address on tun0 (created by an openvpn server daemon):

  module(load="imtcp")
  input(type="imtcp" port="514" address="10.8.9.1")

This machine rebooted (due to unattended upgrades) overnight and when it came back up, rsyslog was not listening on the port. The logs show:

  Oct 24 03:00:16 foo rsyslogd-2077: Could not create tcp listener, ignoring port 514 bind-address 10.8.9.1. [v8.16.0 try http://www.rsyslog.com/e/2077 ]

A restart of rsyslogd was required to address the issue. I'm curious what the right solution to this problem would be, as the bind interface is configuration-dependent.

[Christian Reis (kiko)]

I discussed with slangasek, who said:

  - The workaround is to either to configure systemd to restart rsyslog when the socket is available, or to have openvpn restart rsyslog when the VPN comes up.
  - The definitive fix would require rsyslogd to grow netlink support (or be managed by something else that does).

For now, I'm working around this by using openvpn config.