I am running an x86 server at head office (now running 4.16) and a number of RB433 remote offices 'dial-in' via OpenVPN.
One link is perpetual. It seems to stay up all of the time. _very_ occasionally it dies and has to be manually re-started. It is running over ADSL 2+.
The second link is very unreliable. It is running over an ADSL 1 link which appears to be running with variable signal to noise and occasionally has bursts of errors. Needless to say the telco cannot/will not fix this.
What happens is that the OVPN link stops running and eventually OSPF stalls. Then OVPN re-dials but often (more often than not) re-starts with errors in the cypher meaning that the link runs as a dynamic link and does not properly start. Manually re-starting solves the problem but on occasions this link can drop several times in ten minutes (and on other occasions it can survive for much longer).
It is important that the link is encrypted to a good degree.
Can anyone suggest either a rapid way of detecting an error and re-dialling or, alternatively, some way of getting a VPN link to run over another interface (PPP?) or even several in parallel to minimise the overall link down-time?
It would in any event be desirable for the link to auto-re-start which it does not do at the moment.