GRE link state detection

[thegeezer]

It's OSPF, and  it has the setting tick 2; but even after 2 seconds the
lossy link is not detected
tcpdump definitely confirms that traffic is sending on routerA down vpn0
but not receiving at the routerB vpn0

the issue is showing up with SIP phones that are using the vpn, and the
udp stream continues to push out through the lossy link.
however new connections are then sent via another i.e. vpn1


On 09/05/2013 04:30 PM, Matthias Cramer wrote:
> Hi
>
> What routing protocol do you run ? Would it be possible to just lower the hellos in the used
> protocoll so it detects a link loss faster ?
>
> Regards
>
>   Matthias
>
> On 05/09/13 17:07, thegeezer wrote:
>> Howdy all,
>> i was wondering if anyone has any idea if there is a means by which i
>> can detect GRE link state ?
>>
>> what i have is two sites each with two very unstable internet links
>> in order to vpn between them i have ipsec tunnels linking each side
>> twice (four ipsec tunnels in total)
>> i then have 4x GRE tunnels over the top of those in order that i have a
>> secured routable VPN this gives me net.vpn0 net.vpn1 net.vpn2 and net.vpn3
>> finally i run BIRD over the top which works very well, and synchronises
>> routing tables between the two sites, and allows for me to do such fun as
>> # /etc/init.d/net.vpn0 stop
>> and watch all traffic automagically cut over to another link.
>>
>> so far so awesome.
>>
>> however, as i said the internet links are very unstable, and sometimes
>> just blackhole. so what i was hoping to do is just enable keepalives on
>> the gre tunnel. which sadly seems to be cisco only.
>>
>> can anyone suggest a way of detecting if the GRE is not fully connected ?
>> BIRD only fails over if the net.vpn0 device is down (ifconfig up/down)
>> and for the life of me i cannot find how to detect if a GRE tunnel is
>> 'connected', it seems to just blindly send packets to the remote IP. is
>> my only choice to use L2TP instead ?
>>
>> thanks in advance for any suggestions;
>> PS if i'm on the wrong list i'd welcome a friendly push to the correct list
>>
>