HI all, that´s the thing, this morning when I was reviewing the log in a router I saw what the ospf and bgp neighbors were down for one minute approxly but there is not any log about an interface down or something like that, the link is a Mono Mode fiber link. I don´t know if this behaviur is ok for this link type, it is the first time I work with fiber links.
Thanks in advance.
Do you control the routers at both ends of the link ? If not, it's possible that some work could have been done on the router at the other end which impacted your adjacencies.
We manage the both routers and there was not any schwduled job... I was thinking about the posibility of any work in a fiber patch pannel of our provider (I don´t know about the technical word to this place, I want to refer to the place where all the fiber links are connected in order to get connectivity between two remote routers, this place is managed by the company who manage the circuits).
Here is an example of about what I want to explain:
PE router --- Provider Patch Pannel ----- Provider Patch Pannel -- Provider Patch Pannel ---- PE router
There is points between the two ends managed by the telecoms provider and I think that a job or issue or whatever in the middle of the path between our PE routers could be the reason of lossing neighbors adjacency while interfaces were UP.
Thanks for your quick response!
Fibre links have issues with being able to pass traffic in one direction but not the other. A uni directional link failure might have taken place. In this case, the interface would remain in up / up state but data would not flow in both directions which could explain the neighbor loss.
In switched networks, UDLD (uni directional link detection) features are provided for all switchports. The same might be applicable on routed fibre interfaces as well.
I have been taking a look but there is no way to enable udld in a 7304 router (our PE´s routers), it could be enabled in the P routers (7606) but the issue is in the PE devices...
Looks like a tough call then.
What you can do is to raise a call with your provider about this issue and ask them to give you a report using their testing equipment to confirm that the link is stable. Dont bother with that if youre in India though .. coz all the reports are fudged (atleast my providers are) :)
I can´t raise calls for the moment, we are deploying the network, the only thing is try to get a response speaking directly with the team what manage the circuits... I am afraid because this type of issues can turn me crazy when the network will start runnning...
Thanks a lot for the commnets ;-)
I agree with Arvind that most providers won´t tell you if there is a short outage, although I found that with a lot of asking, and being lucky enough to talk to the right person, you sometimes actually are able to get them to check their logs...
Anyway, in your case, since the interfaces did not go down, the problem might simply have been congestion on the link. You could use a tool such as PingPlotter (www.pingplotter.com) and send a ping to the other side, the tool reports delays and missed pings. If you experience the lost neighbor issue again, you can check that against the data you get from the tool.
There wasn´t congestion in the link because the network isn´t been used, we´r deploying it and client traffic is not coming in the network. I´ve been tryin contact with someone for an answer from the service provider but there is no way... I´ll continue looking for phantoms inside my network ;)
I have seen the same issues where you get paged at night cause ospf lost a neighbor but the interface is not down which does not make sense cause the dead interval is 40 sec and keepalive on the interface is 40 sec . My only explaintation is that the ospf packes got droped by the service provider for some reason . i am not very sure but i myself would be happy to find out why this happens