I have two CISCO 2811 routers connected. I have two Metro-Ethernet links between them.
I've configured OSPF and when one link goes down, OSPF send all traffic through another link. That's working fine.
But my problem is, when one link is poor, not down, but loosing packtes, OSPF still continuos send traffic through this one, what
make my applications working bad.
I would like some Routing Protocol, if its possible, to observe the link condition. If its loosing packet anyway, the protocol
doesn't send traffic through it.
Is it possible?
Is there another way to do that?
The condition you describe is independent of the routing protocol being used.
After a routing protocol has converged, it will have determined the best path to a destination network and forward the traffic out a certain interface.
The best way to avoid a single point of failure is to design a network with redundant, equal-cost paths and redundant equipment. When a circuit is failing miserably, you can manually shut down the interface and allow the traffic to take the alternate path.
That having been said, there are automated mechanisms in existence to mitigate the effects of a flapping interface. You can use IP Event Dampening.
IP Event Dampening is a health-check mechanism that you can confgiure on the router to monitor the "flapping" events of an interface. Basically, when the event threshold surpasses a pre-determined number, the interface is effectively shut down, its neighbor relationship with the peer router is removed and all routes learned through the dampened interface are not used.
This helps improve network stability and convergence times and reduces the load on your router's CPU because it doesnt have to cconstantly re-establish neighbor relationships and re-converge everytime the interface flaps.
When the circuit is back online and error free enough to NOT bring the neighbor relationship down, the interface will be restored.
Lastly, if your circuit is running errors, you should aggressively pursue it with your TELCO and demand that they fix it. It's as simple as that. That is the TRUE fix.
You can configure an IP SLA monitor to track thehealth of the device sitting on the other end of the circuit and create conditions under which the router should use an alternate path, but this is typically used with static routing.
When I realize the link is loosing packet, I ping its neighboor and I loose 60% of packets.
Does configuration you sent me monitor the link and can see these "loosing ping packets"?
Thanks for the valuable information. I learn quite a bit from experts like yourself @ this forum. Thanks and keep up the awesome posts (very helpful)
The routing protocol isn't designed to detect poor link conditions. You may need PfR
the main idea of my post is not to use Routing Protocol. Instead is offer higher avaibility to my applications.
Thank you for your help. I'll read more about PfR.
Your concerns are perfectly clear to me, and I would imagine Edison as well.
If you have a circuit that is running errors, pursue it aggressively with your TELCO and have them fix it.
If you would like to dploy a mechanism that will lessen the effects of a flapping interface, as a result of a bad circuit, I have offered you a potential solution as IP event dampening.
If you are not using a dynamic routing protocol, you could have investigated the use of IP SLA monitors.
Since you are using a dyanmic routing protocol, and routing protocols do NOT take into consideration the quality of a link (unless, of course, the link fails altogether), then Edison's solution may be one worthy of investigation.
In your example, I could figure out that the Router's interface is flapping, but its not what happening. Right?
What happen is the link is really poor, many packet loss and its not because the router interface, but because the Service Provider.... and a really call to the TELCO and they solve the problem, but it takes so long.
I would like some automatic solution...
What do u think?
Im sorry, I dont know what else to say. I have hit it from all angles.
I understand that you are concerned with the integrity of the link. The interface will flap if the link is really bad.
If not, look into PfR.
The interface will flap if the link is really bad.
Not with Metro-E as your switch is connected to a local CPE device.
In NYC, Verizon uses this product a lot:
Necessito trabajo nuevo. Hay algo?
So, this device terminates the ethernet link from the router at the CPE?
Not sure why you say a bad link wont cause the interface to flap...
This unit connects to the Service Provider network via the fiber port and the it has a copper handoff to the customer.
The copper handoff needs to flap in order for the ip dampening at the customer device to work.
The fiber may flap but that's unknown to the customer, and it's really rare to have a flapping condition when the device is connected 5ft apart :)
Ok, then of course its obvious.
I wasnt thinking that the metro-E conenction was terminated on the CP itself. If thats the case, then of course it wont know about what happens beyond that device because that device is providing the layer 1 and 2 handshake for the ethernet interface on the router.
BTW, you never answered my question about...
Read the link I posted earlier, it has all the information under Q&A.
Personally, I know the concept on PfR but I'm not well versed on it. The technology is very new.
I would highly recommend (PFR) Performance routing as suggested by other colleages.
Performance routing consist of the following:
1- 2 or more boarder routers.
2- 2 or more egress interfaces.
3- Master Controller.
Performance routing takes into account the following to determine the best path in a Network.
a) packet loss.
c) delay variation (jitter)
3- Cost minimization.
Unlike normal IGPs where they normally determine best paths based on Cost calculation, Performance routing takes all the above when calculation the Cost of the Path.
Performance routing is supported on some series of routers, 2800 is one of the supported Modular ISR for (PFR).