04-15-2009 11:32 PM - edited 03-04-2019 04:23 AM
Hi All,
We are running an EOS link between 2 routers which is backed up by another EOS link. EOS links are terminated in the switch and extended to the routers on a single trunk port with 2 VLAN's and sub-interfaces.
OSPF is configured to run over the 2 EOS links. OSPF network point to point is configured on the routers.
The problem that we are facing now is when any one of the EOS link goes down, OSPF takes 40 sec (default dead interval timer) to go down as the router's interface always stays up.
Dead interval is confiugred to minimal of 1 sec (hello interval is automatically reduced according to the dead interval that we set) under the sub-interfaces in routers and now if the EOS link goes down, OSPF also goes down in 3 sec and everything is fine.
The issue that we are facing now is if the EOS backbone has got any reliability issue and packet drops, OSPF on EOS between 2 routers of ours is continuously flapping. This causes high CPU load on our routers.
Would like to know:
a. Is there any configuration or feature enhancement kind of thing where if the OSPF on a particular link flaps continuously we can put that link into passive for certain amount of time (should be done automatically once we configured)
b. Is there best way to have a workaround for this problem other than manipulating hello and dead interval.
(Note: "Terminating EOS links directly on the router so that if EOS goes down brings the router interface also down" This option is not feasible because of the compatiblitiy issues on media between us and provider).
thanks in advance.
Arun
04-16-2009 12:33 AM
Hi Arun,
Nice thought. Since your provider does not support something called as Link Loss signalling. Please try the following.
1. IP Event Dapening
2. Bidirectional Forwarding Detection ( BFD)
is something which you can try rather than tweaking the OSPF hello timers. BFD is a very light weight protocol which can perform the same stuff.
Please go through cisco docs and learn the technology and implement it.If you do not understand it. Please feel free to contact me.
Regards,
Srinath M
Cable&Wireless
+91 99809 30364
04-16-2009 04:23 AM
As Srinath describes, IP dampening, might be an ideal solution other than hello and dead interval management. Beside his suggestion of BFD, another option might be to increase the number of missed OSPF hellos. This would increase the lost detection time, but that can be offset by decreasing the hello timer. (Since you're already at 1 second, you would need to determine whether your IOS supports subsecond OSPF hellos if you didn't want to increase your failure detection time.)
04-19-2009 10:10 PM
Hi Srinath & Joseph,
Thanks for your valuable inputs. I cannot use event dampening as the physical interface in the router will always stays up (EOS terminates on switch). BFD will be an ideal option in our network setup and i m testing this.
thanks
Arun Kumar
04-20-2009 12:28 AM
Hi Arun,
Please let us know the results.
Regards,
Srinath M
+9199809 30364
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide