Re: How are you configuring BFD timers?
--- saku@ytti.fi wrote: From: Saku Ytti <saku@ytti.fi> ...but I have far more BFD caused problems than BFD solved problems, spanning multiple vendors. (CAT7600, ASR9k, MX). ---------------------------------------- Yes, that's for sure. Also, it's hard to scale when you're tweaking knobs on each session trying to get the time down w/o causing failure unnecessarily. scott
No sure if this link has been provided yet, but this is how I learned BFD - https://supportforums.cisco.com/t5/service-providers-documents/bfd-support-o... My only experience with BFD has been with short paths using grey optics and interstate DWDM spans. I found 3x50ms echo mode worked well, but you need to watch out for QoS on the remote side as the packet that hairpins back to the sender is subject to queuing. As the link becomes saturated, the BFD packet goes in the queue with everyone else as the far end router hairpins it and can cause a false link down condition if it goes in the bit bucket. I saw timers as low as 3x10ms echo mode with QoS work really well on a strictly ASR9k network. I never tried to run it on bundle links or over layer 2. I did try to run it on some Dell Z9100 and S4048 boxes running FTOS 9 and it failed miserably even with very conservative timers. I haven't had a chance to test it with IPI OcNOS 1.3.3 on the same boxes, or with JunOS. On Thu, Mar 22, 2018 at 3:16 PM, Scott Weeks <surfer@mauigateway.com> wrote:
--- saku@ytti.fi wrote: From: Saku Ytti <saku@ytti.fi>
...but I have far more BFD caused problems than BFD solved problems, spanning multiple vendors. (CAT7600, ASR9k, MX). ----------------------------------------
Yes, that's for sure. Also, it's hard to scale when you're tweaking knobs on each session trying to get the time down w/o causing failure unnecessarily.
scott
participants (2)
-
Ben Bartsch
-
Scott Weeks