Recently I have been part of the design of a new MPLS core network, consisting of four NCS-6008. A bundle is configured on each side of the square, which consists of x2 HundredGig links, giving us enough capacity for all our applications and services.
In case of a failure of one of the two bundle member, as per default operation , the bundle will stay up and run half of the bandwidth. However, despite having on each physical interface the recommended carrier delay and dampening value, we preferred to shut down the entire bundle in case of a single bundle-member flap. This is achieved by setting the minimum number of active links to the number of links configured.
bundle minimum-active links <em>[total number of links]</em>
The reason behind this design choice, is to avoid any kind of link congestion (and thus, potential jitter) affecting our real-time traffic, like IPTV, VoD and VoIP. Plus, inside a DWDM system, a single failed link might also be a symptom of a wider fiber issue, which could degrade apparently healthy links with CRC and frame drop. IGP convergence time (is-is in our case) is quick enough to provide a smooth traffic reroute (especially if Remote LFA is in place)
In short, if your network design permit so, is better to reroute traffic on a geographical parallel link, instead of risking a worse side effect.
P.S. I know I have been absent for a long time, but it’s due to my intense CCIE studies which eventually led me to the yearned number! :) - I will write soon on how I prepared the lab.
2da5b52 @ 2019-07-20