Mesh-Group Flooding Optimization in a Seven-Router Full Mesh
A metro service provider runs a full mesh of seven PE routers (three 7750 SR-12, four 7750 SR-7) connected by 21 point-to-point 10G Ethernet links. All routers are in the same IS-IS level-2 area and advertise about 1,200 prefixes in roughly 450 LSPs. Each PE has six IS-IS adjacencies. After the mesh grew from four to seven routers, LSP flooding and SPF load increased sharply: show router isis statistics shows LSP retransmissions climbing, and SPF runs 18-25 times per minute on the SR-12s. There is no alternate path outside the mesh, and the provider cannot add a route reflector because IS-IS does not use one. The design goal is to cut redundant flooding without creating a single point of failure or losing reachability when one mesh link fails.
1. Evaluate whether IS-IS mesh groups are appropriate here. Explain the flooding benefit and the exact failure mode they introduce.
Read the worked answer
Mesh groups are appropriate only if the operator accepts a flooding trade-off. An IS-IS interface assigned to a mesh group does not flood an LSP back into the same mesh group. In a seven-router full mesh, a router normally receives the same LSP from up to six neighbors and floods it to the other five. With all 21 interfaces in one mesh group, each router receives an LSP from the originator directly and does not forward it to the other mesh members. Flooding drops from roughly six copies per router to one, which reduces LSP processing and retransmissions. The failure mode is subtle: if one mesh link fails, the full mesh is no longer complete. The two routers at the ends of the failed link are no longer directly adjacent, and the other routers will not flood LSPs into the mesh group to reach them. New or changed LSPs originated by one of those routers may never reach the other until the link is restored or the mesh group is changed. The adjacency remains up on the other links, so the problem can appear as a silent blackhole rather than a visible down state. Mesh groups therefore reduce flooding but remove the mesh's automatic LSP detour.
2. Give the SR OS configuration and verification steps to place the mesh interfaces into a single mesh group, and state what you would check before doing it in production.
Read the worked answer
On each router, under the IS-IS interface context, assign a mesh-group number, for example configure router isis 0 interface "to-PE2" mesh-group 1. Repeat for each of the six point-to-point interfaces that should be in the group. The mesh-group value is locally significant, but the design should be symmetric so that every router treats the same set of links as one group. Before changing production, verify that all 21 links are truly point-to-point, that every router has a direct adjacency to every other router, and that no interface is accidentally excluded. Check show router isis interface "to-PE2" detail to confirm the mesh group and adjacency state, and show router isis adjacency to confirm the full mesh. After the change, compare show router isis statistics for LSP retransmissions and SPF runs, and use show router isis database detail to confirm that LSPs from each router are still present on every router. Do not apply the group to a partially meshed topology or to links that rely on a third router for flooding.
3. If the provider insists on retaining mesh groups, how would you redesign them so that a single link failure does not isolate LSPs?
Read the worked answer
Do not put every interface into one mesh group. A safer design is to use two mesh groups and ensure that each router has at least one interface in each group, so an LSP received in group 1 is flooded into group 2. That preserves a flooding path when one link fails, because the remaining group still provides an alternate route. A simpler variant is to leave one interface per router outside any mesh group as a dedicated backup flooding path. The trade-off is that the backup interface carries more LSPs and may become a congestion point, so it should be sized and monitored. In either design, verify the failure case explicitly: shut one mesh link in a maintenance window and confirm that every router still receives a fresh LSP originated by each of the two formerly adjacent routers. Use show router isis database detail and show router isis statistics to confirm that the LSP is present and that retransmissions do not spike. If the topology cannot tolerate any mesh-group failure mode, the correct answer is to avoid mesh groups and instead reduce LSP churn or SPF frequency rather than suppress flooding.