OSPF load balancing problem

Hi all,

I have the following topology:

A --54/54-- B --54/54-- C --2/2-- D --55/55-- E --55/55-- A

where A, B, C, D, E are MikroTik devices of verious versions 2.9.x.
Node A has been included intentionally twice indicating that the topology is a closed loop. The “x/y” notation indicates that the OSPF cost for node X to Y is x while the cost from node Y to X is y.

My aim is to achieve load balancing between nodes A and D while still keeping dynamic routing (OSPF).

As can be seen the total OSPF cost for both available routing paths between nodes A and D is 110/110. Consequently, one would expect to observe in the route lists of nodes A and D a dual gateway to either of both nodes, i.e.:

  1. A to D via B, E .
  2. D to A via C, E

Node D (MTv2.9.2) manages to maintain dual gateways to A as described above while node A (MTv2.9.14) maintains dual gateways only for approximatelly 15min and after that sticks to B.

Why is node A giving up gateway E?

This is a list of all the MT versions running on each machine:

  1. A MTv2.9.14
  2. B MTv2.9.14
  3. C MTv2.9.6
  4. D MTv2.9.2
  5. E MTv2.9.11

Thanks in advance
nemo

maybe you noticed some connectedness?

I am not sure I understand what you mean.

The arrangement described in my first post works only for 15min. After that one of the load balancing endpoints loses one of the gateways and load balancing seizes.

If the OSPF costs were all right then load balancing should resume indefinitely. In the contrary, if OSPF costs were wrong then load balancing should not be available in the first place.