MSTP with no single point of failure with 4 switches and 2 regions

Question

I'm trying to come up with a plan to get our core network design a bit more resilient to failure.

Here is a drawing of what I would like to achieve:

The final design would probably involve one more instance, but the drawing got way too cluttered with a third line all around the place.

What I've figured out on my own is how to set the MST instance root for each instance. What I don't quite get is how I'm supposed to design this that RACKSW02 takes over root for MST2 if RACKSW01 dies etc. Is it just a matter of setting the correct MST root priorities around the setup?

Do I need to set custom link costs if I want to override the default metrics, as RACKSW01<->RACKSW02 is 20gbps, while CORESW01/02 only has 10gbps?

The design goal behind all this is to balance out the load across the links, as 99.9% of the MST2 instance is only between the two top-of-rack switches.

Answer 1

Yes, it's that simple, just set priorities appropriately :-) So you configure your Core 1 with highest and Core 2 with lower priority for instance 1, and configure instance 2 similar way. You leave other devices in the network to their default priorities. The only way to overwrite the default link costs is to set them manually.

I would consider a different configuration. There's no point setting your STP root to the edge. Rather leave the core switches as roots for different MSTP instances; for example Core 1 for odd and Core 2 for even instances. This way the root ports are core uplinks so non-blocking, and you don't have to care about the link costs on the edge. Another advantage is that it's easier to keep track on which ports are blocking on which instances.