How to properly organize a backup channel from the provider to the DC?

Question

Computer networks

Good day, dear All.
Do not kick hard with your feet if the question is stupid. Haven't come across anything like this before.
There is a rack in the DC, stuffed with servers. All this is connected to 2 switches, the connections are organized in such a way that the failure of any switch does not interrupt the service, but ... with the exception of the link from the provider. If that switch dies, where the provider is plugged in, then ops.
In fact, there are two providers, they are plugged into different switches, and in my mind I understand that I need to get confused with organizing my own AS (network / 24 is rented) and raising BGP, but in fact the client is not yet so big and fat.
Accordingly, the question is: how to make a channel reservation in the aspect of protection against the death of the switch? Will it be correct to ask the provider to tighten another cable onto the rack, plug it into the second switch and agree on raising STP, for example?
Or is this problem solved in other ways? Or does no one do it at all and you need to immediately dig towards BGP?

Answer 1

According to your mind, you should have a stack of 2 routers, in which both uplinks from providers and switches are connected, on which the rules for balancing and forwarding the economy to the world are set.
Z.Y. RIPE no longer provides PI addresses in IPv4.

Answer 2

By the way, HA is not fault tolerance, it is high availability. Fault tolerance is FT.
If your service is available outside, then only your own AS will help you. Or a semi-crutch option that some Dyn implements - if a certain IP is unavailable, it changes A-records to a backup IP from your other provider. And on the router, port-mapping or DNAT is already sorting out who is where.
If you need to reserve the exit of your servers to the outside, then I don’t see any problems at all - all servers have a router set as the default gateway, the router monitors the availability of channels, and when one of them fails, it switches to reserve.

Answer 3

Here is your schema. You can implement it through one L3 segment + PBR & HSRP, and if you have your own AS using BGP, I will announce what you need and what to go through.