Proper way to configure RSTP/Loop protection

Hello,

I have some problems with my infrastructure.
I have 2 CRS354 connected together via qsfp.

I need to connect a debian server to both, via two bonded interfaces, one to the first CRS, the other to the other.

Below is a picture of the infrastructure and configuration.

The problem is that when I raise the interface on the server, everything goes into a loop, router raise100% CPU, because there is a lot of dhcp requests.

CRS’s had configured 802.3ad bonding ports between each other, and to server + that bonded ports are added to main bridge with RSTP
Zrzut ekranu 2020-03-12 o 13.15.19.png
Zrzut ekranu 2020-03-12 o 13.13.03.png
Zrzut ekranu 2020-03-12 o 13.08.59.png

In the Linux system, create a bridge, add the bonds as bridge ports and enable STP on the server.

Stock Linux will do STP so you get 45 second failover.

Additionally your config snippet doesn’t match the drawing for the Linux NIC numbering.

Also, remove the check mark from BPDU guard. We want the Linux device to participate in STP in the end.

It doesen’t work.. After 30 seconds i have loop storm on my switches, and they goes down. Additionally linux bridge didn’t set root bridge id from one of my mt switches..