Hi martijnk2,

What version is the switch firmware on? Is there a reason you are designing it this way instead of doing port channels on both ports? So it doesn’t create a loop but still has redundancy.

Yes I do port channels lacp but for a cluster this wasn't supported, only active/passive.

But I have the same issue with lacp port-channels, reboot the server and the port gets shut down for some reason.

Firmware is 6.6.0.17.

There might be a misunderstanding of the (STP) loop guard feature.

From page 829 of the manual

Loop GuardLoop guard protects a network from forwarding loops induced by BPDU packet loss. The reasons for failing to receive packets are numerous, including heavy traffic, software problems, incorrect configuration, and unidirectional link failure. When a non-designated port no longer receives BPDUs, the spanning tree algorithm considers the link to be loop free and transitions the link from blocking to forwarding. Once in the forwarding state, the link may create a loop in the network.
Enabling loop guard prevents such accidental loops. When a port is no longer receiving BPDUs and the max age timer expires, the port is moved to a loop-inconsistent blocking state. In the loop-inconsistent blocking state, traffic is not forwarded so the port behaves as if it is in the blocking state; that is, it discards received traffic, does not learn MAC addresses, and is not part of the active topology. The port will remain in this state until it receives a BPDU. It will then transition through the normal spanning tree states based on the information in the received BPDU.

Means this feature is not intended to pretect agains looped end points.

Please have a look at the "loop protection" feature described in the CLI guide from page 633.

Be aware the "loop protection" feature does not play well with spanning-tree so please use it exclusively on edge ports, STP can stay enabled on connections to other switches.