There is a cluster using GPU pass-through. It consists of 11 nodes, including 3 GPU nodes. Servers utilizing GPU pass-through are billed servers and should ideally have minimal or no downtime.
Due to the current End of Service Life (EOSL), an upgrade is being considered.
For this, the plan involves excluding one GPU node from the cluster, upgrading it to AOS and AHV standalone. Then, using cluster expansion to add it back, turning off the power of the GPU pass-through server, migrating it, and subsequently upgrading the remaining nodes through LCM.
Is this plan feasible?