I see the same behavior in a single node k3s installation, but the eBPF error is not present this time. So the 2 things seems not related.
a
abundant-fish-48120
10/14/2022, 3:24 PM
looks like a simple network issue but I am not familiar with proxmox
abundant-fish-48120
10/14/2022, 3:25 PM
connection timeout could be anywhere from a routing table issue to iptables issue
c
cuddly-egg-57762
10/17/2022, 7:09 AM
it looks like the problem was caused by the massive I/O that was happening on the disks. This because all the VMs was sharing a single datastore which was overloaded of I/O.
Checking the k3s.service log I've seens a lot or warning regarding the ETCD server replies which were too slow.
Givind a dedicated disk to ETCD seems has solved the problem.
Thanks a lot for your help @abundant-fish-48120