i'm in charge of a small r&d cluster that is currently running with 1 mixed control/infra/compute node(formerly "all in one" stack), 1 storage node and 1 compute node. The cluster was working fine. I recently upgraded the cluster to ubuntu 20.04 and the victoria release . since then, it appear the instances that are spawned on the dedicated compute node, cannot get ipv6 connectivity (ipv4 works fine), the one that spawn on the mixed node get ipv4 and v6 fine, i've tried to manually assign the address that are allocated but without success. the only error i can see is in neutron-linuxbridge-agent.log ERROR oslo_messaging.rpc.server neutron.privileged.agent.linux.ip_lib.InvalidArgument: Invalid parameter/value used on interface vxlan-2021, namespace None
however when i tcpdump on this interface on the compute node i can see the dhcpv6 request and replies so i assume ipv6 works up to that point. i've checked the ipv6 forwarding and it seems to be enabled, but considering the OS configuration are mirrored between the mixed node and the compute node and the instance are properly working on the mixed node, i have no idea what i might've missed.
everything is installed from the package install for ubuntu https://docs.openstack.org/install-guide/environment-packages-ubuntu.html. it has been upgraded a few times without issues.
any help is welcome.