Closed larsks closed 3 years ago
Because the smaug worker nodes have only a single network interface, this ends up being a little trickier than I had hoped. The interface is a member of the br-ex OVS bridge, which is maintained by the SDN subsystem (OVN-Kubernetes). On the one hand, this is advantageous: because the primary interface is a member of an OVS bridge, we can just add a tagged port to the bridge and we don't need to worry about the fact that the systems have different interface names, because OpenShift has already taken care of this for us.
The downside is that we can't use a declarative nmstate configuration, because nmstate would want to define the OVS bridge.
The alternative is installing a systemd unit that will perform the necessary network configuration whenever the system boots. This will require deploying a machineconfig across the cluster, which will require a rolling reboot of all cluster nodes.
(Please ignore the previous content of this comment.)
Working now!
We need to provide persistent storage to the smaug cluster. Ideally this would be via the Ceph service provided by NESE, but their engineer (yes, singular) appears to be swamped and we have been unable to tease an ETA out of them. In the interim, one option is to connect smaug to the existing NFS server.