Glossary & concepts

Concepts

Short explanations of concepts used across the test plans — networking, storage, secrets management, and more — for anyone who does not work with them every day.

macvtap (NOT supported) vs. OVN-K localnet (official path)

macvtap gives a VM its own MAC — but is NOT officially supported in OpenShift Virtualization. The correct approach is OVN-K localnet.

WaitForFirstConsumer (StorageClass binding mode)

A PVC stays deliberately 'Pending' until a pod/VM actually uses it — not an error!

NodeNetworkConfigurationPolicy (NNCP)

Declarative node network configuration via the Kubernetes API — instead of manual nmcli commands.

OVN-Kubernetes & br-ex

OpenShift's default network plugin uses an OVS bridge (br-ex) that takes over the physical NIC.

HCP: does every hosted cluster need its own public IP?

No — only the API server (with the KubeVirt/on-prem provider) needs a dedicated IP or dedicated port. OAuthServer, Konnectivity, and Ignition already share the management cluster's existing Ingress route today.

Sealed vs unsealed — why pods stay 0/1 Ready

A sealed OpenBao holds only encrypted data and refuses almost every request — the pod runs but is not Ready. Unsealing reconstructs the barrier key.

KV v2 — versions, soft delete, and check-and-set (CAS)

KV v2 keeps a version history per secret and can require a version number on every write — a write is a whole new version, not a field update.

Kubernetes auth method — SA token in, OpenBao token out

Pods authenticate with their projected ServiceAccount JWT; OpenBao validates it against the kube-apiserver and returns a short-lived token carrying the role's policies.

OpenBao vs Vault — why /vault/ still appears everywhere

OpenBao is the Linux Foundation fork of HashiCorp Vault and keeps API and tooling compatibility — so UI routes, sidecar paths, and third-party integrations still say 'vault'.

Raft integrated storage — leader, quorum, snapshots

OpenBao stores data itself and replicates it with the Raft consensus protocol — no external database, but quorum rules apply.

With Raft storage every server pod keeps a full copy of the data on its own PVC and the cluster elects one leader; writes go through the leader and are replicated to followers. A 3-node cluster tolerates the loss of one node (quorum = 2 of 3). This is why TC-OPENBAO-HA-001 can step down the leader and a follower takes over within seconds, and why removing a peer without wiping its PVC causes stale-state trouble — on a StatefulSet the PVC, not the pod, holds the Raft identity. Raft also provides the backup primitive: a snapshot is a consistent, complete copy of everything (secrets, policies, auth config) in one file.

✓ ADVANTAGES

No external storage backend to operate — PVCs are enough
One-file backups via Raft snapshot (TC-OPENBAO-HA-002, TC-OPENBAO-BK-001) and full-state restore (TC-OPENBAO-BK-002)
Automatic leader election — failover needs no operator action

⚠ LIMITATIONS

Quorum maths: 2 nodes are worse than 1; always run odd numbers
Deleting a pod is not enough to reset a member — the PVC keeps stale Raft state (TC-OPENBAO-HA-001 re-join step)
All nodes hold all data — size PVCs accordingly

📚 Glossary & concepts

Glossary & concepts