Service Mesh and mTLS
Traffic between the cluster and external systems is already encrypted with automatically managed certificates. Traffic between workloads inside the cluster is not, unless you enable a service mesh.
OLIVE supports mutual TLS between services through an Istio-based service mesh, switchable at deployment time. The supported and recommended shape is ambient mode.
ambient has been testedambient is the mode OLIVE is developed and tested against. sidecar is not thoroughly tested: the charts render the policy shape it needs, but no OLIVE release has been validated end to end on a sidecar mesh, so treat it as unsupported in practice. none is tested, since it is what a development deployment uses.
If your cluster already runs sidecars and moving it to ambient is not an option, raise it with OLIVE support at olive-support@boc-group.com before you plan the deployment rather than after.
Why ambient mode
In the traditional sidecar model, every pod runs an additional Envoy proxy. In a microservice architecture with many horizontally scaled workloads, that is one extra container per replica, with the CPU and memory cost that implies.
Ambient mode runs a ztunnel as a DaemonSet — one proxy per node rather than one per pod. Because a cluster has far fewer nodes than pods, the resource footprint is substantially smaller. For workloads enrolled in the ambient mesh, ztunnel provides exactly what OLIVE needs:
- mTLS encryption between workloads
- workload identity and authentication
- secure communication within and across nodes
Ambient also keeps the mesh transparent to the OLIVE charts. Workloads are enrolled through namespace or workload configuration; the application charts do not need to know about mTLS, certificates or proxies. Nothing is injected, so no workload restarts on enrollment.
Waypoint proxies — the additional Istio component for Layer 7 features such as HTTP-level authorization, retries or traffic routing — are not required for in-cluster encryption and are not currently used.
Configuration
global:
serviceMesh:
mode: ambient # none | sidecar | ambient
vaultMeshed: true
vault:
namespace: 'oli-hashicorp-vault'
serviceName: 'oli-vault'
port: 8200
scheme: 'http'
| Mode | Behaviour |
|---|---|
none | No mesh. Charts render exactly as they did before any mesh work existed. In-cluster traffic is not encrypted. |
ambient | A node-local ztunnel captures traffic and tunnels it over HBONE on TCP 15008. Every NetworkPolicy rule whose peer is a meshed pod also allows 15008. Nothing is injected; no workload restarts. |
sidecar | Not thoroughly tested — see the warning above. An Envoy proxy is injected into every pod. Traffic stays on the application port, so no NetworkPolicy port changes are needed — but every pod's proxy must reach istiod on 15012, and injection rewrites the pod spec, so every workload in the namespace restarts. |
The two meshed modes require opposite NetworkPolicy changes. Setting ambient on a cluster running sidecars (or the reverse) produces policies that allow the wrong ports. The failure mode is a timeout, not an error message. Set it once in the shared values file, never per chart, and confirm with your cluster operator which mode is actually installed.
vaultMeshed
HBONE is used only when both ends are meshed. Vault is one cluster-wide deployment enrolled on its own schedule, separately from the tenant namespaces, so it has its own flag:
global:
serviceMesh:
vaultMeshed: false # until the Vault namespace carries istio.io/dataplane-mode=ambient
Leave it false until the Vault namespace is actually enrolled, then set it true.
The vault-agent-injector is a mutating admission webhook called by the kube-apiserver. The apiserver is never in the mesh and has no SPIFFE identity, so a mesh-wide STRICT PeerAuthentication will reject it — and pod creation then fails in every tenant namespace that runs CouchDB or SeaweedFS — which is every tenant, because their pod templates carry the vault.hashicorp.com/agent-inject annotation. Apply the PERMISSIVE exemption for the injector before turning on strict peer authentication.
Activating the mesh
- Confirm with the cluster operator that Istio ambient is installed and which namespaces are enrolled. See Cluster Operator → Service mesh and mTLS.
- Make sure network policies are already stable with
serviceMesh.mode: none. Enabling both at once makes a failure impossible to attribute. - Set
global.serviceMesh.mode: ambientin the shared values file so the shared runtime pool and every tenant get it together. - Sync. No workloads restart in ambient mode.
- Verify that probes still pass and that traffic is actually tunnelled.
Deactivating the mesh
global:
serviceMesh:
mode: none
vaultMeshed: false
Charts revert to the non-mesh rendering and in-cluster traffic is no longer encrypted. Coordinate with the cluster operator: removing the namespace's ambient label is their step, and doing one without the other leaves policies and data plane disagreeing.
Network policies still apply
Ambient mode does not bypass Kubernetes NetworkPolicy. Policies are enforced by the CNI, independently of Istio, and Istio documents that ambient respects them. What changes is the network path: traffic between ambient workloads is tunnelled over HBONE on TCP 15008, so a policy that says "only allow port 8080 into this pod" would inadvertently block mesh traffic. The charts add 15008 automatically when serviceMesh.mode: ambient, which is why the value must be correct.
What mTLS does and does not give you
It gives you: encryption of in-cluster traffic, workload identity, and mutual authentication between services — so "data is protected in transit" holds inside the cluster too, not only at the edge.
It does not give you: encryption at rest. Volumes, backups, object storage and etcd depend on the cloud environment and are the cluster operator's responsibility.