Skip to main content

Infrastructure Charts

OLIVE depends on infrastructure it does not contain: an identity provider, a secrets manager, an ingress controller, and — optionally — a service mesh and an admission policy engine. These have to exist and be configured in particular ways before the platform charts will work.

To make that easier, five Helm charts are made available that set them up. They are not part of the OLIVE platform, and they are not mandatory.

ChartSets upRequired?
vault-infraHashiCorp Vault, initialised, unsealed and bootstrapped for OLIVEYes — OLIVE cannot run without a Vault
keycloak-infraKeycloak and its PostgreSQL databaseYes for authenticated deployments
service-mesh-setupCluster-wide Istio mesh configurationOptional
kyverno-infraKyverno and the OLIVE admission policies — three policy families (image signatures, registry allow-list, pod privileges)Optional
traefik-infraTraefik behind an AWS NLB, with the required redirectErrors plugin vendoredOptional — only where Traefik is not already installed, typically EKS. On RKE2 you add the plugin to the bundled Traefik instead

Between them, the four charts an OLIVE cluster typically installs — Vault, Keycloak, Kyverno and Traefik — reserve 3.4 cores and 3.7 GiB of requests in steady state. The per-chart and per-pod figures are in Measured Resource Footprint.

Why these pages are in the cluster operator guide​

Installing these components is cluster work, and it needs cluster rights. Between them they install CRDs, cluster-scoped webhooks, ClusterRoles and an IngressClass, write into kube-system and the mesh root namespace, and need resources that live outside Kubernetes entirely — a KMS key, IAM roles, an S3 bucket, a load balancer, DNS, a publicly trusted certificate.

A Platform Administrator generally does not hold those rights, and should not need to. This matches the responsibility matrix, which already assigns the engine, the mesh and the Vault instance to the Cluster Operator.

HalfOwned byExamples
The infrastructureCluster operatorEverything on these five pages: installing Istio and Kyverno, the KMS key and IAM credentials, storage classes for Vault's volumes, PostgreSQL, DNS and a certificate for the Keycloak host, a CNI that enforces NetworkPolicy, Argo CD and its AppProject destinations, and every cluster-scoped object — CRDs, ClusterRoles, admission webhooks
The OLIVE contractPlatform AdministratorWhat has to be true afterwards: the Vault bootstrap policy and role, the KV mount being named secret, the global.vault address block, the Keycloak issuer matching what the services expect, realm and MFA policy, the Kyverno keyless subject and the enrolment label keys

In short: operated by the cluster operator, specified by the platform administrator. The infrastructure half is ordinary cluster work. The contract half is OLIVE-specific and means nothing to a cluster operator who has not read it — which is why it is written down separately, in Infrastructure Prerequisites.

Start from the platform administrator's handover

The Administrator's prerequisites page is the checklist of what each component must provide. Read it alongside these pages: it says what has to be true, these say one way of making it true.

Chart versions in this release​

These are the versions published alongside the current OLIVE release; a wrapper chart that still names an older one keeps deploying that older one.

ChartVersionInstalls
vault-infra0.7.14Vault 2.0.4 (upstream chart 0.33.0), Vault Secrets Operator 1.5.1
keycloak-infra1.4.9Keycloak 26.7.5 (codecentric chart 7.2.3)
service-mesh-setup0.5.3configuration only — Istio is installed separately
kyverno-infra1.0.10Kyverno 1.19.0 (upstream chart 3.9.0)
traefik-infra0.1.5Traefik v3.7.13 (upstream chart 41.5.0)
These charts are not part of the OLIVE platform

Read this before using them.

They are not a product. They are not part of the OLIVE platform, not covered by the platform release, and not part of what OLIVE delivers, supports or maintains. They are shared because they save work, in the spirit of "here it is, take it or leave it".

They are provided as-is. There is no commitment that they are maintained, patched, updated for new versions of the components they install, or fixed when something in them breaks. They may lag behind the upstream components. They may lag behind OLIVE itself.

Security updates for them are not tracked. The components they install — HashiCorp Vault, Keycloak, Istio, Kyverno — are third-party software under their own licences, with their own release and security processes. Watching those projects for advisories, and updating your installation, is yours to do whether or not you used these charts.

Questions about the components go to their projects. How Vault should be sealed, how Keycloak federates an identity provider, how Istio is tuned, how Traefik is configured — these are questions for the upstream documentation.

What is part of the platform: the requirements​

The distinction that matters is between the charts and the contract.

Status
What OLIVE requires of Vault, Keycloak, Traefik, the mesh and KyvernoPart of the platform. Documented, versioned with the release, and something you can hold OLIVE to. If a requirement is wrong, incomplete or changes between releases, that is an OLIVE issue
The five setup chartsNot part of the platform. A convenience, as-is, unsupported

Use them, or don't​

If your organisation already runs a Keycloak, already has a hardened Vault, already operates a service mesh, or has its own opinion about how Kyverno should be installed — use what you have. What matters is that certain things are true afterwards, not how they came to be true.

If the charts work for you, use them. If they do not fit your environment, correct them, replace them, or install the components yourself. Nobody is expected to adopt them to get support for the platform — only to meet the contract.

Using the charts is not a support boundary

Choosing not to use these charts does not reduce support for the OLIVE platform, and choosing to use them does not extend support to cover them.

Versions have to fit

Whichever route you take, the component versions must match what the OLIVE release was validated against. See 1.3.0 → Compatibility Matrix for the tested versions, and Platform Technology Overview for what each component is for.

All five are Argo CD-friendly​

Each chart is designed to be deployed and managed as its own Argo CD Application — one per component, because they are independent and installed on their own schedules. (The OLIVE platform itself is the opposite: one Application for the whole olive-platform umbrella chart, since that chart carries the ordering internally. See How to lay out the Applications.)

The usual pattern is a thin wrapper chart per component that declares the OLIVE chart as a dependency:

apiVersion: v2
name: vault-infra-wrapper
description: Helm chart for setting up hashicorp vault
type: application
version: 1.0.0
appVersion: '1.0'
dependencies:
- name: vault-infra
version: 0.7.14
repository: oci://registry.gitlab.com/olive6224315/helm-charts/charts
condition: services.global.vault-infra

with a values.yaml beside it that switches the dependency on and configures it:

services:
global:
vault-infra: true

The condition is what lets one repository layout carry all of them and enable only the ones a given cluster needs. Each page shows a complete working example.

All five create cluster-scoped objects — CRDs, ClusterRoles, webhooks, an IngressClass — so a restricted AppProject refuses them with ... is not permitted in project. What each one needs permitted is on its own page; the platform's own project is a different list, in Argo CD AppProject.

Shared values​

The blocks below appear in more than one of these charts and in the platform charts themselves. Keep them in one shared values file that every deployment includes — a cluster where two charts disagree about where Vault lives, or about which mesh mode is installed, fails in ways that look like something else entirely.

The Administrator owns what goes in here; see Infrastructure Prerequisites.

global:
vault:
namespace: 'oli-hashicorp-vault'
serviceName: 'oli-vault'
port: 8200
scheme: 'http'
serviceMesh:
mode: ambient # none | sidecar | ambient
vaultMeshed: true
# The three Kyverno enrolment label keys. Separate on purpose — a namespace
# can enforce one control while still only auditing another.
imageVerification:
labelKey: 'olive/image-verification'
imageRegistry:
labelKey: 'olive/image-registry'
podPrivileges:
labelKey: 'olive/pod-privileges'
# Node confinement. Set appArmor false on hosts without AppArmor (SELinux —
# RHEL, Rocky, OpenShift — and some managed clouds), where the kubelet
# otherwise rejects the pods with "Cannot enforce AppArmor". seccomp stays on.
appArmor: true
seccomp: true
# Master switch for NetworkPolicy across the charts in a namespace. Before
# turning it on, confirm the CNI enforces NetworkPolicy and that the
# ingress-controller selector matches real pods — a selector that matches
# nothing is not an error, it is zero peers.
networkPolicy:
enabled: false
ingressController:
namespace: kube-system # 'traefik' for a separate install
podLabels:
app.kubernetes.io/name: rke2-traefik # 'traefik' upstream — no rke2- prefix
# StorageClass for every volume in the namespace, including the CloudNativePG
# clusters. '' leaves each chart's own value; 'default' omits the field so
# Kubernetes picks the default class; otherwise name one ('longhorn', 'gp3').
storageClassName: ''
Existing installs and storageClassName

A StatefulSet's volumeClaimTemplates are immutable. Charts that previously rendered longhorn explicitly must keep getting longhorn — rendering anything else makes the StatefulSet update fail.

Order of installation​

  1. Traefik, if it is not already installed — Keycloak needs an ingress to be reachable at its public host name.
  2. Vault — everything else stores credentials in it, including Keycloak's admin password.
  3. Keycloak — needs Vault to exist first; its admin credential is seeded into Vault and read back from there.
  4. Service mesh configuration — after Istio itself is installed, and before you enrol namespaces.
  5. Kyverno — any time; it evaluates namespaces that carry the enrolment label.

Then the OLIVE platform charts are deployed, which is the platform administrator's step.