Platform Technology Overview
| Function | Selected Technology | Version | Purpose |
|---|---|---|---|
| Kubernetes Distribution | RKE2 (1.33.8+rke2r1) and Amazon EKS — both tested on this release | see note | Provides the Kubernetes control plane and worker infrastructure. Any CNCF-conformant Kubernetes that meets the requirements below should work; these two are the ones this release was tested on. |
| Identity & Access Management | Keycloak | 26.7.5 | Central identity provider (IdP) for authentication and identity federation (OIDC/SAML). |
| Secrets Management | HashiCorp Vault | 2.0.4 | Centralized secrets management and secure secret injection into workloads. |
| Database Operator (PostgreSQL) | CloudNativePG (CNPG) operator | 1.27.1 | Manages PostgreSQL clusters (currently used for Temporal). |
| Workflow Orchestration | Temporal | 1.31.4 | Orchestrates datapoints and provides full workflow execution capabilities; chosen to support reliable orchestration and long-running workflows. |
| Storage | Longhorn Storage | 1.8.0 | Provides persistent volume provisioning, backup, and disaster recovery via PVCs; avoids per-storage backup solutions using Longhorn-managed snapshots and S3 backups. Any other StorageClass works as well — see Storage Requirements. |
| Ingress | Traefik | Managed via RKE2, or installed separately (e.g. on EKS) | Handles external traffic routing, request handling, and authentication integration at the ingress level. Traefik is supported starting with OLIVE platform version 1.2.0. See Ingress Requirements for a Traefik that is not RKE2's bundled one. |
| TLS & Certificate Management | cert-manager (Jetstack) | 1.18.2 | Automates TLS certificate provisioning and management within the cluster (e.g. via Let’s Encrypt or a custom CA). |
| Ingress Authentication Bridge | OAuth2 Proxy | 7.15.3 | Handles authentication flows at the ingress level and integrates with Keycloak via OIDC. |
| Session Storage (Auth) | Redis | 7.4.11-bookworm | Stores session data and full JWT tokens for OAuth2 Proxy. |
| Compute Runtime (Node.js) | Node.js (Distroless image) | nodejs24-debian13 (nonroot) | Primary runtime for compute workloads (e.g. Express, TypeScript services). |
| Compute Runtime (Python) | Python (Distroless image) | python3-debian13 (nonroot) | Runtime for Python-based components; currently the auth-proxy session-purge job. |
| NoSQL Document Storage | CouchDB | 3.5.2.1 | JSON-based NoSQL document store (e.g. OLIVE connectivity MM repository instances). |
| Object Storage | SeaweedFS | 4.44 | S3-compatible object storage for large files and connectivity job artifacts; used for the file storage service and large connectivity job entries. |
| Service Mesh (optional) | Istio, ambient mode | Cluster-provided | Provides mutual TLS, workload identity and authentication between in-cluster workloads via a per-node ztunnel DaemonSet. Enabled per deployment with global.serviceMesh.mode: ambient. See Service Mesh & mTLS. |
| Admission Policy Engine (optional) | Kyverno | Cluster-provided (kyverno-infra chart) | Enforces container image signature verification and the image registry allow-list at admission, and hosts operator-written policies. See Admission Control & Image Integrity. |
| Artefact Signing | Sigstore cosign | Build pipeline | Signs the container images OLIVE builds; signatures and attestations are stored as OCI artefacts alongside the image. |
What OLIVE requires of Vault, Keycloak, Traefik, the Istio mesh configuration and Kyverno is documented regardless of how you install them. Optional setup charts are also available for all five, provided as-is and outside the platform's support and maintenance. See Infrastructure Charts.
Kubernetes 1.31 or later is required.
RKE2 and Amazon EKS are the distributions this release was tested on. OLIVE is a standard Kubernetes application and does not depend on a distribution: any CNCF-conformant cluster that meets the requirements in this guide — the CRDs, a CNI that enforces NetworkPolicy, an RWO StorageClass, Traefik with the redirectErrors plugin — should run it. Running on something else is not unsupported, but it is unvalidated, so budget time to confirm those requirements rather than assuming them.
The chart defaults describe RKE2 with its bundled Traefik and Longhorn. On EKS and elsewhere, review global.ingress.className, global.networkPolicy.ingressController and global.storageClassName before the first sync — see Cluster-specific settings.
OLIVE's network policies are only effective if the CNI enforces them. Cilium and Calico do. See Networking.
Kubernetes resources
For the API groups OLIVE creates objects in, which CRDs have to exist before it will sync, the cluster-scoped objects it creates, and a ready-made Argo CD AppProject allow-list, see Kubernetes Resources.
Container images
Every container image OLIVE deploys — the third-party ones it redistributes and the ones it builds itself — is referenced by SHA-256 digest, not by tag. A tag is a mutable pointer: anyone able to write to the registry can repoint it at different content, and every cluster that pulls it then runs something nobody reviewed, with no change in the manifest and nothing in the deployment history to show it. A digest names the exact artefact, and a substituted image simply fails to match.
The table below lists the versions in this release and which chart pulls each image. The digests pinned at release are recorded per release on its release notes page — for the current release, Container images in this release.
The digests themselves are deliberately not reproduced here. They change with every rebuild of an upstream image, and a digest copied into documentation is wrong within weeks while still looking precise. The authoritative value for any deployment is the digest pinned in that release's chart values.
Use this table to know what is deployed and where it comes from; read the chart values to know exactly which artefact.
Third-party images
| Image | Version | Used by |
|---|---|---|
quay.io/keycloak/keycloak | 26.7.5 | keycloak-infra |
quay.io/oauth2-proxy/oauth2-proxy | v7.15.3 | auth-proxy |
redis | 7.4.11-bookworm | auth-proxy |
hashicorp/vault | 2.0.4 | couchdb, keycloak-infra, keycloak-setup, seaweedfs, vault-infra, vault-secrets-setup |
hashicorp/vault-k8s | 1.7.6 | vault-infra |
couchdb | 3.5.2.1 | couchdb |
chrislusf/seaweedfs | 4.44 | seaweedfs |
temporalio/server | 1.31.4 | workflow-engine-postgres |
temporalio/admin-tools | 1.31.4 | workflow-engine-postgres |
temporalio/ui | 2.57.0 | workflow-engine-postgres — development only; disabled by default in production |
alpine/curl | 8.21.0 | couchdb, seaweedfs |
alpine | 3.22.5 | seaweedfs |
alpine/k8s | 1.36.5 | vault-secrets-setup |
docker.io/busybox | 1.38.0 | keycloak-infra |
The images above fall into two groups, and the answer to "can we move this to a newer version ourselves?" differs between them.
Keycloak, Vault and Vault Secrets Operator (quay.io/keycloak/keycloak, hashicorp/vault, hashicorp/vault-k8s) are not part of the OLIVE platform. They are infrastructure OLIVE integrates with, and the infrastructure charts that install them are optional and provided as-is. Those versions are yours to manage: if a CVE is published and the fix is upstream, pin the newer image and roll it out without waiting for an OLIVE release. Keep it within the same major version — OLIVE depends on the behaviour of the Keycloak and Vault APIs it talks to, and a major upgrade can change them. Pin the replacement by digest rather than tag, for the reasons above.
Everything else in the table ships inside the OLIVE platform charts — CouchDB, Redis, SeaweedFS, Temporal, and the small helper images. Some of these can technically be overridden in values.yaml, but doing so is not recommended and not supported. These versions are part of the combination the release was tested as a whole; replacing one on your own can produce inconsistencies and behaviour that is hard to diagnose, and nothing in the release validates the result.
If one of them needs a newer version — a CVE, a compliance finding, an upstream fix you depend on — raise it with OLIVE support at olive-support@boc-group.com rather than overriding it. Depending on the component and the reason, the answer is either a supported override for your deployment or an updated platform release.
The base images under Images OLIVE builds cannot be changed at deployment time at all: they are baked into OLIVE's service images at build time, so moving them requires a rebuild.
Images OLIVE builds
| Image | Version | Used by | Note |
|---|---|---|---|
gcr.io/distroless/nodejs24-debian13 | nonroot | All Express services | Runtime base image |
gcr.io/distroless/python3-debian13 | nonroot | auth-proxy | Runtime base image for the session-purge job |
node | 24.22.1-slim | — | Build stage only, in some services |
node | 24.22.1-alpine3.24 | — | Build stage only, in some services |
job-secrets-image | 0.0.2 | keycloak-setup, schedule-trigger-cron-job | Purpose-built job image, published by OLIVE |
job-secrets-image exists because those jobs used to start from a generic base image as root and install curl, jq, coreutils and gettext at runtime. See Workload Hardening for why that pattern was replaced.
Image digest pinning does not extend to Helm chart dependencies, which cannot currently be pinned by OCI digest. If a chart tag were mutable, the chart could be replaced and reference different images. The risk is reduced because OLIVE's charts live in an OCI registry under its control. Mirroring third-party images and charts into a registry of your own would remove the dependency on external registries, and some organisations choose to; it is not a documented or supported OLIVE procedure. See Admission Control & Image Integrity.