Configure the platform
The KoiosPlatform resource describes the whole platform. This page covers its settings topic by
topic; every field is listed in the KoiosPlatform reference. The
koios-installer repository has an example for each common case in kustomize/platform/examples/
and Secret templates in common/secrets/; each example is a patch over the minimal platform in
kustomize/platform/base.
The operator reconciles the resource continuously: change the KoiosPlatform, not the Deployments,
Services or configuration files it creates, as it puts them back.
Address and TLS
The platform is reached through the platform gateway, published by a Route on OpenShift or an
Ingress on Kubernetes (ingress_type: Route, Ingress or none; the default is Route on
OpenShift and none on Kubernetes).
OpenShift
: The Route is created by default. route_host sets its host (empty lets OpenShift assign one),
and route_tls_secret gives the certificate (tls.crt, tls.key and optionally ca.crt). TLS
terminates at the Route (Edge); passthrough is not supported.
Kubernetes
: Set ingress_type: Ingress, the public hostname and ingress_tls_secret, a Secret with
tls.crt and tls.key that you create in the platform namespace (example ingress).
ingress_class_name selects the IngressClass (empty uses the cluster default) and
ingress_annotations adds annotations, for example for cert-manager.
Without an Ingress, service_type publishes the gateway Service as LoadBalancer or NodePort
(nodeport_port); set hostname to the address users reach. public_base_url sets the public base
URL of the platform when it is not the one of the Route, Ingress or load balancer, for example
behind an external proxy.
The address the platform uses is in status.URL. Automation controller uses its own Base URL of
the service setting (TOWER_URL_BASE) for the links in notifications. From 2026.11.12 the operator sets it to that
address while the setting still has its default value (https://platformhost) or the value the operator set before,
and keeps it in step when the address changes; it is not changed when the platform has no public address. To use
another address, set it in Settings > Automation Execution > System: the operator does not change a value set by an
administrator.
Database
By default the operator deploys and manages a PostgreSQL 15 for the whole platform, with one
database and user per component. database.storage_class and database.storage_size choose its
volume.
To use your own PostgreSQL (15 or later), create one database and user per component (the database
of automation hub needs the hstore extension) and give each component a Secret with its connection
(template common/secrets/database.example.yaml, example external-db):
Component |
Field |
|---|---|
Platform gateway |
|
Automation controller |
|
Event-Driven Ansible |
|
Event-Driven Ansible event streams |
|
Automation hub |
|
For a verified TLS connection, set sslmode: verify-full (or verify-ca) in the Secrets and trust
the CA of the server with bundle_cacert_secret. The managed PostgreSQL has no TLS.
The database mode (database.managed) cannot change on an existing platform. To move a platform
between a managed and an external database, back it up and restore it into a new platform (Backup
and restore).
Redis
The gateway uses Redis. By default the operator deploys a single Valkey pod (redis_mode: standalone). redis_mode: cluster deploys six Valkey nodes, three masters and three replicas
(example redis-cluster).
To use your own Redis instead, give its connection in redis.redis_secret (template
common/secrets/redis.example.yaml): host, port, password and optional username, redis_tls,
redis_ssl_cert_reqs and redis_mode (standalone or cluster). The operator then does not deploy
Redis. With TLS and a private CA, trust the CA with bundle_cacert_secret. Two platforms must not
share one Redis.
Components
Automation controller is always deployed. Event-Driven Ansible and automation hub are deployed
unless eda.disabled: true or hub.disabled: true; disabling one keeps its resource and removes it
from the gateway.
controller, eda and hub also take the fields of the upstream resource of each component
(AWX, EDA and Galaxy), which the operator merges with the ones the platform requires; see the
reference.
Automation hub storage
Automation hub stores content on a ReadWriteMany volume by default, so that its pods can run on
any node. Without a ReadWriteMany StorageClass (for example k3s local-path), use a single-node
volume with hub.file_storage_access_mode: ReadWriteOnce (example single-node-storage).
Scheduling and resources
Each part of the platform has its own placement and resources:
gateway:
api.replicas,api.resource_requirements,api.node_selector,api.tolerationsandapi.topology_spread_constraints;managed PostgreSQL:
database.resource_requirements,database.node_selector,database.tolerationsanddatabase.priority_class;managed Redis:
redis.resource_requirements,redis.node_selectorandredis.tolerations;automation controller, Event-Driven Ansible and automation hub: the fields of their upstream resources under
controller,edaandhub.
Administrator and keys
The platform administrator is admin_user (admin by default), created once. Its password is
generated in the Secret <platform>-admin-password, or taken from your own Secret named in
admin_password_secret (key password).
The platform gateway encrypts the secrets it stores with a key generated in
<platform>-db-fields-encryption, or taken from db_fields_encryption_secret (key secret_key).
Automation controller, Event-Driven Ansible and automation hub have keys of their own, also generated
by the operator. Do not change these keys, neither the Secret a platform uses nor its value: the data
encrypted with them would become unreadable. Backups include them (Backup and
restore).
Settings and feature flags
extra_settings passes settings to the platform gateway and controller.extra_settings to
automation controller, as setting/value pairs (Troubleshooting
has an example). feature_flags sets gateway feature flags (FEATURE_* keys); the operator
validates every key.
Idle mode
idle_aap: true scales the platform down without losing data: the gateway, Redis and the managed
PostgreSQL go to zero (the database stays up with database.idle_disabled: true), and automation
controller, Event-Driven Ansible and automation hub are idled. Set it back to false to resume.
Image pull secrets
image_pull_secrets names existing Secrets to pull the images of the private registry, and the
operator passes them to the components. The pods of jobs and activations use the default
ServiceAccount of the platform namespace: the installer’s common/pull-secrets.sh attaches the pull
Secret to it. Container groups in other namespaces and execution nodes outside the cluster need
their own access to the registry. See Registry access.