2019-08-02 11:06:26 +02:00
|
|
|
# Design
|
|
|
|
|
|
|
|
|
|
## Controller design
|
|
|
|
|
|
2022-09-22 07:52:52 -05:00
|
|
|
The controller listens for Custom Resource which defines client registration
|
|
|
|
|
request. Once Custom resource is created, the controller register oauth2 client
|
|
|
|
|
in hydra using hydra's REST API. Client Id, Client Secret and Identifier of the
|
|
|
|
|
client in hydra are be stored in the kubernetes as a secret and referenced in
|
|
|
|
|
the applied CR. Reference is used to identify in which kubernetes secret are
|
|
|
|
|
stored mentioned properties. Secret iscreated in the same namespace of applied
|
|
|
|
|
CR. By default controller should be deployed in the same pod as hydra. Service
|
|
|
|
|
discovery will come in place in the future.
|
2019-08-02 11:06:26 +02:00
|
|
|
|
2022-09-22 07:52:52 -05:00
|
|
|
Custom Resource should be Namespace scoped to enable isolation in k8s. It is
|
|
|
|
|
represented in the diagram
|
2019-08-02 11:12:44 +02:00
|
|
|
|
|
|
|
|
|
2019-08-02 11:06:26 +02:00
|
|
|
|
|
|
|
|
## Synchronization mode
|
|
|
|
|
|
2022-09-22 07:52:52 -05:00
|
|
|
Additionally, controller supports synchronization mode, where it tries to
|
|
|
|
|
register all clients in hydra. Synchronization is an optional mode, enabled via
|
|
|
|
|
config, which is meant for use cases where hydra is deployed with in memory
|
|
|
|
|
storage. If hydra pod is restarted for some reason then it does not have client
|
|
|
|
|
in its storage. With synchronization mode the controller makes sure that hydra
|
|
|
|
|
has up to date clients. Synchronization is done by making POST request to hydra
|
|
|
|
|
with payload describing all client information including clientID,clientSecret
|
|
|
|
|
and Identifier of last applied client. If client exists in hydra storage 409 is
|
|
|
|
|
returned which is considered as ok and synchronization continues with other
|
|
|
|
|
clients.
|
|
|
|
|
|
|
|
|
|
|
