This repo will be deprecated and unsupported starting April 2022
Discontinuing this operator in favor of recommending ExternalSecrets which is widely used.
EncryptedResource
Razeedeploy: component to transport and decrypt secret Kubernetes resources. Currently supports PGP keys and encrypted resources.
Note: testing has been done using and RSA key generated by gpg (GnuPG) 2.2.16
Install
kubectl apply -f "https://github.com/razee-io/EncryptedResource/releases/latest/download/resource.yaml"
Resource Definition
Sample
apiVersion: "deploy.razee.io/v1alpha2"
kind: EncryptedResource
metadata:
name: <name>
namespace: <namespace>
spec:
clusterAuth:
impersonateUser: razeedeploy
auth:
privateKeyRef:
valueFrom:
secretKeyRef:
name: <secret_name>
namespace: <secret_namespace>
key: <secret_key>
passphraseRef: # optional
valueFrom:
secretKeyRef:
name: <secret_name>
namespace: <secret_namespace>
key: <secret_key>
resources: # must be base64 encoded
- <encrypted-base64_encoded-resource>
Spec
Path: .spec
Description: spec
is required and must include sections resources
and auth
.
Schema:
spec:
type: object
required: [resources, auth]
properties:
clusterAuth:
type: object
...
auth:
type: object
...
resources:
type: array
...
Resources
Path: .spec.resources[]
Description: Resources to be decrypted and applied to the cluster. There must be at least one resource in the list.
Note: All resources must be base64 encoded, whether they are ascii armored or not.
Schema:
resources:
type: array
minItems: 1
items:
type: string
User Impersonation
Path: .spec.clusterAuth.impersonateUser
Description: Impersonates a user for the given resource. This includes all actions the controller must make related to the resource (fetching envs, getting resources, applying resources, etc.). The RazeeDeploy resource must be created in the razeedeploy namespace in order to use impersonateUser, all other namespaces will ignore impersonateUser and default to the razeedeploy user (eg. no user impersonation). ImpersonateUser only applies to the single RazeeDeploy resource that it has been added to.
Note:: If cluster owners want to prevent users, with direct cluster access, from using user-impersonation, they should prevent those users from creating RazeeDeploy resources in the razeedeploy namespace. In the future we will have an Admission Controller that should improve security and eliminate the need for the razeedeploy namespace scoping. razeedeploy-core #189
Schema:
properties:
clusterAuth:
type: object
properties:
impersonateUser:
type: string
Default: 'razeedeploy'
Authentication
Path: .spec.auth
Description: Authentication to be able to decrypt the resources. Auth must include a private key, but optionally can also include a passphrase that the private key is encrypted with. You may specify the key/passphrase as a secret reference or hardcoded as a string (hardcoded strings are not recommended, but added to assist in development and testing).
Note: If using hardcoded strings, they must be base64 encoded, whether they are ascii armored or not.
Schema:
auth:
type: object
oneOf:
- required: [privateKey]
- required: [privateKeyRef]
properties:
privateKey:
type: string
privateKeyRef:
type: object
required: [valueFrom]
...
passphrase:
type: string
passphraseRef:
type: object
required: [valueFrom]
...
Managed Resource Labels
Reconcile
Child resource: .metadata.labels[deploy.razee.io/Reconcile]
- DEFAULT:
true
- A razeedeploy resource (parent) will clean up a resources it applies (child) when either the child is no longer in the parent resource definition or the parent is deleted.
-
false
- This behavior can be overridden when a child's resource definition has
the label
deploy.razee.io/Reconcile=false
.
- This behavior can be overridden when a child's resource definition has
the label
Resource Update Mode
Child resource: .metadata.labels[deploy.razee.io/mode]
Razeedeploy resources default to merge patching children. This behavior can be
overridden when a child's resource definition has the label
deploy.razee.io/mode=<mode>
Mode options:
- DEFAULT:
MergePatch
- A simple merge, that will merge objects and replace arrays. Items previously defined, then removed from the definition, will be removed from the live resource.
- "As defined in RFC7386, a Merge Patch is essentially a partial representation of the resource. The submitted JSON is "merged" with the current resource to create a new one, then the new one is saved. For more details on how to use Merge Patch, see the RFC." Reference
-
StrategicMergePatch
- A more complicated merge, the kubernetes apiServer has defined keys to be able to intelligently merge arrays it knows about.
- "Strategic Merge Patch is a custom implementation of Merge Patch. For a detailed explanation of how it works and why it needed to be introduced, see StrategicMergePatch." Reference
- Kubectl Apply Semantics
-
EnsureExists
- Will ensure the resource is created and is replaced if deleted. Will not enforce a definition.
Debug Individual Resource
.spec.resources.metadata.labels[deploy.razee.io/debug]
Treats the live resource as EnsureExist. If any razeedeploy component is enforcing
the resource, and the label deploy.razee.io/debug: true
exists on the live
resource, it will treat the resource as ensure exist and not override any changes.
This is useful for when you need to debug a live resource and don't want razeedeploy
overriding your changes. Note: this will only work when you add it to live resources.
If you want to have the EnsureExist behavior, see Resource Update Mode.
- ie:
kubectl label rr <your-rr> deploy.razee.io/debug=true
Lock Cluster Updates
Prevents the controller from updating resources on the cluster. If this is the
first time creating the razeedeploy-config
ConfigMap, you must delete the running
controller pods so the deployment can mount the ConfigMap as a volume. If the
razeedeploy-config
ConfigMap already exists, just add the pair lock-cluster: true
.
export CONTROLLER_NAME=encryptedresource-controller && export CONTROLLER_NAMESPACE=razee
kubectl create cm razeedeploy-config -n $CONTROLLER_NAMESPACE --from-literal=lock-cluster=true
kubectl delete pods -n $CONTROLLER_NAMESPACE $(kubectl get pods -n $CONTROLLER_NAMESPACE | grep $CONTROLLER_NAME | awk '{print $1}' | paste -s -d ',' -)