Skip to content

Razeedeploy: component to download and manage kubernetes resources

License

Notifications You must be signed in to change notification settings

razee-io/RemoteResource

Repository files navigation

RemoteResource

Build Status GitHub

RemoteResource is the foundation for implementing continuous deployment with razeedeploy. It retrieves and applies the configuration for all resources.

Install

Razee Deploy Delta is the recommended way to install RemoteResource.

Optional: Advanced Controller Options

Warning: By default, Razeedeploy runs as cluster wide admin. Any user that has permission to create a razeedeploy resource (RemoteResource or MustacheTemplate) has the ability to escalate their privileges. To prevent privilege escalation, cluster owners should restrict which users are allowed to create razeedeploy resources. Alternatively, you can setup ImpersonationWebhook and then enable user impersonation to ensure razeedeploy only allows users to do operations that they have already been granted access to do.

Resource Definition

Sample

apiVersion: "deploy.razee.io/v1alpha2"
kind: RemoteResource
metadata:
  name: <remote_resource_name>
  namespace: <namespace>
spec:
  clusterAuth:
    impersonateUser: razeedeploy
  backendService: generic
  requests:
    - options:
        url: https://<source_repo_url>/<file_name1>
        headers:
          <header_key1>: <header_value1>
          <header_key2>: <header_value2>
          <header_key3>:
            valueFrom:
              secretKeyRef:
                name: <name of secret resource>
                namespace: <optional namespace of secret resource>
                key: <key in secret>
    - optional: true
      options:
        url: http://<source_repo_url>/<file_name2>

Spec

Path: .spec

Description: spec is required and must include section requests.

Schema:

spec:
  type: object
  required: [requests]
  properties:
    clusterAuth:
      type: object
      ...
    backendService:
      type: string
      enum: [generic, s3, git]
    auth:
      type: object
      ...
    requests:
      type: array
      ...

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.). ImpersonateUser only applies to the single RazeeDeploy resource that it has been added to.

Enable Feature: Enable User Impersonation

Warning: By default, Razeedeploy runs as cluster wide admin. Any user that has permission to create a razeedeploy resource (RemoteResource or MustacheTemplate) has the ability to escalate their privileges. To prevent privilege escalation, cluster owners should restrict which users are allowed to create razeedeploy resources. Alternatively, you can setup ImpersonationWebhook and then enable user impersonation to ensure razeedeploy only allows users to do operations that they have already been granted access to do.

Schema:

clusterAuth:
  type: object
  properties:
    impersonateUser:
      type: string

Default: 'razeedeploy'

Backend Service

Path: .spec.backendService

Description: This field tells RemoteResource which service logic it should use. Currently the choices are generic, s3 and git, where s3 has special logic for using HMAC keys or IAM api keys for fetching access tokens. See S3 or Git.

Schema:

backendService:
  type: string
  enum: [generic, s3, git]

Default: 'generic'

Request Options

Path: .spec.requests[].options

Description: All options defined in an options object will be passed as-is to the http request. This means you can specify things like headers for authentication in this section. See S3 for authenticating with an S3 object store. See Git for configuring a git request option.

Note: You can reference secret data in your header options by setting the value of any key in the headers to be valueFrom.secretKeyRef. ie. .spec.requests[].options.headers.<yourHeaderKey> = .valueFrom.secretKeyRef.{name, namespace, key}

Schema:

options:
  type: object
  oneOf:
    - required: [url]
    - required: [uri]
    - required: [git]
  properties:
    url:
      type: string
      format: uri
    uri:
      type: string
      format: uri
    git: 
      type: object
      required: [provider, repo, branch, filePath]
      properties:
        provider:
          type: string
          enum: [github, gitlab]
        repo:
          type: string
        branch: 
          type: string
        filePath:
          type: string
    headers:
      type: object
      x-kubernetes-preserve-unknown-fields: true
    headersFrom:
      type: array
      items:
        type: object
        oneOf:
          - required: [configMapRef]
          - required: [secretMapRef]
          - required: [genericMapRef]
        properties:
          configMapRef:
            type: object
            required: [name]
            properties:
              name:
                type: string
              namespace:
                type: string
          secretMapRef:
            type: object
            required: [name]
            properties:
              name:
                type: string
              namespace:
                type: string
          genericMapRef:
            type: object
            required: [apiVersion, kind, name]
            properties:
              apiVersion:
                type: string
              kind:
                type: string
              name:
                type: string
              namespace:
                type: string

Optional Request

Path: .spec.requests[].optional

Description: if download or applying child resource fails, RemoteResource will stop execution and report error to .status. You can allow execution to continue by marking a reference as optional.

Schema:

optional:
  type: boolean

Default: false

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.

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: Apply (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
  • AdditiveMergePatch
    • Similar to the default Apply (MergePatch), this is a simple merge, that will merge objects and replace arrays. The difference is that it will not remove fields from the live resource when they are removed from the definition. eg. This will only add/update fields, it wont remove fields.
    • If you are using this mode and find that you need to remove a field, you can do so manually, by setting the field in the yaml defintion to have a value of null. When the null value is merged with the live resource, it will effectively delete the field.
    • This mode is useful if you have very large resources and require that the last-applied-configuration annotation is not injected into the resource.
  • 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

Cluster Wide Controls

The optional razeedeploy-config ConfigMap can be used to customize the controller for cluster wide actions.

Because the ConfigMap is optional, if it is created the first time, you must restart controller pods, so the deployment can mount the ConfigMap as a volume.

Example:

apiVersion: v1
kind: ConfigMap
metadata:
  name: razeedeploy-config
  namespace: razeedeploy
data:
  lock-cluster: "false"
  enable-impersonation: "false"

Lock Cluster

Key: lock-cluster

Options:

  • DEFAULT: false
    • Allows the controller to continue normal operations on the cluster.
  • true
    • Prevents the controller from updating resources on the cluster.

Enable Cluster Wide User Impersonation

Key: enable-impersonation

Options:

  • DEFAULT: false
    • Prevents the controller from performing user impersonation in all namespaces. Continues to allow user impersonation in the razeedeploy namespace. To prevent privildge escalation, users should be restricted from creating razeedeploy resources.
  • true
    • Allows the controller to perform user impersonation in all namespaces. See important note below about steps that should be taken to properly configure this feature before enabling.

IMPORTANT: it is highly advised to set up ImpersonationWebhook before enabling cluster-wide impersonation. If ImpersonationWebhook is not installed before enabling impersonation, any user on the cluster that is allowed to create razeedeploy resources will be able to impersonate any other user. Once the ImpersonationWebhook controller is installed and all necessary config and authorizations in place, impersonation can be safely enabled in the razeedeploy-config configmap.