Skip to main content
Version: Unreleased 🚧

RBAC

Botkube allows plugins to access Kubernetes API by defining RBAC rules. Based on this configuration Botkube generates a temporary kubeconfig with user and/or group impersonation.

Architecture

Botkube uses its own cluster credentials to generate a temporary kubeconfig, and the kubeconfig uses user/group impersonation.

For source plugins, the kubeconfig is generated once - during plugin startup. For executor plugins, the kubeconfig is generated every time a command is sent to the plugin, which allows greater flexibility, such as including the name of the channel the command was sent from.

diagram

This kubeconfig is available to plugins in the Execute and Stream methods, as long as the plugin has the context.rbac property defined.

Configuration

Botkube Cloud

In Botkube Cloud, the RBAC is configured within the "Permissions" tab for each plugin.

Cloud RBAC defaults

Self-hosted Botkube installation

For each executor and source plugin, you can define a context.rbac configuration. This config is used to generate a dedicated kubeconfig.

executors:
"kubectl":
botkube/kubectl@v1:
enabled: true
# ...
context:
rbac:
user:
type: Static # Static or ChannelName
static: # applicable only for "Static" user mapping type
value: botkube-internal-static-user
prefix: "" # optional prefix for user name; useful especially for channel name mapping
group:
type: Static # Static or ChannelName
static: # applicable only for "Static" group mapping type
values:
- "my-group1"
- "my-group2"
prefix: "" # optional prefix for all group names; useful especially for channel name mapping

Mapping types

For both user and group, the following mapping types are supported:

  • Static

    For user, it uses a single static value. For group, it uses a list of static values. The value is prepended with an optional prefix.

  • ChannelName

    Channel name is used as subject for user or group impersonation. The channel name is prepended with an optional prefix. This mapping is only available for executor plugins.

Default configuration

When a given plugin have context.rbac property undefined, Botkube doesn't generate a kubeconfig for this plugin. To request kubeconfig generation, define context.rbac property with empty value:

executors:
"kubectl":
botkube/kubectl@v1:
enabled: true
# ...
context:
rbac: {} # enable kubeconfig generation

However, such configuration will generate a kubeconfig with empty impersonation config, which effectively means an anonymous access to the Kubernetes API.

During Botkube installation, Botkube generates Kubernetes ClusterRole and ClusterRoleBinding resources with read-only access for the default group botkube-plugins-default. This group is used by default across the values.yaml for all default plugins.

rbac:
# ...
groups:
"botkube-plugins-default":
create: true
rules:
- apiGroups: ["*"]
resources: ["*"]
verbs: ["get", "watch", "list"]

See the values.yaml for more details.

Defaults for user mapping when group mapping is used

Kubernetes requires user for group impersonation. That's why when a group mapping is user without context.rbac.user mapping defined, Botkube uses botkube-internal-static-user user name for impersonation. For example, when the following configuration is used:

executors:
"kubectl":
botkube/kubectl@v1:
enabled: true
# ...
context:
rbac:
# no user mapping defined
group:
type: Static
static:
value: botkube-plugins-default

It is equivalent to:

executors:
"kubectl":
botkube/kubectl@v1:
enabled: true
# ...
context:
rbac:
user:
type: Static
static:
value: botkube-internal-static-user
group:
type: Static
static:
value: botkube-plugins-default

Defaults for Botkube Cloud

When configuring plugin on Botkube Cloud, the "Default" permissions mean that the botkube-plugins-default group will be used, which have read-only access to Kubernetes API and is configured during Botkube installation. See the Default configuration section.

Cloud RBAC defaults

Examples

This paragraph contains examples of RBAC configuration for different use cases.

tip

You can use rbac.groups or extraObjects overrides during Botkube installation to create custom RBAC resources. See the values.yaml for more details.

Kubectl executor with read-only Pod access based on static group mapping

In this example an executor plugin is defined with static RBAC that maps to group read-pods.

  1. Consider the following self-hosted Botkube config:

    # ...
    executors:
    "kubectl-read-only":
    botkube/kubectl@v1:
    enabled: true
    # ...
    context:
    rbac:
    group:
    type: Static
    static:
    values: [read-pods]

Let's assume this plugin is bound to at least one channel.

  1. Consider the following Kubernetes RBAC configuration:

    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRole
    metadata:
    name: kubectl-read-pods
    rules:
    - apiGroups: [""]
    resources: ["pods"]
    verbs: ["get", "watch", "list"]
    ---
    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRoleBinding
    metadata:
    name: kubectl-read-pods
    roleRef:
    apiGroup: rbac.authorization.k8s.io
    kind: ClusterRole
    name: kubectl-read-pods
    subjects:
    - kind: Group
    name: read-pods # <-- this is the group name used in Botkube config
    apiGroup: rbac.authorization.k8s.io

In a result, when this executor plugin is invoked, Botkube generates a kubeconfig impersonating group read-pods and passes it to the plugin. The plugin then can authenticate with the API server with identity of group read-pods. In that way, the plugin can use read-only operations on Pods.

Kubernetes source plugin with read-only access based on static user mapping

In this example a single source plugin is defined with static RBAC that maps to user kubernetes-read-only.

  1. Consider the following self-hosted Botkube config:

    sources:
    "kubernetes":
    botkube/kubernetes@v1:
    enabled: true
    # ...
    context:
    rbac:
    user:
    type: Static
    static:
    value: kubernetes-read-only
  2. Consider the following Kubernetes RBAC configuration:

    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRole
    metadata:
    name: reader
    rules:
    - apiGroups: ["*"]
    resources: ["*"]
    verbs: ["get", "watch", "list"]
    ---
    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRoleBinding
    metadata:
    name: reader
    roleRef:
    apiGroup: rbac.authorization.k8s.io
    kind: ClusterRole
    name: reader
    subjects:
    - kind: User
    name: kubernetes-read-only # <-- this is the username used in Botkube config
    apiGroup: rbac.authorization.k8s.io

In a result, the source plugin can access all Kubernetes resources with read-only permissions.

Kubectl executor plugin with different permissions based on channel name mapping

In this example kubectl executor plugin is configured with channel name mapping and bound to two channels, ch-1 and ch-2. In Kubernetes RBAC resources, group ch-1 is given write access, while group ch-2 is given only read access.

  1. Consider the following self-hosted Botkube config:

    executors:
    "kubectl":
    botkube/kubectl@v1:
    # ...
    enabled: true
    context:
    rbac:
    group:
    type: ChannelName
    communications:
    "default-group":
    socketSlack:
    enabled: true
    # ...
    channels:
    "ch-1":
    name: ch-1
    bindings:
    executors:
    - kubectl
    "ch-2":
    name: ch-2
    bindings:
    executors:
    - kubectl
    # ...
  2. Consider the following Kubernetes RBAC configuration:

    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRole
    metadata:
    name: editor
    rules:
    - apiGroups: ["*"]
    resources: ["*"]
    verbs: ["get", "watch", "list", "update", "create", "delete"]
    ---
    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRoleBinding
    metadata:
    name: editor
    roleRef:
    apiGroup: rbac.authorization.k8s.io
    kind: ClusterRole
    name: editor
    subjects:
    - kind: Group
    name: ch-1 # <-- channel name used in Botkube config
    apiGroup: rbac.authorization.k8s.io
    ---
    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRole
    metadata:
    name: read-only
    rules:
    - apiGroups: ["*"]
    resources: ["*"]
    verbs: ["get", "watch", "list"]
    ---
    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRoleBinding
    metadata:
    name: read-only
    roleRef:
    apiGroup: rbac.authorization.k8s.io
    kind: ClusterRole
    name: read-only
    subjects:
    - kind: Group
    name: ch-2 # <-- channel name used in Botkube config
    apiGroup: rbac.authorization.k8s.io

In a result, users in channel ch-1 can execute all kubectl commands, while users in channel ch-2 can only execute read-only commands.

Limitations

This paragraph contains limitations of the current implementation.

Shared file system

Botkube runs plugin processes in the same container within the same Pod. Therefore, all plugins share the same file system.

If you're a plugin developer and decide to write kubeconfig to the file system, be aware that it can be accessible by all plugins in the container.

Supported RBAC mappings

While Executor plugins support multiple mapping types, there are the following limitations:

  • Source plugins support only the Static mapping.
  • Automated actions support only the Static mapping.

RBAC configuration merging

The same executor plugins with different RBAC configuration cannot be bound to the same channel. This is validated during Botkube startup and will result in an error.

For example, the following self-hosted Botkube configuration is invalid:

communications:
"default-group":
socketSlack:
enabled: true
# ...
channels:
"ch-1":
name: ch-1
bindings:
executors:
- kubectl
- kubectl-read-only
executors:
"kubectl":
botkube/kubectl@v1:
enabled: true
# ...
context:
rbac: # Different RBAC configuration
group:
type: ChannelName
"kubectl-read-only":
botkube/kubectl@v1:
enabled: true
# ...
context:
rbac: # Different RBAC configuration
user:
type: Static
static:
value: kubectl-read-only

Troubleshooting

In most cases troubleshooting Botkube RBAC issues means troubleshooting Kubernetes RBAC, where kubectl auth command can help.

If you see the following error:

Error: create: failed to create: secrets is forbidden: User "botkube-internal-static-user" cannot create resource "secrets" in API group "" in the namespace "default"

that means the RBAC rules configured for a given plugin are insufficient in a given context.

Firstly, ensure what user/group is used for impersonation. To do that, check your configuration against the mapping description from the Configuration section.

Checking available actions for a given user/group

After obtaining proper user and group, use the following command to list all available actions for a given user and/or group:

kubectl auth can-i --as {user} --as-group {group} --list

For example, to list all available actions for user botkube-internal-static-user and group private-channel use:

kubectl auth can-i --as botkube-internal-static-user --as-group private-channel --list

Checking if a given user/group can perform a given action

To verify if a given user and/or group can perform a given action, use:

kubectl auth can-i get pod -n botkube --as {user} --as-group {group}

For example, to verify if user botkube-internal-static-user and group private-channel can get Secret in namespace botkube use:

kubectl auth can-i get secret -n botkube --as botkube-internal-static-user --as-group private-channel

Plugin development

If you are a plugin developer and want to learn how to use generated kubeconfig in the plugin codebase, refer to Using kubeconfig document.