Deckhouse Kubernetes Platform includes a set of templates for creating projects:
defaultis a template for basic project use cases:- resource limits
- network isolation
- automatic alerts and log collection
- selecting a security profile
- configuring project administrators
secureincludes all capabilities of thedefaulttemplate and adds:- configuring allowed UID/GID ranges for the project
- audit rules for Linux users’ interactions with the kernel
- scanning container images at runtime for known vulnerabilities (CVEs)
secure-with-dedicated-nodesincludes all capabilities of thesecuretemplate and adds:- defining a node selector for all pods in the project: when a pod is created, its node selector is automatically replaced with the project’s node selector
- defining default tolerations for all pods in the project: when a pod is created, the default tolerations are automatically added to it
To list all available parameters for a project template, run:
d8 k get projecttemplates <PROJECT_TEMPLATE_NAME> -o jsonpath='{.spec.parametersSchema.openAPIV3Schema}' | jq
Project creation
- To create a project, create a custom resource Project with the project template name specified in the .spec.projectTemplateName field.
-
In the .spec.parameters parameter, specify the values for the .spec.parametersSchema.openAPIV3Schema section of the custom resource ProjectTemplate.
An example of creating a project using Project from the
defaultProjectTemplate is shown below:apiVersion: deckhouse.io/v1alpha2 kind: Project metadata: name: my-project spec: description: This is an example from the Deckhouse documentation. projectTemplateName: default parameters: resourceQuota: requests: cpu: 5 memory: 5Gi storage: 1Gi limits: cpu: 5 memory: 5Gi networkPolicy: Isolated podSecurityProfile: Restricted extendedMonitoringEnabled: true administrators: - subject: Group name: k8s-admins -
To check the project status, run the following command:
d8 k get projects my-projectA successfully created project should have the
Deployed(synced) status. If the status isError, add the-o yamlflag to the command (for example,d8 k get projects my-project -o yaml) to get more details about the error cause.
Automatic project creation for a namespace
You can create a new project for an existing namespace. To do this, add the projects.deckhouse.io/adopt annotation to the namespace. For example:
-
Create a new namespace:
d8 k create ns test -
Annotate it with the following command:
d8 k annotate ns test projects.deckhouse.io/adopt="" -
Make sure the project has been created:
d8 k get projectsA new project matching the namespace will appear in the list of projects:
NAME STATE PROJECT TEMPLATE DESCRIPTION AGE deckhouse Deployed virtual This is a virtual project 181d default Deployed virtual This is a virtual project 181d test Deployed empty 1m
You can change the template of an existing project to another available template.
Note that changing the template may cause resource conflicts: if the template chart defines resources that already exist in the namespace, the template cannot be applied.
Creating a custom project template
The default project templates cover common baseline scenarios and also serve as examples of what templates can do.
To create your own template:
- Use one of the default templates as a starting point, for example,
default. -
Export it to a separate file, for example,
my-project-template.yaml, using the following command:d8 k get projecttemplates default -o -
Edit the
my-project-template.yamlfile and apply the required changes.You must update not only the template itself, but also the input parameters schema to match it.
Project templates support all Helm templating functions.
-
Change the template name in the
.metadata.namefield. -
Apply the resulting template with the following command:
d8 k apply -f my-project-template.yaml -
Check that the new template is available by running:
d8 k get projecttemplates <NEW_TEMPLATE_NAME>
Using labels to manage resources
When creating resources in ProjectTemplate, you can use special labels to control how the multitenancy-manager processes these resources.
Skipping creation of the heritage: multitenancy-manager label
By default, all resources created from ProjectTemplate receive the label heritage: multitenancy-manager.
This label prohibits changes to resources by users or any other controller except multitenancy-manager.
If you need to allow resource modification (for example, for compatibility with other systems, or if implementing your own control over the created objects), add the label projects.deckhouse.io/skip-heritage-label to the resource.
Example:
apiVersion: v1
kind: ConfigMap
metadata:
name: my-config
namespace: {{ .projectName }}
labels:
projects.deckhouse.io/skip-heritage-label: "true"
app: my-app
data:
key: value
In this case, the resource will receive the labels projects.deckhouse.io/project and projects.deckhouse.io/project-template, but will not receive the label heritage: multitenancy-manager.
Excluding resources from management by multitenancy-manager
If you need to exclude a resource from management by multitenancy-manager (for example, if the resource should be managed manually or by another controller), add the label projects.deckhouse.io/unmanaged to the resource.
Example:
apiVersion: v1
kind: Secret
metadata:
name: external-secret
namespace: {{ .projectName }}
labels:
projects.deckhouse.io/unmanaged: "true"
type: Opaque
data:
token: <base64-encoded-value>
Resources with the label projects.deckhouse.io/unmanaged:
- Will be created only once when the project is created;
- Will not be updated with subsequent template changes or updates;
- Will not be monitored in the project’s status;
- Will receive the labels
projects.deckhouse.io/projectandprojects.deckhouse.io/project-templatebut will not receive the labelheritage: multitenancy-manager.
Once a resource is marked as unmanaged, it will be created on initial installation but not updated when the ProjectTemplate is changed.
After creation, the resource becomes fully independent and must be managed manually.
Implementing validation of object changes with a custom label
The multitenancy-manager module uses ValidatingAdmissionPolicy to protect resources labeled heritage: multitenancy-manager from manual changes.
You can implement similar validation for resources with any label.
How validation works in multitenancy-manager
Validation occurs for objects labeled heritage: multitenancy-manager.
The following components are used for this:
- ValidatingAdmissionPolicy: Defines validation rules:
- Operations:
UPDATEandDELETE. - Check: only operations on behalf of the controller’s service account are allowed.
- Applies to all resources and API groups.
- Operations:
- ValidatingAdmissionPolicyBinding: Defines which objects the validation applies to:
- Uses
namespaceSelectorandobjectSelectorto select resources by the labelheritage: multitenancy-manager.
- Uses
Creating your own validation
To implement validation for resources with a different label (for example, heritage: my-custom-label):
-
Create a file with the ValidatingAdmissionPolicy and ValidatingAdmissionPolicyBinding resource manifests:
apiVersion: admissionregistration.k8s.io/v1 kind: ValidatingAdmissionPolicy metadata: name: my-custom-label-validation spec: failurePolicy: Fail matchConstraints: resourceRules: - apiGroups: ["*"] apiVersions: ["*"] operations: ["UPDATE", "DELETE"] resources: ["*"] scope: "*" validations: - expression: 'request.userInfo.username == "system:serviceaccount:my-namespace:my-service-account"' # Replace with your service account reason: Forbidden messageExpression: 'object.kind == ''Namespace'' ? ''This resource is managed by '' + object.metadata.name + '' system. Manual modification is forbidden.'' : ''This resource is managed by '' + object.metadata.namespace + '' system. Manual modification is forbidden.''' --- apiVersion: admissionregistration.k8s.io/v1 kind: ValidatingAdmissionPolicyBinding metadata: name: my-custom-label-validation spec: policyName: my-custom-label-validation validationActions: [Deny, Audit] matchResources: namespaceSelector: matchLabels: heritage: my-custom-label objectSelector: matchLabels: heritage: my-custom-label -
Configure the validation parameters:
policyName: Unique policy name (must match in Policy and Binding).request.userInfo.username: The name of the service account allowed to change resources (replace with your service account).heritage: my-custom-label: The value of theheritagelabel for your resources (replace with your value). The use of the valuesmultitenancy-manager,deckhouseis prohibited.failurePolicy: Fail: Policy on validation failure.Fail: Reject the request on validation failure.Ignore: Ignore validation errors.
validationActions: Validation actions:Deny: Deny unauthorized operations.Audit: Record operations in the audit log.
-
Apply the policy:
d8 k apply -f my-validation-policy.yaml -
Ensure your resources have the corresponding
heritagelabel:apiVersion: v1 kind: ConfigMap metadata: name: my-resource labels: heritage: my-custom-label
Granting cluster-scoped resources to projects
The multitenancy-manager lets cluster administrators control which cluster-scoped resources (for example StorageClass) may be referenced from within project namespaces.
To do this, custom resources are used:
- GrantableClusterResourceDefinition (cluster-scoped) — registers a cluster resource that can be
granted to projects: which resource it is (
grantedResource), where references to it are validated (usageReferences), the baseline availability (defaultAvailability), and how the per-project default is discovered (defaultFrom). Each reference opts into defaulting individually withdefault: true— set it only for a field whose value the resource always needs (such as aPersistentVolumeClaim’sstorageClassName). Leave it off for a reference whose absence is meaningful, such as an annotation that merely toggles a feature; that reference is still validated and counted, just never filled in. - ClusterResourceGrantPolicy (cluster-scoped) — selects projects (by namespace labels via
projectSelector) and, per resource (resourceName), the granted names (allowed,allowedSelector) and the per-projectdefault. An allow-list restricts the resource to it. - AvailableClusterResource (namespaced, read-only, short name
available) — the controller-rendered catalog of what a project may use; tenants read it to discover the available names. - ClusterResourceGrant (namespaced) — the per-project object-quota pool (limits on object count and on measured quantities such as requested storage); its status reports current usage.
apiVersion: multitenancy.deckhouse.io/v1alpha1
kind: GrantableClusterResourceDefinition
metadata:
name: storageclasses
spec:
grantedResource:
apiGroup: storage.k8s.io
kind: StorageClass
enforcement: Managed
defaultAvailability: All
defaultFrom:
annotationKey: storageclass.kubernetes.io/is-default-class
usageReferences:
- rule:
apiGroups:
- ""
apiVersions:
- v1
resources:
- persistentvolumeclaims
fieldPath: $.spec.storageClassName
default: true
---
apiVersion: multitenancy.deckhouse.io/v1alpha1
kind: ClusterResourceGrantPolicy
metadata:
name: production-storage
spec:
projectSelector:
matchLabels:
environment: production
resources:
- resourceName: storageclasses
default: fast-ssd # Overrides the annotation-based default.
allowed:
- fast-ssd
- standard
allowedSelector: # Plus any StorageClass with label shared=true.
matchLabels:
shared: "true"
Enforcement notes:
- The validating webhook denies creating/updating objects in matched projects whose referenced value is not granted. On update, values already present in the object are grandfathered in, so pre-existing objects are not broken.
- The defaulting webhook fills in the granted default on creation only, and only into references
marked
default: true. References left without it (such as feature-toggling annotations) are never filled in. - A grant that matches no project, or a project with no matching grant, imposes no restriction.