Server-Side Apply

Introduction

Server Side Apply helps users and controllers manage their resources through declarative configurations. Clients can create and modify their objects declaratively by sending their fully specified intent.

A fully specified intent is a partial object that only includes the fields and values for which the user has an opinion. That intent either creates a new object or is combined, by the server, with the existing object.

The system supports multiple appliers collaborating on a single object.

Changes to an object's fields are tracked through a "field management" mechanism. When a field's value changes, ownership moves from its current manager to the manager making the change. When trying to apply an object, fields that have a different value and are owned by another manager will result in a conflict. This is done in order to signal that the operation might undo another collaborator's changes. Conflicts can be forced, in which case the value will be overridden, and the ownership will be transferred.

If you remove a field from a configuration and apply the configuration, server side apply checks if there are any other field managers that also own the field. If the field is not owned by any other field managers, it is either deleted from the live object or reset to its default value, if it has one. The same rule applies to associative list or map items.

Server side apply is meant both as a replacement for the original kubectl apply and as a simpler mechanism for controllers to enact their changes.

If you have Server Side Apply enabled, the control plane tracks managed fields for all newly created objects.

Field Management

Compared to the last-applied annotation managed by kubectl, Server Side Apply uses a more declarative approach, which tracks a user's field management, rather than a user's last applied state. This means that as a side effect of using Server Side Apply, information about which field manager manages each field in an object also becomes available.

For a user to manage a field, in the Server Side Apply sense, means that the user relies on and expects the value of the field not to change. The user who last made an assertion about the value of a field will be recorded as the current field manager. This can be done either by changing the value with POST, PUT, or non-apply PATCH, or by including the field in a config sent to the Server Side Apply endpoint. When using Server-Side Apply, trying to change a field which is managed by someone else will result in a rejected request (if not forced, see Conflicts).

When two or more appliers set a field to the same value, they share ownership of that field. Any subsequent attempt to change the value of the shared field, by any of the appliers, results in a conflict. Shared field owners may give up ownership of a field by removing it from their configuration.

Field management is stored in amanagedFields field that is part of an object's metadata.

A simple example of an object created by Server Side Apply could look like this:

apiVersion: v1
kind: ConfigMap
metadata:
  name: test-cm
  namespace: default
  labels:
    test-label: test
  managedFields:
  - manager: kubectl
    operation: Apply
    apiVersion: v1
    time: "2010-10-10T0:00:00Z"
    fieldsType: FieldsV1
    fieldsV1:
      f:metadata:
        f:labels:
          f:test-label: {}
      f:data:
        f:key: {}
data:
  key: some value

The above object contains a single manager in metadata.managedFields. The manager consists of basic information about the managing entity itself, like operation type, API version, and the fields managed by it.

Nevertheless it is possible to change metadata.managedFields through an Update operation. Doing so is highly discouraged, but might be a reasonable option to try if, for example, the managedFields get into an inconsistent state (which clearly should not happen).

The format of the managedFields is described in the API.

Conflicts

A conflict is a special status error that occurs when an Apply operation tries to change a field, which another user also claims to manage. This prevents an applier from unintentionally overwriting the value set by another user. When this occurs, the applier has 3 options to resolve the conflicts:

• Overwrite value, become sole manager: If overwriting the value was intentional (or if the applier is an automated process like a controller) the applier should set the force query parameter to true and make the request again. This forces the operation to succeed, changes the value of the field, and removes the field from all other managers' entries in managedFields.

• Don't overwrite value, give up management claim: If the applier doesn't care about the value of the field anymore, they can remove it from their config and make the request again. This leaves the value unchanged, and causes the field to be removed from the applier's entry in managedFields.

• Don't overwrite value, become shared manager: If the applier still cares about the value of the field, but doesn't want to overwrite it, they can change the value of the field in their config to match the value of the object on the server, and make the request again. This leaves the value unchanged, and causes the field's management to be shared by the applier and all other field managers that already claimed to manage it.

Managers

Managers identify distinct workflows that are modifying the object (especially useful on conflicts!), and can be specified through the fieldManager query parameter as part of a modifying request. It is required for the apply endpoint, though kubectl will default it to kubectl. For other updates, its default is computed from the user-agent.

Apply and Update

The two operation types considered by this feature are Apply (PATCH with content type application/apply-patch+yaml) and Update (all other operations which modify the object). Both operations update the managedFields, but behave a little differently.

For instance, only the apply operation fails on conflicts while update does not. Also, apply operations are required to identify themselves by providing a fieldManager query parameter, while the query parameter is optional for update operations. Finally, when using the apply operation you cannot have managedFields in the object that is being applied.

An example object with multiple managers could look like this:

apiVersion: v1
kind: ConfigMap
metadata:
  name: test-cm
  namespace: default
  labels:
    test-label: test
  managedFields:
  - manager: kubectl
    operation: Apply
    apiVersion: v1
    fields:
      f:metadata:
        f:labels:
          f:test-label: {}
  - manager: kube-controller-manager
    operation: Update
    apiVersion: v1
    time: '2019-03-30T16:00:00.000Z'
    fields:
      f:data:
        f:key: {}
data:
  key: new value

In this example, a second operation was run as an Update by the manager called kube-controller-manager. The update changed a value in the data field which caused the field's management to change to the kube-controller-manager.

If this update would have been an Apply operation, the operation would have failed due to conflicting ownership.

Merge strategy

The merging strategy, implemented with Server Side Apply, provides a generally more stable object lifecycle. Server Side Apply tries to merge fields based on the actor who manages them instead of overruling based on values. This way multiple actors can update the same object without causing unexpected interference.

When a user sends a "fully-specified intent" object to the Server Side Apply endpoint, the server merges it with the live object favoring the value in the applied config if it is specified in both places. If the set of items present in the applied config is not a superset of the items applied by the same user last time, each missing item not managed by any other appliers is removed. For more information about how an object's schema is used to make decisions when merging, see sigs.k8s.io/structured-merge-diff.

A number of markers were added in Kubernetes 1.16 and 1.17, to allow API developers to describe the merge strategy supported by lists, maps, and structs. These markers can be applied to objects of the respective type, in Go files or in the OpenAPI schema definition of the CRD:

If listType is missing, the API server interprets a patchMergeStrategy=merge marker as a listType=map and the corresponding patchMergeKey marker as a listMapKey.

The atomic list type is recursive.

These markers are specified as comments and don't have to be repeated as field tags.

Compatibility across topology changes

On rare occurrences, a CRD or built-in type author may want to change the specific topology of a field in their resource without incrementing its version. Changing the topology of types, by upgrading the cluster or updating the CRD, has different consequences when updating existing objects. There are two categories of changes: when a field goes from map/set/granular to atomic and the other way around.

When the listType, mapType, or structType changes from map/set/granular to atomic, the whole list, map, or struct of existing objects will end-up being owned by actors who owned an element of these types. This means that any further change to these objects would cause a conflict.

When a list, map, or struct changes from atomic to map/set/granular, the API server won't be able to infer the new ownership of these fields. Because of that, no conflict will be produced when objects have these fields updated. For that reason, it is not recommended to change a type from atomic to map/set/granular.

Take for example, the custom resource:

apiVersion: example.com/v1
kind: Foo
metadata:
  name: foo-sample
  managedFields:
  - manager: manager-one
    operation: Apply
    apiVersion: example.com/v1
    fields:
      f:spec:
        f:data: {}
spec:
  data:
    key1: val1
    key2: val2

Before spec.data gets changed from atomic to granular, manager-one owns the field spec.data, and all the fields within it (key1 and key2). When the CRD gets changed to make spec.data granular, manager-one continues to own the top-level field spec.data (meaning no other managers can delete the map called data without a conflict), but it no longer owns key1 and key2, so another manager can then modify or delete those fields without conflict.

Custom Resources

By default, Server Side Apply treats custom resources as unstructured data. All keys are treated the same as struct fields, and all lists are considered atomic.

If the Custom Resource Definition defines a schema that contains annotations as defined in the previous "Merge Strategy" section, these annotations will be used when merging objects of this type.

Using Server-Side Apply in a controller

As a developer of a controller, you can use server-side apply as a way to simplify the update logic of your controller. The main differences with a read-modify-write and/or patch are the following:

• the applied object must contain all the fields that the controller cares about.
• there is no way to remove fields that haven't been applied by the controller before (controller can still send a PATCH/UPDATE for these use-cases).
• the object doesn't have to be read beforehand, resourceVersion doesn't have to be specified.

It is strongly recommended for controllers to always "force" conflicts, since they might not be able to resolve or act on these conflicts.

Transferring Ownership

In addition to the concurrency controls provided by conflict resolution, Server Side Apply provides ways to perform coordinated field ownership transfers from users to controllers.

This is best explained by example. Let's look at how to safely transfer ownership of the replicas field from a user to a controller while enabling automatic horizontal scaling for a Deployment, using the HorizontalPodAutoscaler resource and its accompanying controller.

Say a user has defined deployment with replicas set to the desired value:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
  labels:
    app: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx:1.14.2

And the user has created the deployment using server side apply like so:

kubectl apply -f https://k8s.io/examples/application/ssa/nginx-deployment.yaml --server-side

Then later, HPA is enabled for the deployment, e.g.:

kubectl autoscale deployment nginx-deployment --cpu-percent=50 --min=1 --max=10

Now, the user would like to remove replicas from their configuration, so they don't accidentally fight with the HPA controller. However, there is a race: it might take some time before HPA feels the need to adjust replicas, and if the user removes replicas before the HPA writes to the field and becomes its owner, then apiserver will set replicas to 1, its default value. This is not what the user wants to happen, even temporarily.

There are two solutions:

• (basic) Leave replicas in the configuration; when HPA eventually writes to that field, the system gives the user a conflict over it. At that point, it is safe to remove from the configuration.

• (more advanced) If, however, the user doesn't want to wait, for example because they want to keep the cluster legible to coworkers, then they can take the following steps to make it safe to remove replicas from their configuration:

First, the user defines a new configuration containing only the replicas field:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
spec:
  replicas: 3

The user applies that configuration using the field manager name handover-to-hpa:

kubectl apply -f https://k8s.io/examples/application/ssa/nginx-deployment-replicas-only.yaml \
  --server-side --field-manager=handover-to-hpa \
  --validate=false

If the apply results in a conflict with the HPA controller, then do nothing. The conflict indicates the controller has claimed the field earlier in the process than it sometimes does.

At this point the user may remove the replicas field from their configuration.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
  labels:
    app: nginx
spec:
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx:1.14.2

Note that whenever the HPA controller sets the replicas field to a new value, the temporary field manager will no longer own any fields and will be automatically deleted. No clean up is required.

Transferring Ownership Between Users

Users can transfer ownership of a field between each other by setting the field to the same value in both of their applied configs, causing them to share ownership of the field. Once the users share ownership of the field, one of them can remove the field from their applied configuration to give up ownership and complete the transfer to the other user.

Comparison with Client Side Apply

A consequence of the conflict detection and resolution implemented by Server Side Apply is that an applier always has up to date field values in their local state. If they don't, they get a conflict the next time they apply. Any of the three options to resolve conflicts results in the applied configuration being an up to date subset of the object on the server's fields.

This is different from Client Side Apply, where outdated values which have been overwritten by other users are left in an applier's local config. These values only become accurate when the user updates that specific field, if ever, and an applier has no way of knowing whether their next apply will overwrite other users' changes.

Another difference is that an applier using Client Side Apply is unable to change the API version they are using, but Server Side Apply supports this use case.

Upgrading from client-side apply to server-side apply

Client-side apply users who manage a resource with kubectl apply can start using server-side apply with the following flag.

kubectl apply --server-side [--dry-run=server]

By default, field management of the object transfers from client-side apply to kubectl server-side apply without encountering conflicts.

This behavior applies to server-side apply with the kubectl field manager. As an exception, you can opt-out of this behavior by specifying a different, non-default field manager, as seen in the following example. The default field manager for kubectl server-side apply is kubectl.

kubectl apply --server-side --field-manager=my-manager [--dry-run=server]

Downgrading from server-side apply to client-side apply

If you manage a resource with kubectl apply --server-side, you can downgrade to client-side apply directly with kubectl apply.

Downgrading works because kubectl server-side apply keeps the last-applied-configuration annotation up-to-date if you use kubectl apply.

This behavior applies to server-side apply with the kubectl field manager. As an exception, you can opt-out of this behavior by specifying a different, non-default field manager, as seen in the following example. The default field manager for kubectl server-side apply is kubectl.

kubectl apply --server-side --field-manager=my-manager [--dry-run=server]

API Endpoint

With the Server Side Apply feature enabled, the PATCH endpoint accepts the additional application/apply-patch+yaml content type. Users of Server Side Apply can send partially specified objects as YAML to this endpoint. When applying a configuration, one should always include all the fields that they have an opinion about.

Clearing ManagedFields

It is possible to strip all managedFields from an object by overwriting them using MergePatch, StrategicMergePatch, JSONPatch, or Update, so every non-apply operation. This can be done by overwriting the managedFields field with an empty entry. Two examples are:

PATCH /api/v1/namespaces/default/configmaps/example-cm
Content-Type: application/merge-patch+json
Accept: application/json
Data: {"metadata":{"managedFields": [{}]}}

PATCH /api/v1/namespaces/default/configmaps/example-cm
Content-Type: application/json-patch+json
Accept: application/json
Data: [{"op": "replace", "path": "/metadata/managedFields", "value": [{}]}]

This will overwrite the managedFields with a list containing a single empty entry that then results in the managedFields being stripped entirely from the object. Note that setting the managedFields to an empty list will not reset the field. This is on purpose, so managedFields never get stripped by clients not aware of the field.

In cases where the reset operation is combined with changes to other fields than the managedFields, this will result in the managedFields being reset first and the other changes being processed afterwards. As a result the applier takes ownership of any fields updated in the same request.