SKS-2345: Add support for system disk expansion (#169)

api/v1beta1/conditions_consts.go

@@ -121,6 +121,29 @@ const (
 	// WaitingForAvailableHostWithEnoughGPUsReason (Severity=Info) documents an ElfMachine
 	// waiting for an available host with enough GPUs to create VM.
 	WaitingForAvailableHostWithEnoughGPUsReason = "WaitingForAvailableHostWithEnoughGPUs"
+
+	// ResourcesHotUpdatedCondition documents the status of the hot updating resources of a VM.
+	ResourcesHotUpdatedCondition = "ResourceHotUpdated"
+
+	// WaitingForResourcesHotUpdateReason (Severity=Info) documents an ElfMachine waiting for updating resources.
+	WaitingForResourcesHotUpdateReason = "WaitingForResourcesHotUpdate"
+
+	// ExpandingVMDiskReason documents (Severity=Info) ElfMachine currently executing the expand disk operation.
+	ExpandingVMDiskReason = "ExpandingVMDisk"
+
+	// ExpandingVMDiskFailedReason (Severity=Warning) documents an ElfMachine controller detecting
+	// an error while expanding disk; those kind of errors are usually transient and failed updating
+	// are automatically re-tried by the controller.
+	ExpandingVMDiskFailedReason = "ExpandingVMDiskFailed"
+
+	// ExpandingRootPartitionReason documents (Severity=Info) ElfMachine currently executing the
+	// adding new disk capacity to root directory operation.
+	ExpandingRootPartitionReason = "ExpandingRootPartition"
+
+	// ExpandingRootPartitionFailedReason (Severity=Warning) documents an ElfMachine controller
+	// detecting an error while adding new disk capacity to root directory; those kind of errors are
+	// usually transient and failed updating are automatically re-tried by the controller.
+	ExpandingRootPartitionFailedReason = "ExpandingRootPartitionFailed"
 )
 
 // Conditions and Reasons related to make connections to a Tower. Can currently be used by ElfCluster and ElfMachine

api/v1beta1/elfmachine_types.go

@@ -22,6 +22,7 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	clusterv1 "sigs.k8s.io/cluster-api/api/v1beta1"
 	capierrors "sigs.k8s.io/cluster-api/errors"
+	"sigs.k8s.io/cluster-api/util/conditions"
 )
 
 const (
@@ -37,6 +38,9 @@ const (
 
 	// VMDisconnectionTimestampAnnotation is the annotation identifying the VM of ElfMachine disconnection time.
 	VMDisconnectionTimestampAnnotation = "cape.infrastructure.cluster.x-k8s.io/vm-disconnection-timestamp"
+
+	// VMFirstBootTimestampAnnotation is the annotation identifying the VM of ElfMachine first power on time.
+	VMFirstBootTimestampAnnotation = "cape.infrastructure.cluster.x-k8s.io/vm-first-boot-timestamp"
 )
 
 // ElfMachineSpec defines the desired state of ElfMachine.
@@ -124,6 +128,10 @@ type ElfMachineStatus struct {
 	// +optional
 	GPUDevices []GPUStatus `json:"gpuDevices,omitempty"`
 
+	// Resources records the resources allocated for the machine.
+	// +optional
+	Resources ResourcesStatus `json:"resources,omitempty"`
+
 	// FailureReason will be set in the event that there is a terminal problem
 	// reconciling the Machine and will contain a succinct value suitable
 	// for machine interpretation.
@@ -241,6 +249,16 @@ func (m *ElfMachine) IsFailed() bool {
 	return m.Status.FailureReason != nil || m.Status.FailureMessage != nil
 }
 
+// IsHotUpdating returns whether the machine is being hot updated.
+func (m *ElfMachine) IsHotUpdating() bool {
+	if conditions.Has(m, ResourcesHotUpdatedCondition) &&
+		conditions.IsFalse(m, ResourcesHotUpdatedCondition) {
+		return true
+	}
+
+	return false
+}
+
 func (m *ElfMachine) SetVMDisconnectionTimestamp(timestamp *metav1.Time) {
 	if m.Annotations == nil {
 		m.Annotations = make(map[string]string)
@@ -318,6 +336,35 @@ func (m *ElfMachine) GetVMDisconnectionTimestamp() *metav1.Time {
 	return nil
 }
 
+func (m *ElfMachine) SetVMFirstBootTimestamp(timestamp *metav1.Time) {
+	annotations := m.GetAnnotations()
+	if annotations == nil {
+		annotations = map[string]string{}
+	}
+	m.Annotations[VMFirstBootTimestampAnnotation] = timestamp.Format(time.RFC3339)
+	m.SetAnnotations(annotations)
+}
+
+func (m *ElfMachine) GetVMFirstBootTimestamp() *metav1.Time {
+	if m.Annotations == nil {
+		return nil
+	}
+
+	if _, ok := m.Annotations[VMFirstBootTimestampAnnotation]; ok {
+		timestampAnnotation := m.Annotations[VMFirstBootTimestampAnnotation]
+		timestamp, err := time.Parse(time.RFC3339, timestampAnnotation)
+		if err != nil {
+			return nil
+		}
+
+		firstBootTimestamp := metav1.NewTime(timestamp)
+
+		return &firstBootTimestamp
+	}
+
+	return nil
+}
+
 func (m *ElfMachine) RequiresGPUDevices() bool {
 	return m.RequiresPassThroughGPUDevices() || m.RequiresVGPUDevices()
 }

api/v1beta1/types.go

@@ -196,6 +196,11 @@ type GPUStatus struct {
 	Name  string `json:"name,omitempty"`
 }
 
+// ResourcesStatus records the resources allocated to the virtual machine.
+type ResourcesStatus struct {
+	Disk int32 `json:"disk,omitempty"`
+}
+
 //+kubebuilder:object:generate=false
 
 // PatchStringValue is for patching resources.

config/crd/bases/infrastructure.cluster.x-k8s.io_elfmachines.yaml

@@ -403,6 +403,13 @@ spec:
               ready:
                 description: Ready is true when the provider resource is ready.
                 type: boolean
+              resources:
+                description: Resources records the resources allocated for the machine.
+                properties:
+                  disk:
+                    format: int32
+                    type: integer
+                type: object
               taskRef:
                 description: TaskRef is a managed object reference to a Task related
                   to the machine. This value is set automatically at runtime and should

config/crd/bases/kubesmart.smtx.io_hostconfigs.yaml

@@ -0,0 +1,153 @@
+---
+apiVersion: apiextensions.k8s.io/v1
+kind: CustomResourceDefinition
+metadata:
+  annotations:
+    controller-gen.kubebuilder.io/version: v0.12.0
+  name: hostconfigs.kubesmart.smtx.io
+spec:
+  group: kubesmart.smtx.io
+  names:
+    categories:
+    - kubesmart
+    kind: HostConfig
+    listKind: HostConfigList
+    plural: hostconfigs
+    shortNames:
+    - hc
+    singular: hostconfig
+  scope: Namespaced
+  versions:
+  - additionalPrinterColumns:
+    - description: the current phase of HostConfig
+      jsonPath: .status.phase
+      name: Phase
+      type: string
+    - description: the last execution time
+      jsonPath: .status.lastExecutionTime
+      name: LastExecutionTime
+      type: string
+    - description: Time duration since creation of HostConfig
+      jsonPath: .metadata.creationTimestamp
+      name: Age
+      type: date
+    name: v1alpha1
+    schema:
+      openAPIV3Schema:
+        description: HostConfig is the Schema for the HostConfig API.
+        properties:
+          apiVersion:
+            description: 'APIVersion defines the versioned schema of this representation
+              of an object. Servers should convert recognized schemas to the latest
+              internal value, and may reject unrecognized values. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources'
+            type: string
+          kind:
+            description: 'Kind is a string value representing the REST resource this
+              object represents. Servers may infer this from the endpoint the client
+              submits requests to. Cannot be updated. In CamelCase. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds'
+            type: string
+          metadata:
+            type: object
+          spec:
+            properties:
+              config:
+                properties:
+                  ansible:
+                    description: Ansible 通过 ansible playbook 完成配置
+                    properties:
+                      localPlaybook:
+                        description: LocalPlaybook 本地的 playbook，单个 yaml 文件， secret
+                          引用或者 yaml 字符串
+                        properties:
+                          content:
+                            description: Content is the inline yaml text.
+                            format: yaml
+                            type: string
+                          secretRef:
+                            description: SecretRef specifies the secret which stores
+                              yaml text.
+                            properties:
+                              name:
+                                description: name is unique within a namespace to
+                                  reference a secret resource.
+                                type: string
+                              namespace:
+                                description: namespace defines the space within which
+                                  the secret name must be unique.
+                                type: string
+                            type: object
+                            x-kubernetes-map-type: atomic
+                        type: object
+                      remotePlaybook:
+                        description: RemotePlaybook 在远端的 playbook，单个 .tar.gz 压缩包，内容可以是单个
+                          yaml 文件，也可以符合 ansible 要求的目录
+                        properties:
+                          md5sum:
+                            description: MD5sum 压缩包的 MD5，填写了会进行校验，已经下载过的 playbook
+                              校验通过后跳过重复下载
+                            type: string
+                          name:
+                            description: Name 要执行的 playbook 文件名，相对于压缩包顶层的位置
+                            type: string
+                          url:
+                            description: URL playbook 在远端的地址，支持 https
+                            type: string
+                        required:
+                        - name
+                        - url
+                        type: object
+                      values:
+                        description: Values 执行 playbook 的参数，yaml 格式，可以是 secret 引用或者
+                          yaml 字符串
+                        properties:
+                          content:
+                            description: Content is the inline yaml text.
+                            format: yaml
+                            type: string
+                          secretRef:
+                            description: SecretRef specifies the secret which stores
+                              yaml text.
+                            properties:
+                              name:
+                                description: name is unique within a namespace to
+                                  reference a secret resource.
+                                type: string
+                              namespace:
+                                description: namespace defines the space within which
+                                  the secret name must be unique.
+                                type: string
+                            type: object
+                            x-kubernetes-map-type: atomic
+                        type: object
+                    type: object
+                  timeout:
+                    description: Timeout 执行一次配置的超时时间
+                    type: string
+                type: object
+              nodeName:
+                type: string
+            required:
+            - config
+            - nodeName
+            type: object
+          status:
+            properties:
+              failureMessage:
+                type: string
+              failureReason:
+                type: string
+              lastExecutionTime:
+                description: LastExecutionTime 最后执行的时间戳
+                format: date-time
+                type: string
+              phase:
+                description: Phase 当前状态
+                type: string
+            required:
+            - phase
+            type: object
+        type: object
+    served: true
+    storage: true
+    subresources:
+      status: {}