The cluster-api-provider-metal3 (CAPM3)
is one of the controllers involved in
managing the life cycle of kubernetes clusters on Metal3 Machines. This document
describes the components involved and their roles. It also discusses the flow of
information from one CR to another with the help of the controllers. As to avoid
ambiguity, we will refer to the physical or virtual machines managed by the
controllers as Bare Metal Servers. And, the kubernetes resources (CRs)
representing them as Metal3 Machines (M3M).
The cluster-api-provider-metal3 (CAPM3)
controller is responsible for watching
and reconciling multiple resources. However, it is important to see other
controllers and custom resources (CRs) involved in the process. The ultimate
goal of the interaction between the controllers and CRs is to provision a
kubernetes cluster on Bare Metal Servers. To that end, the controllers perform
different actions on one or more relevant CRs.
The following diagram shows the different controllers and CRs involved. The
CAPI, CABPK and CAPM3 controllers are beyond the scope of this document. With
respect to CAPM3, we focus on what CRS it watches
and reconciles
. The arrows
in black show which CRs the controller reconciles
while the one in red show
that a related controller is watching
another CR.
As shown in the above Components' diagram, CAPM3 is watching for changes in
Machine
CR and upon change, it makes changes on the Metal3Machine
CR.
Similarly, it watches Cluster
CR and makes changes on a related
Metal3Cluster
CR.
The left most components, BMO controller and BareMetalHost(BMH) CR, are the closest to the Bare Metal Server. If one wants to changes the state of a Metal3 Machine, they modify the BMH CR. Upon change to the BMH, BMO interacts with Ironic to make changes on the Bare Metal Server.
During the initial introspection and state changes, the above logic works in the opposite direction as well. Information gathered during introspection or any state changes on the Bare Metal Server, results in BMO learning about the change(s) via Ironic and a chain of events starts. Once BMO learns about the changes, it makes the required changes on the BMH.
As discussed above, the management of Bare Metal Servers requires the
interaction of multiple controllers via multiple CRs. However, the interaction
is performed on a specified number of fields on each CR. i.e. A controller
watches
a specified number of fields in each CR and makes changes on a set of
fields. Before seeing the relationship, we need to see the CRs at two stages,
before and after provisioning a control plane machine.
Cluster, User provided Configuration
apiVersion: cluster.x-k8s.io/v1beta1
kind: Cluster
metadata:
name: test1
spec:
clusterNetwork:
services:
cidrBlocks: ["10.96.0.0/12"]
pods:
cidrBlocks: ["192.168.0.0/18"]
serviceDomain: "cluster.local"
infrastructureRef:
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Cluster
name: test1
Cluster, after reconciliation
apiVersion: cluster.x-k8s.io/v1beta1
kind: Cluster
metadata:
name: test1
namespace: metal3
spec:
clusterNetwork:
pods:
cidrBlocks:
- 192.168.0.0/18
serviceDomain: cluster.local
services:
cidrBlocks:
- 10.96.0.0/12
|----------------------------------------------------------------------------|
|# infrastructureRef comes from 'Metal3Cluster' and is added by 'CAPM3' |
| infrastructureRef: |
| apiVersion: infrastructure.cluster.x-k8s.io/v1beta1 |
| kind: Metal3Cluster |
| name: test1 |
| namespace: metal3 |
|-----------------------------------------------------------------------------
status:
apiEndpoints:
- host: 192.168.111.249
port: 6443
controlPlaneInitialized: true
infrastructureReady: true
phase: provisioned
Metal3Cluster, User provided Configuration for
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Cluster
metadata:
name: test1
spec:
apiEndpoint: https://192.168.111.249:6443
noCloudProvider: true
Metal3Cluster, after reconciliation
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Cluster
metadata:
name: test1
namespace: metal3
|----------------------------------------------------------------------------|
|# ownerReferences refers to the linked Cluster and is added by 'CAPM3' |
|ownerReferences: |
|- apiVersion: cluster.x-k8s.io/v1beta1 |
| kind: Cluster |
| name: test1 |
| uid: 193ec580-89db-46cd-b6f7-ddc0cd79636d |
|----------------------------------------------------------------------------|
spec:
apiEndpoint: https://192.168.111.249:6443
noCloudProvider: true
status:
apiEndpoints:
- host: 192.168.111.249
port: 6443
ready: true
Machine, User provided Configuration
apiVersion: cluster.x-k8s.io/v1beta1
kind: Machine
metadata:
name: test1-controlplane-0
labels:
cluster.x-k8s.io/control-plane: "true"
cluster.x-k8s.io/cluster-name: "test1"
spec:
version: v1.18.0
bootstrap:
configRef:
apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
kind: KubeadmConfig
name: test1-controlplane-0
infrastructureRef:
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Machine
name: test1-controlplane-0
Machine, after reconciliation
apiVersion: cluster.x-k8s.io/v1beta1
kind: Machine
metadata:
labels:
cluster.x-k8s.io/cluster-name: test1
cluster.x-k8s.io/control-plane: "true"
name: test1-controlplane-0
namespace: metal3
ownerReferences:
- apiVersion: cluster.x-k8s.io/v1beta1
kind: Cluster
name: test1
uid: 193ec580-89db-46cd-b6f7-ddc0cd79636d
spec:
bootstrap:
configRef:
apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
kind: KubeadmConfig
name: test1-controlplane-0
|----------------------------------------------------------------------------|
|# dataSecretName comes from Secret 'test1-controlplane-0' |
|# & is added by 'CAPI' |
| dataSecretName: test1-controlplane-0 |
|----------------------------------------------------------------------------|
infrastructureRef:
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Machine
name: test1-controlplane-0
providerID: metal3://8e16d3b6-d48c-41e0-af0f-e43dbf5ec0cd
version: v1.18.0
status:
addresses:
- address: 172.22.0.10
type: InternalIP
- address: 192.168.111.21
type: InternalIP
- address: node-1
type: Hostname
- address: node-1
type: InternalDNS
bootstrapReady: true
infrastructureReady: true
nodeRef:
name: node-1
phase: running
Metal3Machine, User provided Configuration
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Machine
metadata:
name: test1-controlplane-0
spec:
image:
url: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img
checksum: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img.sha256sum
Metal3Machine, after reconciliation
apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: Metal3Machine
metadata:
name: test1-controlplane-0
namespace: metal3
# ownerReferences refers to the linked Machine and is added by 'CAPM3'
ownerReferences:
- apiVersion: cluster.x-k8s.io/v1beta1
kind: Machine
name: test1-controlplane-0
spec:
hostSelector: {}
image:
checksum: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img.sha256sum
url: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img
providerID: metal3://8e16d3b6-d48c-41e0-af0f-e43dbf5ec0cd
|----------------------------------------------------------------------------|
|# dataTemplate comes from CR 'dataTemplate' and is added by 'CAPM3' |
|dataTemplate: |
| name: test1-workers-template |
| namespace: metal3 |
|----------------------------------------------------------------------------|
status:
addresses:
- address: 172.22.0.10
type: InternalIP
- address: 192.168.111.21
type: InternalIP
- address: node-1
type: Hostname
- address: node-1
type: InternalDNS
ready: true
BareMetalHost, User provided Configuration
apiVersion: metal3.io/v1alpha1
kind: BareMetalHost
metadata:
name: node-1
spec:
online: true
bootMACAddress: 00:b2:8c:ee:22:98
bmc:
address: ipmi://192.168.111.1:6231
credentialsName: node-1-bmc-secret
BareMetalHost, after reconciliation
apiVersion: metal3.io/v1alpha1
kind: BareMetalHost
metadata:
name: node-1
namespace: metal3
spec:
bmc:
address: ipmi://192.168.111.1:6231
credentialsName: node-1-bmc-secret
bootMACAddress: 00:b2:8c:ee:22:98
|-------------------------------------------------------------------------------|
|# consumerRef refers to the linked Metal3Machine is added by 'CAPM3' |
|consumerRef: |
| apiVersion: infrastructure.cluster.x-k8s.io/v1beta1 |
| kind: Metal3Machine |
| name: test1-controlplane-0 |
| namespace: metal3 |
|# Image comes from 'Metal3Machine' and is added by 'CAPM3' |
|image: |
| checksum: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img.sha256sum|
| url: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img |
|online: true |
|# UserData comes from 'Metal3Machine' and is added by 'CAPBK' |
|userData: |
| name: test1-controlplane-0-user-data |
| namespace: metal3 |
|-------------------------------------------------------------------------------|
status:
goodCredentials:
credentials:
name: node-1-bmc-secret
namespace: metal3
hardware: <Hardware related data>
hardwareProfile: unknown
operationHistory: ....
operationalStatus: OK
poweredOn: true
provisioning:
ID: 75450b93-0476-45b7-8629-d1eebbdc558b
image:
checksum: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img.sha256sum
url: http://172.22.0.1/images/bionic-server-cloudimg-amd64.img
state: provisioned
triedCredentials:
credentials:
name: node-1-bmc-secret
namespace: metal3
credentialsVersion: "1435"
KubeadmConfig, user provided Configuration
apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
kind: KubeadmConfig
metadata:
name: test1-controlplane-0
spec:
initConfiguration:
nodeRegistration:
name: "{{ ds.meta_data.name }}"
kubeletExtraArgs:
node-labels: "metal3.io/uuid={{ ds.meta_data.uuid }}"
preKubeadmCommands: <list of commands>
postKubeadmCommands: <list of commands>
files: <list of files>
KubeadmConfig, after reconciliation
apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
kind: KubeadmConfig
metadata:
name: test1-controlplane-0
namespace: metal3
ownerReferences:
- apiVersion: cluster.x-k8s.io/v1beta1
kind: Machine
name: test1-controlplane-0
spec:
clusterConfiguration:
apiServer: {}
apiVersion: kubeadm.k8s.io/v1beta1
certificatesDir: /etc/kubernetes/pki
clusterName: test1
controlPlaneEndpoint: 192.168.111.249:6443
controllerManager: {}
dns:
type: ""
etcd: {}
imageRepository: ""
kind: ClusterConfiguration
kubernetesVersion: v1.32.0
networking:
dnsDomain: cluster.local
podSubnet: 192.168.0.0/18
serviceSubnet: 10.96.0.0/12
scheduler: {}
files: <list of files>
initConfiguration:
nodeRegistration:
kubeletExtraArgs:
node-labels: metal3.io/uuid={{ ds.meta_data.uuid }}
name: '{{ ds.meta_data.name }}'
postKubeadmCommands: <list of commands>
status:
|----------------------------------------------------------------------------|
|bootstrapData: <user data in base64 format> |
|----------------------------------------------------------------------------|
ready: true
As was shown on the above CRs, some of the fields are introduced in one CR and they travel through multiple CRs to reach the BMH, which is the closest to the Bare Metal Server. There is also a movement of information from the virtual/physical machines towards the CRs, but this is beyond the scope of this document. Some of the fields are added by users, while the others are by the relevant controllers.
We have added the source of relevant fields as comments in the above yaml files. The following sequence diagram shows the flow of information (fields) across multiple CRs with the help of controllers.
apiEndpoint:
IP:Port of a load balancer (keepalived VIP)
image:
OS image for the Metal3 Machine
The following fields are used to make a relationship between CRs.
infrastructureRef
ownerReferences
consumerRef
The following fields refer to the cloud-init data, but the names change across different CRs.
bootstrap
bootstrap.data
userData
bootstrapData