Kubernetes-native declarative infrastructure for Metal3.
The Cluster API brings declarative, Kubernetes-style APIs to cluster creation, configuration and management. The API itself is shared across multiple cloud providers. Cluster API Provider Metal3 is one of the providers for Cluster API and enables users to deploy a Cluster API based cluster on top of bare metal infrastructure using Metal3.
CAPM3 version | Cluster API version | CAPM3 Release | CAPI Release |
---|---|---|---|
v1alpha5 | v1alpha4 | v0.5.X | v0.4.X |
v1beta1 | v1beta1 | v1.1.X | v1.1.X |
v1beta1 | v1beta1 | v1.2.X | v1.2.X |
v1beta1 | v1beta1 | v1.3.X | v1.3.X |
v1beta1 | v1beta1 | v1.4.X | v1.4.X |
v1beta1 | v1beta1 | v1.5.X | v1.5.X |
The recommended method is using Clusterctl.
Starting from v0.5.0
release of Cluster API Provider Metal3, Baremetal
Operator is decoupled from Cluster API Provider Metal3 deployments when deployed
via clusterctl
. For this reason, Baremetal Operator will not be installed when
initializing the Metal3 provider with clusterctl, and its CRDs and controller
need to be manually installed. Example flow of installing Metal3 provider:
-
Install Cluster API core, bootstrap and control-plane providers. This will also install cert-manager if it is not already installed. To have more verbose logs you can use the -v flag when running the clusterctl and set the level of the logging verbose with a positive integer number, ie. -v5.
clusterctl init --core cluster-api:v1.6.7 --bootstrap kubeadm:v1.6.7 \ --control-plane kubeadm:v1.6.7 -v5
-
Install Metal3 provider. This will install the latest version of Cluster API Provider Metal3 CRDs and controllers.
clusterctl init --infrastructure metal3
You can also specify the provider version by appending a version tag to the provider name as follows:
clusterctl init --infrastructure metal3:v1.6.1
-
Deploy Baremetal Operator manifests and CRDs. You need to install cert-manager for Baremetal Operator, but since step 1 already does it, we can skip it and only install the operator. Depending on whether you want TLS, or basic-auth enabled, kustomize paths may differ. Check operator dev-setup doc for more info.
git clone https://github.com/metal3-io/baremetal-operator.git kubectl create namespace baremetal-operator-system cd baremetal-operator kustomize build config/default | kubectl apply -f -
-
Install Ironic. There are a couple of ways to do it.
- Run within a Kubernetes cluster as a pod, refer to the deploy.sh script.
- Outside of a Kubernetes cluster as a container. Please refer to the run_local_ironic.sh script.
Please refer to the getting-started for more info.
Starting from v0.5.0
release of Cluster API Provider Metal3, Baremetal
Operator is decoupled from Cluster API Provider Metal3 deployments when deployed
via clusterctl
. For that reason, when performing clusterctl move
, custom
objects outside of the Cluster API chain or not part of CAPM3 will not be
pivoted to a target cluster. An example of those objects is BareMetalHost, or
user created ConfigMaps and Secrets which are reconciled by Baremetal Operator.
To ensure that those objects are also pivoted as part of clusterctl move
,
clusterctl.cluster.x-k8s.io
label needs to be set on the BareMetalHost CRD
before pivoting. If there are other CRDs also need to be pivoted to the target
cluster, the same label needs to be set on them.
All the other objects owned by BareMetalHost, such as Secret and ConfigMap don't require this label to be set, because they hold ownerReferences to BareMetalHost, and that is good enough for clusterctl to move all the hierarchy of BareMetalHost object.
There are multiple ways to setup a development environment:
- Using Tilt
- Other management cluster
- See metal3-dev-env for an
end-to-end development and test environment for
cluster-api-provider-metal3
and baremetal-operator.
See the API Documentation for details about the objects used with this Cluster API provider. You can also see the cluster deployment workflow for the outline of the deployment process.
The architecture with the components involved is documented here
To trigger e2e test on a PR, use the following phrases:
Release-1.6 branch:
- /test metal3-ubuntu-e2e-integration-test-release-1-6 runs integration e2e tests with CAPM3 API version v1beta1 and branch release-1.6 on Ubuntu
- /test metal3-centos-e2e-integration-test-release-1-6 runs integration e2e tests with CAPM3 API version v1beta1 and branch release-1.6 on CentOS
Unlike integration tests, basic tests focus on the target cluster creation without involving pivoting from the bootstrap cluster. To run basic tests use:
- /test metal3-centos-e2e-basic-test-release-1-6 runs basic e2e tests on release-1.6 branch with centos
Release-1.6 branch:
- /test metal3-ubuntu-e2e-feature-test-release-1-6 runs e2e feature tests with CAPM3 API version v1beta1 and branch release-1.6 on Ubuntu
- /test metal3-centos-e2e-feature-test-release-1-6 runs e2e feature tests with CAPM3 API version v1beta1 and branch release-1.6 on CentOS
CAPM3 tests upgrade from all supported release to the current one.
- /test metal3-e2e-clusterctl-upgrade-test-release-1-6 runs e2e clusterctl upgrade tests on release-1.6 with Ubuntu
CAPM3 tests upgrading kubernetes between last 3 releases.
The trigger takes the format:
/test metal3-e2e-<from-minor-k8s-v>-<to-minor-k8s-v>-upgrade-test-<branch>
- /test metal3-e2e-1-28-1-29-upgrade-test-release-1-6
- /test metal3-e2e-1-27-1-28-upgrade-test-release-1-6
- /test metal3-e2e-1-26-1-27-upgrade-test-release-1-6
Note:
- Triggers follow the pattern:
/test metal3-<image-os>-e2e-<test-type>-test-<branch>
More info about e2e test can be found here