docs

Gitops + ClusterAPI = Kubernetes

Repo structure

Directories are listed roughly in the order in which they are applied.

/clusters/

Each subdir represents everything required to create and run a Kubernetes cluster - these are the base dirs which are reconciled by Flux running in each cluster.

/clusters/bootstrap/

Ephemeral cluster created in order to bootstrap the management cluster. Usually a single-node Talos cluster running in Docker which is removed once the MC becomes self-managing. See bootstrap-mc.md.

/workloads/cni/

All resources required to install and manage a CNI in a cluster (currently Cilium running in strict mode)

/workloads/controllers/

Controllers defined here are required for each cluster, depending on whether the cluster is a management or workload cluster.

/workloads/flux/

Configs to install and manage Flux and all core components in a cluster; all clusters manage themselves once they have been bootstrapped.

/workloads/addons/

Additional applications which are required to support a cluster.

/kubernetes/resources/

Template manifests for CAPI components used by Sidero to create a cluster.

/kubernetes/clusters/

Subdirs contain manifests for various Sidero-specific components in order to customise the cluster deployment. Kubernetes nodes are configured at this level (networking, low-level Kubernetes settings such as kubelet args etc).

/apps/

Subdirs contain config for optional applications running in clusters.

/apps/base/

Shared app config between all clusters which is inherited and extended by clusters running each app.

/apps/$clustername/

Cluster-specific config for apps in each cluster.

/hack/

Contains various supporting scripts and information used to create clusters (not reconciled by Flux).

Notes

• Certificate authorities for Kubernetes and Talos are generated ahead of cluster creation (see /hack/pki/ for more info).
• Networking has some specific configuration for each cluster (see networking.md).
• Clusters can create Loadbalancer services. This is handled by Cilium which advertises the IP via BGP to Opnsense (see this configmap as an example)
• Node creation is handled externally by Terraform and this repo. Therefore, in order to scale a cluster the VMs must first be created with Terraform.