Skip to content

LCM-only Installation Guide#

This section provides instructions on how to install Pelagia in lifecycle-management-only (lcm-only) mode. This mode allows you to automatically remove Rook Ceph OSD disks and nodes in your Kubernetes cluster.

Prerequisites#

  • A Kubernetes cluster, for example, deployed with k0s.
  • A deployed Ceph cluster managed by Rook.

Installation#

To install Pelagia in lcm-only mode, use the Helm chart provided in the repository:

export PELAGIA_VERSION="<version>"
helm upgrade --install pelagia-ceph oci://registry.mirantis.com/pelagia/pelagia-ceph --version ${PELAGIA_VERSION} --set cephDeployment.enabled=false,rookConfig.configureRook=false -n pelagia --create-namespace

Substitute <version> with the latest stable version of Pelagia Helm chart.

This command installs Pelagia LCM controllers in the pelagia namespace. If the namespace does not exist, Helm will create it. As a result, the following controllers appear in the pelagia namespace:

$ kubectl -n pelagia get pods

NAME                                     READY   STATUS    RESTARTS   AGE
pelagia-lcm-controller-6f858b6c5-4bqqr   3/3     Running   0          9m40s
pelagia-lcm-controller-6f858b6c5-6thhn   3/3     Running   0          9m40s
pelagia-lcm-controller-6f858b6c5-whq7x   3/3     Running   0          9m40s

The installation command above installs Pelagia in lcm-only mode for an existing Rook Ceph cluster placed in the rook-ceph namespace. If you want to manage a Rook Ceph cluster in a different namespace, set the rookConfig.rookNamespace chart value. For example: 

export PELAGIA_VERSION="<version>"
helm upgrade --install pelagia-ceph oci://registry.mirantis.com/pelagia/pelagia-ceph --version ${PELAGIA_VERSION} --set cephDeployment.enabled=false,rookConfig.configureRook=false,rookConfig.rookNamespace=new-rook-ceph -n pelagia --create-namespace

Substitute <version> with the latest stable version of Pelagia Helm chart.

Post-installation#

After the installation, create empty CephDeploymentHealth custom resource. It stores Ceph cluster state and Rook API status.

apiVersion: lcm.mirantis.com/v1alpha1
kind: CephDeploymentHealth
metadata:
  name: pelagia-ceph
  namespace: pelagia
status: {}

Now you can manage Rook Ceph OSD disks and nodes using the Pelagia CephOsdRemoveTask custom resource. For example:

apiVersion: lcm.mirantis.com/v1alpha1
kind: CephOsdRemoveTask
metadata:
    name: remove-osd-4
    namespace: pelagia
spec:
  nodes:
    storage-worker-1:
      cleanupByOsd:
      - id: 4

This example will dry-run the removal of the OSD with ID 4 on the storage-worker-1 node. After the dry-run, the status field of the CephOsdRemoveTask resource will be updated with the results of the dry-run.

To trigger the OSD removal, set the spec.approve field to true and apply the resource again. For example:

apiVersion: lcm.mirantis.com/v1alpha1
kind: CephOsdRemoveTask
metadata:
    name: remove-osd-4
    namespace: pelagia
spec:
  approve: true
  nodes:
    storage-worker-1:
      cleanupByOsd:
      - id: 4

This action triggers the removal of the OSD with ID 4 on the storage-worker-1 node.

The CephOsdRemoveTask workflow is as follows:

  1. The OSD daemon is stopped and data is rebalanced out from the OSD.
  2. The OSD disk is cleaned up on the node.
  3. The OSD daemon is removed from the Ceph cluster.

See also#

For the Pelagia architecture and overview, refer to the Architecture Guide.

For the detailed OSD automated lifecycle management, refer to Ops Guide: Automated Lifecycle Management.