Deploy Ceph Storage Cluster in Kubernetes using Rook

Last Updated:
Deploy Ceph Storage Cluster in Kubernetes using Rook

In this tutorial, you will learn how to deploy Ceph storage cluster in Kubernetes using Rook. Rook is an open-source cloud-native storage orchestrator that provides a platform, framework and support for deploying various distributed storage solutions such as Ceph, NFS, Cassandra, Minio, CockroachDB etc on Kubernetes. This guide will focus on deploying Ceph, a distributed storage system that provides file, block and object storage services, using Rook.

We are using Ubuntu 22.04 LTS server in our cluster.

Deploying Ceph Storage Cluster in Kubernetes using Rook

Why Rook for Ceph Deplyment on Kubernetes Cluster?

Some of the benefits of using Rook to deploy Ceph on Kubernetes cluster:

  • Task automation: Rook automates deployment, bootstrapping, configuration, provisioning, scaling, upgrading, migration, disaster recovery, monitoring, and resource management for your Ceph storage cluster.
  • Self-managing, self-scaling, self-healing: Rook ensures that the storage services are always available and performant. It automatically scales the storage up or down as needed and heals any issues that may arise.
  • Seamless integration with Kubernetes: Rook uses the Kubernetes API and scheduling features to provide a seamless experience for managing your storage. You can use familiar Kubernetes commands and tools to interact with your Rook deployed storage cluster.

Rook Components

Rook is made up of several components that work together to manage a storage cluster:

  • Rook Operator: This is the core component of Rook. It is a Kubernetes operator that is responsible for deploying, configuring, and managing your storage clusters.
  • Rook agents: These are the daemons that run on each node in the Kubernetes cluster. They are responsible for mounting and unmounting storage devices and for managing the lifecycle of storage pods.
  • Rook discover: rook-discover is a dedicated containerized component that runs as a pod within your Kubernetes cluster. It is responsible for actively scanning the cluster for existing Ceph daemons such as MONs, OSDs, and MGRs. Rook Discover informs the Rook Agent about the discovered OSDs.

Other components include but not limited to;

  • Ceph daemons:
    • MONs: Manage the Ceph cluster and store its configuration
    • OSDs: Store data in the Ceph cluster
    • MGRs: Provide management services for the Ceph cluster
  • Custom Resource Definitions (CRDs):
    • Define the desired state of the storage cluster
    • Allow users to configure the storage cluster
  • CSI drivers:
    • Allow Rook to integrate with the Container Storage Interface (CSI).
    • There are three CSI drivers integrated with Rook; Ceph RBD, CephFS and NFS (disabled by default). CephFS and RBD drivers are enabled automatically by the Rook operator.

The Architecture

The Rook architecture is depicted by the screenshot below from the Rooks documentation page.

Rook High Level Architecture

Prerequisites for Deploying Ceph on K8S using Rook

A running Kubernetes Cluster

You need to have a Kubernetes cluster up and running. In this guide, we are using a three worker-node Kubernetes cluster. See our guide below on how to deploy a three node Kubernetes cluster on Ubuntu;

Setup Kubernetes Cluster on Ubuntu 22.04/20.04

These are the details of my K8s cluster;

NodeHostnameIP AddressvCPUsRAM (GB)Storage ResourceOS
Mastermaster192.168.122.1028OS: /dev/vdaUbuntu 22.04
Worker 1worker01192.168.122.1128OS: /dev/vda, OSD: raw /dev/vdb, 100GUbuntu 22.04
Worker 2worker02192.168.122.1228OS: /dev/vda, OSD: raw /dev/vdb, 100GUbuntu 22.04
Worker 3worker03192.168.122.1328OS: /dev/vda, OSD: raw /dev/vdb, 100GUbuntu 22.04

Cluster nodes;

kubectl get nodes
NAME       STATUS   ROLES           AGE     VERSION
master     Ready    control-plane   13m     v1.28.2
worker01   Ready    <none>          5m48s   v1.28.2
worker02   Ready    <none>          5m45s   v1.28.2
worker03   Ready    <none>          5m42s   v1.28.2

Note that Kubernetes v1.22 or higher is supported. You can get the version using the kubectl version command and check the Server version.

Storage Resources:

At least one of these local storage types is required:

  • Raw devices (no partitions or formatted filesystems)
  • Raw partitions (no formatted filesystem)
  • LVM Logical Volumes (no formatted filesystem)
  • Persistent Volumes available from a storage class in block mode

We will be using raw devices with no partitions/filesystem in this guide.

We have attached raw block devices each of 100G to each of the worker nodes in the cluster;

lsblk | grep -v '^loop'
vda                       252:0    0   100G  0 disk 
├─vda1                    252:1    0     1M  0 part 
├─vda2                    252:2    0     2G  0 part /boot
└─vda3                    252:3    0    98G  0 part 
  └─ubuntu--vg-ubuntu--lv 253:0    0    98G  0 lvm  /var/lib/kubelet/pods/d447336a-f345-4629-877f-003053e48c1c/volume-subpaths/tigera-ca-bundle/calico-node/1
vdb                       252:16   0   100G  0 disk

LVM Package

Ceph OSDs have a dependency on LVM in the following scenarios:

  • OSDs are created on raw devices or partitions
  • If encryption is enabled (encryptedDevice: true in the cluster CR)
  • metadata device is specified

LVM is not required for OSDs in these scenarios:

  • Creating OSDs on PVCs using the storageClassDeviceSets

In this guide, since we are using raw block devices for OSDs, then we need to install the LVM package.

Since we are using Ubuntu in our environment, then you can install LVM package as follows;

sudo apt install lvm2

For any other Linux distro, consult their documentation on how to install the LVM package.

Deploy the Rook Operator

Clone Current Rook Release Github Repository

Once you have all the prereqs in place, proceed to deploy Rook operator on the cluster node with full access to the kubeconfig.

So what is kubeconfig? kubeconfig is a an abbreviation for Kubernetes configuration file and is:

  • A YAML file containing the configuration details required to connect to a Kubernetes cluster.
  • Stores information like:
    • Cluster server addresses.
    • User credentials (certificate or token).
    • Current context (the specific cluster and namespace you want to interact with).
  • Used by various tools like kubectl, the Kubernetes command-line tool, to interact with the cluster.
  • By default, kubeconfig is stored in the .kube/config file within the home directory of your user account.
  • You can also specify a custom location for the file using the --kubeconfig flag in kubectl commands.

In our cluster, we have our kubeconfig on the control-plane/worker node.

To deploy Rook operator, you need to clone their Github repository. It is recommended that you clone the current Rook release Github repository. v1.13.0 is the current release as of this writing.

Install git package on your system. On Ubuntu, you can install git by executing the command below;

sudo apt install git

You can then clone the current release branch of Rook as follows;

git clone --single-branch --branch release-1.13

This clones Rook to rook directory in the current working directory.

ls -1 rook/

Deploy the Rook Operator

Next, navigate to the example manifests directory, rook/deploy/examples, and deploy the Rook operator (operator.yaml), the CRDs (Custom Resource Definitions, crds.yaml) and the common resources (common.yaml). In Kubernetes, a manifest is a YAML or JSON file that describes the desired state of a Kubernetes object within the cluster. These objects can include deployment, replica set, service… It includes information like:

  • Kinds: The type of resource being created (e.g., Pod, Deployment, Service).
  • Metadata: Names, labels, annotations, etc., for identification and configuration.
  • Specifications: Detailed configuration of the resource, including containers, volumes, network settings, etc.
cd ~/rook/deploy/examples
kubectl create -f crds.yaml -f common.yaml -f operator.yaml

Sample command output; created created created created created created created created created created created created created created created created created created created
namespace/rook-ceph created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created created
serviceaccount/objectstorage-provisioner created
serviceaccount/rook-ceph-cmd-reporter created
serviceaccount/rook-ceph-mgr created
serviceaccount/rook-ceph-osd created
serviceaccount/rook-ceph-purge-osd created
serviceaccount/rook-ceph-rgw created
serviceaccount/rook-ceph-system created
serviceaccount/rook-csi-cephfs-plugin-sa created
serviceaccount/rook-csi-cephfs-provisioner-sa created
serviceaccount/rook-csi-rbd-plugin-sa created
serviceaccount/rook-csi-rbd-provisioner-sa created
configmap/rook-ceph-operator-config created
deployment.apps/rook-ceph-operator created

Rook uses a default namespace called rook-ceph;

kubectl get ns
NAME               STATUS   AGE
calico-apiserver   Active   3h9m
calico-system      Active   3h9m
default            Active   3h14m
kube-node-lease    Active   3h14m
kube-public        Active   3h14m
kube-system        Active   3h14m
rook-ceph          Active   7s
tigera-operator    Active   3h10m

This therefore means that you have to specify the namespace for all subsequent kubectl commands related to Rook. If you want you can set the rook-ceph namespace as your preferred/default namespace so that you don’t have to specify the namespace when executing kubectl commands related to Rook by using the command, kubectl config set-context --current --namespace=<insert-rook-default-namespace-name-here>.

Deploying Ceph Storage Cluster in Kubernetes using Rook

Once the Rook Operator is deployed, you can now create the Ceph cluster.

Before you can proceed to deploy Ceph storage cluster, ensure that the rook-ceph-operator is in the Running state.

Note that Rook operator is deployed into rook-ceph namespace by default.

kubectl get pod -n rook-ceph

Sample output;

NAME                                  READY   STATUS    RESTARTS   AGE
rook-ceph-operator-598b5dd74c-42gh5   1/1     Running   0          90s

Rook supports different Ceph cluster settings for various environments. There are different manifests for various cluster environments that Rook ships with. These include;

  • cluster.yaml: Cluster settings for a production cluster running on bare metal. Requires at least three worker nodes. Like in our setup.
  • cluster-on-pvc.yaml: Cluster settings for a production cluster running in a dynamic cloud environment.
  • cluster-test.yaml: Cluster settings for a test environment such as minikube.

Now that the Rook operator is running, run the command below to create the cluster.

kubectl create -f cluster.yaml

Output; created.

The cluster will now take a few to initialize.

You can execute the command below to check the status of the cluster pods;

kubectl get pod -n rook-ceph

Sample output;

NAME                                           READY   STATUS    RESTARTS   AGE
csi-cephfsplugin-6cdnk                         2/2     Running   0          11s
csi-cephfsplugin-8klf8                         2/2     Running   0          11s
csi-cephfsplugin-dkkdf                         2/2     Running   0          11s
csi-cephfsplugin-provisioner-fd76b9895-888tm   5/5     Running   0          11s
csi-cephfsplugin-provisioner-fd76b9895-gscgb   5/5     Running   0          11s
csi-rbdplugin-gqfcl                            2/2     Running   0          11s
csi-rbdplugin-lkqpm                            2/2     Running   0          11s
csi-rbdplugin-provisioner-75f66b455d-6xlgh     5/5     Running   0          11s
csi-rbdplugin-provisioner-75f66b455d-7w44t     5/5     Running   0          11s
csi-rbdplugin-tjk2r                            2/2     Running   0          11s
rook-ceph-mon-a-6b97dfd866-44r2k               1/2     Running   0          4s
rook-ceph-operator-598b5dd74c-42gh5            1/1     Running   0          2m6s

Note that the number of osd pods that will be created will depend on the number of nodes in the cluster and the number of devices configured. For the default cluster.yaml above, one OSD will be created for each available device found on each node.

Once everything is initialized, the pod status may look like as shown in the output below;

NAME                                                 READY   STATUS      RESTARTS   AGE
csi-cephfsplugin-6cdnk                               2/2     Running     0          2m10s
csi-cephfsplugin-8klf8                               2/2     Running     0          2m10s
csi-cephfsplugin-dkkdf                               2/2     Running     0          2m10s
csi-cephfsplugin-provisioner-fd76b9895-888tm         5/5     Running     0          2m10s
csi-cephfsplugin-provisioner-fd76b9895-gscgb         5/5     Running     0          2m10s
csi-rbdplugin-gqfcl                                  2/2     Running     0          2m10s
csi-rbdplugin-lkqpm                                  2/2     Running     0          2m10s
csi-rbdplugin-provisioner-75f66b455d-6xlgh           5/5     Running     0          2m10s
csi-rbdplugin-provisioner-75f66b455d-7w44t           5/5     Running     0          2m10s
csi-rbdplugin-tjk2r                                  2/2     Running     0          2m10s
rook-ceph-crashcollector-worker01-9585d87f9-96phf    1/1     Running     0          60s
rook-ceph-crashcollector-worker02-7549954c4b-d5jch   1/1     Running     0          49s
rook-ceph-crashcollector-worker03-675cdbd7f7-pjv7n   1/1     Running     0          47s
rook-ceph-exporter-worker01-6c9cf475fc-jxddz         1/1     Running     0          60s
rook-ceph-exporter-worker02-6b57f48d4-qsvsr          1/1     Running     0          46s
rook-ceph-exporter-worker03-75fb9cc47f-sht56         1/1     Running     0          45s
rook-ceph-mgr-a-65f4bb6685-mh4xr                     3/3     Running     0          80s
rook-ceph-mgr-b-6648b7fb6-dqtdv                      3/3     Running     0          79s
rook-ceph-mon-a-6b97dfd866-44r2k                     2/2     Running     0          2m3s
rook-ceph-mon-b-785dcc4874-wkzm4                     2/2     Running     0          99s
rook-ceph-mon-c-586576df47-9ctwf                     2/2     Running     0          90s
rook-ceph-operator-598b5dd74c-42gh5                  1/1     Running     0          4m5s
rook-ceph-osd-0-55f6c88b9c-l8f2w                     2/2     Running     0          50s
rook-ceph-osd-1-64c6c74db4-zjfn9                     2/2     Running     0          49s
rook-ceph-osd-2-8455dcfb9f-ghrbc                     2/2     Running     0          47s
rook-ceph-osd-prepare-worker01-426s7                 0/1     Completed   0          21s
rook-ceph-osd-prepare-worker02-trvpb                 0/1     Completed   0          18s
rook-ceph-osd-prepare-worker03-b7n6b                 0/1     Completed   0          15s

Check Ceph Cluster Status

To check the Ceph cluster status, you can use Rook toolbox. The Rook toolbox is a container with common tools used for rook debugging and testing and can be run in two modes;

  1. Interactive mode: Start a toolbox pod where you can connect and execute Ceph commands from a shell
  2. One-time job mode: Run a script with Ceph commands and collect the results from the job log

We will use the interactive mode in this example guide. Thus, create a toolbox pod;

kubectl create -f toolbox.yaml

You can check the current status of the toolbox deployment;

kubectl -n rook-ceph rollout status deploy/rook-ceph-tools

Sample output;

deployment "rook-ceph-tools" successfully rolled out


kubectl get pod -n rook-ceph | grep tool

Sample output;

rook-ceph-tools-564c8446db-xh6qp                     1/1     Running     0          71s

After that, connect to the rook-ceph-tools pod and check Ceph cluster status;

kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph -s

Or simply just use the name of the tools pod;

kubectl -n rook-ceph exec -it rook-ceph-tools-564c8446db-xh6qp -- ceph -s

Sample output;

    id:     e1466372-9f01-42af-8ad7-0bfcfa71ef78
    health: HEALTH_OK
    mon: 3 daemons, quorum a,b,c (age 21m)
    mgr: a(active, since 20m), standbys: b
    osd: 3 osds: 3 up (since 20m), 3 in (since 21m)
    pools:   1 pools, 1 pgs
    objects: 2 objects, 449 KiB
    usage:   80 MiB used, 300 GiB / 300 GiB avail
    pgs:     1 active+clean

Or you can login to the pod and execute the commands if you want;

kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash

You can execute any Ceph commands in there.

If you want to remove the toolbox deployment from the rook-ceph namespace;

kubectl -n rook-ceph delete deploy/rook-ceph-tools

Enable Rook Ceph Orchestrator Module

The Rook Orchestrator module provides an integration between Ceph’s Orchestrator framework and Rook. It runs in the ceph-mgr daemon and implements the Ceph orchestration API by making changes to the Ceph storage cluster in Kubernetes that describe desired cluster state. A Rook cluster’s ceph-mgr daemon is running as a Kubernetes pod, and hence, the rook module can connect to the Kubernetes API without any explicit configuration.

While the orchestrator Ceph module is enabled by default, the rook Ceph module is disabled by default.

You need the Rook toolbox to enable Rook ceph orchestrator module. We have already enabled the Rook toolbox above.

Thus, either login directly to Rook toolbox tool (kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash) and execute the Ceph commands to enable the rook orchestrator module or simply execute the ceph commands without logging into the toolbox;

kubectl -n rook-ceph exec deploy/rook-ceph-tools -- ceph mgr module enable rook
kubectl -n rook-ceph exec deploy/rook-ceph-tools -- ceph orch set backend rook

Check the Ceph Orchestrator status;

kubectl -n rook-ceph exec deploy/rook-ceph-tools -- ceph orch status

Sample output;

Backend: rook
Available: Yes

Rook Ceph Storage Cluster Kubernetes Services

Ceph storage cluster Kubernetes services are also created;

kubectl get svc -n rook-ceph
NAME                      TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)             AGE
rook-ceph-exporter        ClusterIP            9926/TCP            34h
rook-ceph-mgr             ClusterIP             9283/TCP            34h
rook-ceph-mgr-dashboard   ClusterIP           8443/TCP            34h
rook-ceph-mon-a           ClusterIP           6789/TCP,3300/TCP   34h
rook-ceph-mon-b           ClusterIP             6789/TCP,3300/TCP   34h
rook-ceph-mon-c           ClusterIP            6789/TCP,3300/TCP   34h

All these services are exposed within the Kubernetes cluster and are only accessible within the cluster via their Cluster IPs and ports.

To see more details about each service, you can describe it. For example;

kubectl describe svc <service-name> -n <namespace>

Accessing Rook Kubernetes Ceph Storage Cluster Dashboard

As you can see above, we have a Ceph manager dashboard service, but is only accessible internally within the K8s cluster via the IP,, and port 8443 (https). To confirm this, get the description of the service.

kubectl describe svc rook-ceph-mgr-dashboard -n rook-ceph
Name:              rook-ceph-mgr-dashboard
Namespace:         rook-ceph
Labels:            app=rook-ceph-mgr
Annotations:       <none>
Selector:          app=rook-ceph-mgr,mgr_role=active,rook_cluster=rook-ceph
Type:              ClusterIP
IP Family Policy:  SingleStack
IP Families:       IPv4
Port:              https-dashboard  8443/TCP
TargetPort:        8443/TCP
Session Affinity:  None
Events:            <none>

As you can see, this service is of type ClusterIP. This, as already mentioned, this exposes the service on a cluster-internal IP. Choosing this value makes the Service only reachable from within the cluster.

How can we externally access Rook Kubernetes Ceph Cluster dashboard?

There are several ways you can use to expose a service externally. One of them is to set the service type to NodePort. This exposes the service on each Node’s IP at a static port. You can let the port be defined automatically or manually set it yourself, in the service manifest file.

By default, Rook ships with a service manifest file that you can use to expose Ceph cluster externally via the cluster Node’s IP. The file is named as dashboard-external-https.yaml under ~/rook/deploy/examples.

cat ~/rook/deploy/examples/dashboard-external-https.yaml
apiVersion: v1
kind: Service
  name: rook-ceph-mgr-dashboard-external-https
  namespace: rook-ceph # namespace:cluster
    app: rook-ceph-mgr
    rook_cluster: rook-ceph # namespace:cluster
    - name: dashboard
      port: 8443
      protocol: TCP
      targetPort: 8443
    app: rook-ceph-mgr
    mgr_role: active
    rook_cluster: rook-ceph # namespace:cluster
  sessionAffinity: None
  type: NodePort

So, to access the Ceph dashboard externally, create the external dashboard service.

cd ~/rook/deploy/examples
kubectl create -f dashboard-external-https.yaml

Confirm the services;

kubectl get svc -n rook-ceph
NAME                                     TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)             AGE
rook-ceph-exporter                       ClusterIP    <none>        9926/TCP            35h
rook-ceph-mgr                            ClusterIP     <none>        9283/TCP            35h
rook-ceph-mgr-dashboard                  ClusterIP   <none>        8443/TCP            35h
rook-ceph-mgr-dashboard-external-https   NodePort   <none>        8443:31617/TCP      25s
rook-ceph-mon-a                          ClusterIP   <none>        6789/TCP,3300/TCP   35h
rook-ceph-mon-b                          ClusterIP     <none>        6789/TCP,3300/TCP   35h
rook-ceph-mon-c                          ClusterIP    <none>        6789/TCP,3300/TCP   35h

As you can see, the service, rook-ceph-mgr-dashboard-external-https, internal port, 8443/TCP is mapped to port 31617/TCP on the each of the cluster host.

Get Kubernetes cluster Nodes IPs using kubectl command;

kubectl get nodes -o wide
master     Ready    control-plane   38h   v1.28.2   <none>        Ubuntu 22.04.2 LTS   5.15.0-76-generic   containerd://1.6.26
worker01   Ready    <none>          38h   v1.28.2   <none>        Ubuntu 22.04.2 LTS   5.15.0-91-generic   containerd://1.6.26
worker02   Ready    <none>          38h   v1.28.2   <none>        Ubuntu 22.04.2 LTS   5.15.0-76-generic   containerd://1.6.26
worker03   Ready    <none>          38h   v1.28.2   <none>        Ubuntu 22.04.2 LTS   5.15.0-76-generic   containerd://1.6.26

So, you can access dashboard on any node IP and port, 31617/TCP, https (Accept the ssl warning and proceed to dashboard).
rook kubernetes ceph dashboard

You can get the login credential for the admin user from the secrets called rook-ceph-dashboard-password;

kubectl -n rook-ceph get secret rook-ceph-dashboard-password -o jsonpath="{['data']['password']}" | base64 --decode && echo

Use the resulting password to login to ceph as user admin.

rook ceph kubernetes dashboard

You can enable Telemetry from the toolbox pod using the command below;

kubectl -n rook-ceph exec -it rook-ceph-tools-564c8446db-xh6qp -- ceph telemetry on --license sharing-1-0

Create Ceph Storage Pools

The ceph storage cluster is now running, however, without any usable storage pools. Ceph supports three types of storage that can be exposed by Rook:

  • Block Storage
    • Represented by Ceph Block Device (RBD)
    • Provides raw block devices for persistent storage in Kubernetes pods.
    • Similar to traditional block storage like hard drives or SSDs.
    • Used for applications requiring low-latency access to raw data, such as databases and file systems.
    • Check the guide on how to provision Block storage, Provision Block Storage for Kubernetes on Rook Ceph Cluster
  • Object Storage
    • Represented by Ceph Object Storage Gateway (RADOS Gateway, RGW)
    • Offers S3-compatible object storage for unstructured data like images, videos, and backups.
    • Applications can access objects directly using the S3 API.
    • Ideal for cloud-native applications and large-scale data management.

Rook ships with manifests for all these types of storage just in case you want to configure and use them.

The manifests of these storage types are provided under the directory, ~/rook/deploy/examples/csi/ for both RBD and CephFS, and under ~/rook/deploy/examples/object.yaml for Object Storage.

You can proceed to provision the storage via Ceph.

Check our guide below on how to provision block storage for Kubernetes on Rook ceph cluster.


Rook Ceph Documentation


We're passionate about sharing our knowledge and experiences with you through our blog. If you appreciate our efforts, consider buying us a virtual coffee. Your support keeps us motivated and enables us to continually improve, ensuring that we can provide you with the best content possible. Thank you for being a coffee-fueled champion of our work!

Photo of author
Linux Certified Engineer, with a passion for open-source technology and a strong understanding of Linux systems. With experience in system administration, troubleshooting, and automation, I am skilled in maintaining and optimizing Linux infrastructure.

Leave a Comment