Connecting Nova to External Ceph
When operating a cloud environment, it is often necessary to use a shared storage system rather than the local compute node for virtual machine disk storage. This can be useful for a number of reasons, such as:
- To provide a scalable storage solution for instances
- To provide a storage solution that is separate from the compute nodes
- To enable live migration capabilities
In this guide, we will show you how to connect Nova to an external Ceph storage system. The examples provided here assume Cinder is also configured for the same external Ceph backend.
Prerequisites
Before you begin, you will need the following:
- A running OpenStack environment
- A running Ceph environment
- A running Nova environment
- A running Cinder environment using External Ceph
- The
ceph.conffile from the Ceph deployment server - An OSD pool named
vmswith adequate PGs - A Ceph client named
cinderwith adequate permissions to thevmspool
Information Needed
The following information is needed to configure Nova to use Ceph as an external storage system.
| Property | Notes |
|---|---|
fsid |
The UUID of the respective Ceph deployment. Contained in ceph.conf. |
mon_host |
Specially-crafted list of Ceph monitor hosts. Contained in ceph.conf. |
Ceph cinder client key |
The Ceph key used to operate as the cinder user in Ceph |
Step 1: Configure Ceph for Nova
Prior to configuring Nova to support an external Ceph deployment, you must first configure users and pools within Ceph. The examples below are provided as reference only, and may not be applicable to your environment. An assumption is made that Ceph is operational and OSDs have been made available for use.
Step 1a: Create and Initialize a Pool
Step 1b: Disable Autoscaling (optional)
Step 1c: Set Placement Groups on the Pool
Please note the value provided for pgs may not be appropriate for your deployment. Please refer to the Ceph PG calculator for additional guidance.
Step 1d: Create a client user for Nova/Cinder
In this example, a common client name cinder will be used for both Nova and Cinder integrations.
ceph auth get-or-create client.cinder mon 'profile rbd' osd 'profile rbd pool=volumes, profile rbd pool=vms, profile rbd-read-only pool=images' mgr 'profile rbd pool=volumes, profile rbd pool=vms'
The following output may be provided:
Please note the key will be unique to your deployment. The key value will be needed for a Kubernetes secret later in this guide.
Step 2: Configure Kubernetes Resources
The OpenStack-Helm charts rely on Kubernetes resources such as ConfigMaps and Secrets to create and populate configuration files during the deployment.
On a Ceph node, locate the ceph.conf file at /etc/ceph/ceph.conf and copy its output.
Step 2a: Create the Kubernetes ConfigMap
On your OpenStack deploy node, create a file at /etc/genestack/manifests/ceph/ceph-etc.yaml containing the following ConfigMap with your respective ceph.conf output:
apiVersion: v1
kind: ConfigMap
metadata:
name: ceph-etc
namespace: openstack
data:
ceph.conf: |
[global]
fsid = 6be518ac-f7b4-11f0-a76d-4bc5024d47a0
mon_host = [v2:172.16.1.70:3300/0,v1:172.16.1.70:6789/0] [v2:172.16.1.71:3300/0,v1:172.16.1.71:6789/0] [v2:172.16.1.72:3300/0,v1:172.16.1.72:6789/0] [v2:172.16.1.73:3300/0,v1:172.16.1.73:6789/0] [v2:172.16.1.74:3300/0,v1:172.16.1.74:6789/0]
Apply the ConfigMap using the following command:
Please note the same ceph-etc ConfigMap may be used for other Ceph integrations such as Glance and Cinder. The name ceph-etc is built in to the template and should not be overridden unless you know what you're doing.
Step 2b: Create the Kubernetes Secret
On your OpenStack deploy node, create a file at /etc/genestack/manifests/ceph/pvc-ceph-keyring.yaml containing the following Secret using the key provided earlier:
apiVersion: v1
kind: Secret
metadata:
name: pvc-ceph-client-key
namespace: openstack
labels:
application: ceph
component: rbd
stringData:
key: AQCQqXJpHtP3AhBBU6rf/yvgq92fuJqBgy3Nxg==
type: Opaque
Apply the Secret using the following command:
Please note the name pvc-ceph-client-key is built in to the template and should not be overridden unless you know what you're doing.
Step 3: Create a Libvirt Secret UUID
Using the uuidgen utility, create a unique UUID that can be used to create a Libvirt secret:
Step 4: Configure Nova to use External Ceph
Update the Nova Helm overrides at /etc/genestack/helm-configs/nova/nova-helm-overrides.yaml with the following configuration to connect Nova to External Ceph.
Note that the values for both secret_uuid and rbd_secret_uuid should be the same UUID generated in the previous step. The keyring value should match the key generated in Step 1d.
---
images:
tags:
nova_compute: "quay.io/airshipit/nova:2024.1-ubuntu_jammy"
conf:
ceph:
enabled: true
cinder:
secret_uuid: eea41bd9-c85e-4c99-879b-65e38ffb3213
keyring: AQCQqXJpHtP3AhBBU6rf/yvgq92fuJqBgy3Nxg==
nova:
libvirt:
images_type: rbd
images_rbd_pool: vms
images_rbd_ceph_conf: /etc/ceph/ceph.conf
rbd_secret_uuid: eea41bd9-c85e-4c99-879b-65e38ffb3213
force_raw_images: true
volume_use_multipath: false
Please note the upstream container images are required for Ceph support at this time.
Step 5: Configure Libvirt to use External Ceph
Update the Libvirt Helm overrides at /etc/genestack/helm-configs/libvirt/libvirt-helm-overrides.yaml with the following configuration to connect Libvirt to External Ceph.
Note that the value for both secret_uuid should be the same UUID generated in Step 3. The keyring value should match the key generated in Step 1d.
---
images:
tags:
libvirt: "docker.io/openstackhelm/libvirt:2024.1-ubuntu_jammy"
conf:
ceph:
enabled: true
cinder:
keyring: AQCQqXJpHtP3AhBBU6rf/yvgq92fuJqBgy3Nxg==
secret_uuid: eea41bd9-c85e-4c99-879b-65e38ffb3213
external_ceph:
enabled: true
user: cinder
secret_uuid: eea41bd9-c85e-4c99-879b-65e38ffb3213
user_secret_name: pvc-ceph-client-key
Please note the upstream container images are required for Ceph support at this time.
Step 6: Apply the Configuration
Apply the configuration to the Nova Helm chart.
Once the configuration has been applied, Nova will be configured to use an external Ceph deployment for image storage. Instances deployed from image (not boot-from-volume) should automatically use the external Ceph RBD backend.