Kube-OpenNebula
Helm chart and OpenNebula images ready to deploy on Kubernetes
Quick start
Control plane
-
Create namespace:
kubectl create namespace opennebula
-
Deploy OpenNebula:
cp helm/opennebula/values.yaml values.yaml vim values.yaml helm install opennebula helm/opennebula -f values.yaml --set oned.createCluster=1 --wait
Compute nodes
-
To deploy external compute node your hosts should have
libvirtd
andqemu-kvm
installed and configured sudoers.
However you can just installopennebula-node
meta-package. -
Get OpenNebula's ssh-key, and place it to
/var/lib/one/.ssh/authorized_keys
on every node to allow OpenNebula login via ssh.kubectl exec opennebula-opennebula-oned-0 -c oned -- ssh-keygen -y -f /var/lib/one/.ssh/id_rsa
-
Create new host via OpenNebula Interface.
-
Check is everything is fine
You should be able login via ssh from oned pod to every node. You can check that by executing the following command:kubectl exec -ti opennebula-opennebula-oned-0 ssh <node>
Customization
Sometimes you need to perform some customization, eg. update sunstone views and addtitional drivers, etc.
All these customizations could be done by updating dockerimages, you can find some examples here, or by simple using extraVolumes
and extraVolumeMounts
in chart values.
Production setup
Production install assumes having persistent storage.
OpenNebula requires one ReadWriteOnce persistent volume per each oned-instance where database files will be stored, even local volumes enough for that, however it also requires one shared (ReadWriteMany) persistent volume for virtual machine logs and vnc tokens, take a look at nfs-server-provisioner if you'r storage does not support ReadWriteMany.
Example production configuration can be found here
Upgrade notes
The minor upgrades can be performed by standard way using rolling update, however major updates should be performed by fully chart reinstallation. You have to remove the old chart, and install new one, however your data should be saved on persistent volumes, thus new images will perform database migration on their first start.
Perform backup:
# Find the leader pod
kubectl get pod -l role=leader
# Perform the backup
kubectl exec <leader_pod> -c oned -- bash -c 'mysqldump -h$DB_SERVER -u$DB_USER -p$DB_PASSWORD $DB_NAME | gzip -9' > backup.sql.gz
Minor upgrade:
helm upgrade opennebula helm/opennebula -f values.yaml -f secrets.yaml --wait
Major upgrade:
# Remove the chart
helm remove opennebula opennebula
# Deploy the new chart
helm install opennebula opennebula -f values.yaml -f secrets.yaml --wait