Helm chart and OpenNebula images ready to deploy on Kubernetes

• Create namespace:

  kubectl create namespace opennebula

• Install Helm repo:

  helm repo add kvaps https://kvaps.github.io/charts

• Deploy OpenNebula:

  # download example values
  helm show values kvaps/opennebula --version 1.2.0 > values.yaml
  
  # install release
  helm install opennebula kvaps/opennebula --version 1.2.0 \
    --namespace opennebula \
    --set oned.createCluster=1 \
    -f values.yaml \
    --wait

• To deploy external compute node your hosts should have libvirtd and qemu-kvm installed and configured sudoers.
  However you can just install opennebula-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>

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 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

Find current leader:

kubectl get pod -n opennebula -l role=leader

Perform backup:

kubectl exec -n opennebula -c oned <leader_pod> -- sh -c 'mysqldump -h$DB_SERVER -u$DB_USER -p$DB_PASSWD $DB_NAME | gzip -9' > opennebula-db.sql.gz

To restore, redeploy release with --set oned.debug=true and:

kubectl exec -n opennebula -i -c oned <each_oned_pod> -- sh -c 'zcat | mysql -h$DB_SERVER -u$DB_USER -p$DB_PASSWD -D$DB_NAME' < opennebula-db.sql.gz

then disable debug

kubectl get pod -n opennebula -l role=leader

Perform backup:

kubectl exec -n opennebula -c oned <leader_pod> -- sh -c 'mysqldump -h$DB_SERVER -u$DB_USER -p$DB_PASSWD $DB_NAME | gzip -9' > opennebula-db.sql.gz

To restore, redeploy release with --set oned.debug=true and:

kubectl exec -n opennebula -i -c oned <each_oned_pod> -- sh -c 'zcat | mysql -h$DB_SERVER -u$DB_USER -p$DB_PASSWD -D$DB_NAME' < opennebula-db.sql.gz

then disable debug

The minor upgrades can be performed by standard way using rolling update, however major updates must be performed by fully chart reinstallation.
You have to remove the old chart, and install new one. No worry as your data should be saved on persistent volumes, thus new images will perform database migration on their first start.

Warning: Don't forget to make backup before the upgrade!

Minor upgrade:

helm upgrade opennebula kvaps/opennebula --version 1.2.0 \
  --namespace opennebula \
  -f values.yaml \
  --wait

Major upgrade:

# Remove the chart
helm remove opennebula \
  --namespace opennebula

# Deploy the new chart
helm upgrade opennebula kvaps/opennebula --version 1.2.0 \
  --namespace opennebula \
  -f values.yaml \
  --wait