libvirt-actuator

README

Libvirt actuator

The command allows to directly interact with the libvirt actuator.

How to prepare environment

For running the libvirt actuator with Fedora 28 Cloud images.

1. Provision instance in packet.net:

   $ ./hack/packet-provision.sh install
   

   and get the IP address (assuming it is 147.75.96.139).

2. Update ~/.ssh/config to contain:

   Host libvirtactuator 147.75.96.139
   Hostname 147.75.96.139
   ControlPersist 10m
   ControlMaster auto
   ControlPath /tmp/%r@%h:%p
   User root
   StrictHostKeyChecking no
   PasswordAuthentication no
   UserKnownHostsFile ~/.ssh/aws_known_hosts
   IdentityFile /tmp/packet_id_rsa
   IdentitiesOnly yes
   Compression yes
   

3. Run init script to:

   • update qemu-kvm to at least 2.4
   • create default volume storage
   • create cloud init volume
   • create fedora_base volume

   $ bash cmd/libvirt-actuator/resources/init.sh
   

   In case the default cloud-init is not sufficient, you can generate new cloud-init.iso with the following steps:

   1. Update content of meta-data file (the filename must be meta-data)
   2. Update content of user-data (the filename must be user-data)
   3. Generate new iso with the cloud init by running:

      $ genisoimage -output cloud-init.iso -volid cidata -joliet -r user-data meta-data
      

   4. Upload the iso to the instance in case you already run the init.sh script:

      $ scp cloud-init.iso libvirtactuator:/var/lib/libvirt/images/cloud-init.iso
      

   WARNING: Please keep the /sys/firmware/qemu_fw_cfg/by_name/opt/actuator.libvirt.io.k8s.sigs/config/raw path in the user-data file as it is. Otherwise, the user data (see below) provided by kubernetes secret will get ignored.

How to prepare machine resources

By default, available user data (under user-data.sh file) contains a shell script that deploys kubernetes master node. Feel free to modify the file to your needs.

The libvirt actuator expects the user data to be provided by a kubernetes secret by setting spec.providerConfig.value.cloudInit.userDataSecret field. See userdata.yml for example.

At the same time, the spec.providerConfig.value.uri needs to be set to libvirt uri. E.g. qemu+ssh://root@147.75.96.139/system in case the libivrt instance is accessible via ssh on 147.75.96.139 IP address.

To build the libvirt-actuator binary:

You'll need to install libvirt-dev[el] installed on the system you are building and running the binary. e.g. apt-get -y install libvirt-dev or yum -y install libvirt-devel

CGO_ENABLED=1 go build -o bin/libvirt-actuator -a github.com/openshift/cluster-api-provider-libvirt/cmd/libvirt-actuator

Create libvirt instance based on machine manifest

$ ./bin/libvirt-actuator create -m examples/machine.yaml -c examples/cluster.yaml

Once the libvirt instance is created you can login inside (username: fedora, password: fedora):

$ virsh -c qemu+ssh://root@147.75.96.139/system console

Meantime you can check /root/user-data.logs to see the progress of deploying kubernetes master node:

$ watch -n 1 sudo tail -20 /root/user-data.logs

Once the deployment is done, you can list the master node:

$ sudo kubectl get nodes
NAME                   STATUS    ROLES     AGE       VERSION
fedora.example.local   Ready     master    17m       v1.11.3

Test if libvirt instance exists based on machine manifest

$ ./bin/libvirt-actuator exists -m examples/machine.yaml -c examples/cluster.yaml

Delete libvirt instance based on machine manifest

$ ./bin/libvirt-actuator delete -m examples/machine.yaml -c examples/cluster.yaml