Installation VM compute
Les noeuds de calculs correspondent aux hyperviseurs. Ils sont pilotés par la brique Nova, et possèdent en général un agent neutron pour la configuration des réseaux dédiés aux VMs.
Installation de la partie Nova
Pour installer les paquets :
yum install openstack-nova-compute
Configuration de Nova
Fichier de configuration /etc/nova/nova.conf :
[DEFAULT] auth_strategy=keystone my_ip=192.168.2.69 use_neutron=true enabled_apis=osapi_compute,metadata firewall_driver = nova.virt.firewall.NoopFirewallDriver rpc_backend=rabbit [api_database] [barbican] [cache] [cells] [cinder] [cloudpipe] [conductor] [cors] [cors.subdomain] [crypto] [database] [ephemeral_storage_encryption] [glance] api_servers = http://glance.internal:9292 [guestfs] [hyperv] [image_file_url] [ironic] [key_manager] [keystone_authtoken] auth_uri = http://controller.internal:5000 auth_url = http://controller.internal:35357 memcached_servers = controller.internal:11211 auth_type = password project_domain_name = default user_domain_name = default project_name = service username = nova password = nova [libvirt] [matchmaker_redis] [metrics] [mks] [neutron] [osapi_v21] [oslo_concurrency] lock_path = /var/lib/nova/tmp [oslo_messaging_amqp] [oslo_messaging_notifications] [oslo_messaging_rabbit] rabbit_host = bdd rabbit_userid = openstack rabbit_password = RABBIT_PASS [oslo_messaging_zmq] [oslo_middleware] [oslo_policy] [placement] [placement_database] [rdp] [remote_debug] [serial_console] [spice] [ssl] [trusted_computing] [upgrade_levels] [vmware] [vnc] enabled = True vncserver_listen = 0.0.0.0 vncserver_proxyclient_address = $my_ip novncproxy_base_url = http://controller.public:6080/vnc_auto.html [workarounds] [wsgi] [xenserver] [xvp]
Activer et demarrer les services Nova de l'hyperviseur :
systemctl enable libvirtd openstack-nova-compute systemctl start libvirtd openstack-nova-compute
Configuration de la partie Neutron
Installer les paquets des agents nécessaires :
yum install openstack-neutron-openvswitch ipset
Configuration de l'agent openvswitch dans /etc/neutron/plugins/ml2/openvswitch_agent.ini :
[DEFAULT] [agent] tunnel_types = gre,vxlan l2_population = True [ovs] local_ip = 192.168.200.4 bridge_mappings = vlan:br-vlan [securitygroup] firewall_driver = iptables_hybrid
Configuration de la partie Neutron dans /etc/neutron/neutron.conf :
[DEFAULT] auth_strategy = keystone rpc_backend = rabbit [agent] [cors] [cors.subdomain] [database] [keystone_authtoken] auth_uri = http://controller.internal:5000 auth_url = http://controller.internal:35357 memcached_servers = controller.internal:11211 auth_type = password project_domain_name = default user_domain_name = default project_name = service username = neutron password = neutron [matchmaker_redis] [nova] [oslo_concurrency] lock_path = /var/lib/neutron/tmp [oslo_messaging_amqp] [oslo_messaging_notifications] [oslo_messaging_rabbit] rabbit_host = bdd rabbit_userid = openstack rabbit_password = RABBIT_PASS [oslo_messaging_zmq] [oslo_policy] [qos] [quotas] [ssl]
Activer et démarrer les services suivants :
systemctl enable neutron-openvswitch-agent openvswitch systemctl start neutron-openvswitch-agent openvswitch
Configuration de la libvirt
Pour permettre les migrations à chaud, il faut ajouter l'option –listen via le fichier /etc/sysconfig/libvirtd. Il faut de plus autoriser le tcp, sans le TLS (à voir pour plus tard), ainsi que d'accepter les connexions sans authentification en TCP. Cela se fait dans /etc/libvirt/libvirtd.conf (auth_tcp = none).