{"id":310,"date":"2016-09-23T23:34:18","date_gmt":"2016-09-23T16:34:18","guid":{"rendered":"http:\/\/www.rickyadams.com\/wp\/?p=310"},"modified":"2016-09-23T23:34:41","modified_gmt":"2016-09-23T16:34:41","slug":"310","status":"publish","type":"post","link":"https:\/\/www.rickyadams.com\/wp\/310\/","title":{"rendered":"OpenStack Environment Architecture"},"content":{"rendered":"
\n
\n

OpenStack Environment Architecture<\/h2>\n
Fuel deploys an OpenStack Environment with nodes that provide a specific set of functionality. Beginning with Fuel 5.0, a single architecture model can support HA (High Availability) and non-HA deployments; you can deploy a non-HA environment and then add additional nodes to implement HA rather than needing to redeploy the environment from scratch.<\/div>\n
The OpenStack environment consists of multiple physical server nodes (or an equivalent VM), each of which is one of the following node types:<\/div>\n
\n
Controller:<\/dt>\n
\n
The\u00a0Controller<\/em><\/a>\u00a0manages all activities in the environment. The\u00a0nova-controller<\/cite>\u00a0maintains the life cycle of the\u00a0OpenStack<\/em><\/a>\u00a0controller.<\/div>\n
\n
Note<\/div>\n
HA<\/em><\/a>\u00a0environment must consist of at least 3 controllers in order to achieve HA for\u00a0MySQL\/Galera<\/em><\/a>\u00a0cluster. And while two controllers could be enough for most of cases, such as HA for highly available OpenStack API services or reliable\u00a0RabbitMQ AMQP<\/em><\/a>\u00a0messaging or resilient virtual IP addresses and load balancing, a third controller is required for quorum-based clusters, such as MySQL\/Galera or\u00a0Corosync\/Pacemaker<\/em><\/a>. The configuration for stateless and statefull services in HA differs a lot. HA environments also contain active\/active and active\/passive components. Please, see\u00a0HA-guide<\/a>\u00a0for more details. Fuel configures all stateless OpenStack API services and RabbitMQ HA cluster as active\/active. The MySQL\/Galera cluster is configured as active\/passive. For database clusters, active\/active is sometimes referred to as multi-master environments. Such environments should be able to successfully handle multi-node writing conflicts. But OpenStack support for multi-node writing to MySQL\/Galera nodes is\u00a0not production ready yet<\/a>. “The simplest way to overcome this issue from the operator\u2019s point of view is to use only one writer node for these types of transactions”. That is why Fuel configures\u00a0HAProxy<\/em><\/a>\u00a0frontend for MySQL\/Galera to use only one active node, while the other nodes in the cluster are retained standby (passive) state.\u00a0Mongo<\/em><\/a>\u00a0DB backend for\u00a0Ceilometer<\/em><\/a>\u00a0is also configured as active\/passive. Note that it is possible to configure MySQL\/Galera HA with two controller nodes and a lightweight arbitrator service running at some other node, but this deployment layout is not supported in Fuel.<\/div>\n<\/div>\n
For more information about how Fuel deploys HA controllers, see\u00a0Multi-node with HA Deployment<\/em><\/a>.<\/div>\n<\/dd>\n
Compute:<\/dt>\n
\n
Compute servers are the workhorses of your installation; they are the servers on which your users’ virtual machines are created.\u00a0nova-compute<\/cite>\u00a0controls the life cycle of these VMs; Neutron Agent and Ceilometer Compute Agent may also run on Compute nodes.<\/div>\n
\n
Note<\/div>\n
In environments that Fuel deploys using vCenter as the hypervisor, the\u00a0Nova-compute<\/em><\/a>\u00a0service can run only on Controller nodes. Because of this, Fuel does not allow you to\u00a0assign<\/em><\/a>\u00a0the “Compute” role to any node when using vCenter.<\/div>\n<\/div>\n<\/dd>\n
Storage:<\/dt>\n
\n
OpenStack requires block and object storage to be provisioned. These can be provisioned as Storage nodes or as roles that run on Compute nodes. Fuel provides the following storage options out of the box:<\/div>\n