kolla-ansible/doc/source/reference/nova-fake-driver.rst
chenxing 0002de177e Upgrade the rst convention of the Reference Guide [3]
We upgrade the rst convention by following Documentation Contributor
Guide[1].

[1] https://docs.openstack.org/doc-contrib-guide

Change-Id: Id480cd24f5eed810e81af0f12e84a4a6db49247d
Partially-Implements: blueprint optimize-the-documentation-format
2018-03-19 01:53:10 +00:00

1.5 KiB

Nova Fake Driver

One common question from OpenStack operators is that "how does the control plane (for example, database, messaging queue, nova-scheduler ) scales?". To answer this question, operators setup Rally to drive workload to the OpenStack cloud. However, without a large number of nova-compute nodes, it becomes difficult to exercise the control performance.

Given the built-in feature of Docker container, Kolla enables standing up many of Compute nodes with nova fake driver on a single host. For example, we can create 100 nova-compute containers on a real host to simulate the 100-hypervisor workload to the nova-conductor and the messaging queue.

Use nova-fake driver

Nova fake driver can not work with all-in-one deployment. This is because the fake neutron-openvswitch-agent for the fake nova-compute container conflicts with neutron-openvswitch-agent on the Compute nodes. Therefore, in the inventory the network node must be different than the Compute node.

By default, Kolla uses libvirt driver on the Compute node. To use nova-fake driver, edit the following parameters in /etc/kolla/globals.yml or in the command line options.

enable_nova_fake: "yes"
num_nova_fake_per_node: 5

Each Compute node will run 5 nova-compute containers and 5 neutron-plugin-agent containers. When booting instance, there will be no real instances created. But nova list shows the fake instances.