Node Configuration¶
This section explains the various ways a user can configure nodes.
Libvirt Node Configuration¶
You configure the overcloud nodes by placing variable definitions in a
YAML file and passing that to ansible using the -N
command line
option, like this:
quickstart.sh -N config/nodes/1ctlr_1comp.yml
Setting number and type of overcloud nodes¶
The overcloud_nodes
variable can be used to change the number and
type of nodes deployed in your overcloud. Begining of the default
(config/nodes/1ctlr_1comp.yml) looks like this.
overcloud_nodes:
- name: control_0
flavor: control
virtualbmc_port: 6230
default_name: overcloud-controller-0
hostnamemap_override: overcloud-controller-foo-0
- name: compute_0
flavor: compute
virtualbmc_port: 6231
default_name: overcloud-novacompute-0
hostnamemap_override: overcloud-novacompute-bar-0
You can use your own config if you want to test a different setup. For example:
overcloud_nodes:
- name: control_0
flavor: control
virtualbmc_port: 6230
- name: control_1
flavor: control
virtualbmc_port: 6231
- name: control_2
flavor: control
virtualbmc_port: 6232
- name: compute_0
flavor: compute
virtualbmc_port: 6233
- name: ceph_0
flavor: ceph
virtualbmc_port: 6234
- name: swift_0
flavor: objectstorage
virtualbmc_port: 6235
Controlling resources¶
These variables set the resources that will be assigned to a node by default, unless overridden by a more specific variable:
default_disk
default_memory
default_vcpu
These variables set the resources assigned to the undercloud node:
undercloud_disk
undercloud_memory
(defaults to 12288)undercloud_vcpu
(defaults to 4)
These variables set the resources assigned to controller nodes:
control_disk
control_memory
control_vcpu
These variables control the resources assigned to compute nodes:
compute_disk
compute_memory
compute_vcpu
These variables control the resources assigned to ceph storage nodes:
ceph_disk
ceph_memory
ceph_vcpu
There is another option extradisks
that can be used to create three
additional blockdevices vdb
, vdc
and vdd
per node. By default it is
only enabled on the objectstorage node flavor. Note that ironic will pick the
smallest disk available in the node when there are multiple disks attached. You
must either set the same size to all the disks using extradisks_size
or
provide root_device_hints
to set in ironic.
An example¶
To create a minimal environment that would be unsuitable for deploying
anything real nova instances, you could place something like the
following in myconfigfile.yml
:
undercloud_memory: 8192
control_memory: 6000
compute_memory: 2048
overcloud_nodes:
- name: control_0
flavor: control
virtualbmc_port: 6230
- name: compute_0
flavor: compute
virtualbmc_port: 6231
And then pass that to the ansible-playbook
command as described at
the beginning of this document.
Baremetal Node Configuration¶
Baremetal deployments are unique from libvirt virtual deployments in that the hardware, the specs, and network settings can not be adjusted via a configuration file. These settings for each individual baremetal deployment are unique and must be stored separately.
What the baremetal node configuration baremetal.yml
does is essentially
ensuring that no libvirt guests are provisioned setting overcloud_nodes to
null:
overcloud_nodes:
The pattern and layout for baremetal hardware configuration can be found in this doc
For additional support with baremetal deployments please visit the #oooq channel on OFTC irc.
OpenStack Virtual Baremetal Node Configuration¶
Using OpenStack Virtual Baremetal is a simple node configuration where the user needs only to define how many cloud instances to run.
For example, you will find the following config in the node configuration files:
# Define the controller node and compute nodes.
# Create three controller nodes and one compute node.
node_count: 4
The remaining configuration for the instances like the flavor types are found in the config/environment/ configuration as this may vary based on your cloud provider. For an example please reference this configuration