Juju Charm - Nova Compute
Go to file
Corey Bryant c8dea9f6da Switch to yoga unit jobs and add py39
Change-Id: If2665b63728eb5a800fcb025a642d0413f54157a
2021-11-17 14:49:03 -05:00
actions Spelling fixes found by codespell. 2021-11-15 21:54:13 +00:00
files Support libvirt group instead of libvirtd when hugepages set up 2018-02-07 12:18:30 +09:00
hooks Spelling fixes found by codespell. 2021-11-15 21:54:13 +00:00
lib Added actions that enable clean removal of nova-compute unit from model 2021-03-24 16:07:58 +01:00
templates Spelling fixes found by codespell. 2021-11-15 21:54:13 +00:00
tests Add yoga bundles and release-tool syncs 2021-11-12 11:18:58 -05:00
unit_tests Spelling fixes found by codespell. 2021-11-15 21:54:13 +00:00
.gitignore Fix up amulet for queens 2018-02-25 16:06:08 +00:00
.gitreview OpenDev Migration Patch 2019-04-19 19:34:02 +00:00
.project [trivial] fixup pydev project 2013-11-17 21:48:39 +00:00
.pydevproject Misc fixes for charm infra 2020-05-26 09:38:07 +01:00
.stestr.conf Switch to stestr for unit execution 2019-03-06 17:19:20 +00:00
.zuul.yaml Added actions that enable clean removal of nova-compute unit from model 2021-03-24 16:07:58 +01:00
actions.yaml Reword security-checklist action definition 2021-08-04 12:08:14 +02:00
bindep.txt Added actions that enable clean removal of nova-compute unit from model 2021-03-24 16:07:58 +01:00
charm-helpers-hooks.yaml Updates to flip all libraries back to master 2021-05-03 16:09:13 +01:00
config.yaml Allow overriding libvirt/num_pcie_ports 2021-09-28 12:31:20 +00:00
copyright Re-license charm as Apache-2.0 2016-07-03 16:37:37 +00:00
hardening.yaml Add hardening support 2016-03-24 11:18:41 +00:00
icon.svg Update charm icon 2017-08-02 18:21:57 +01:00
LICENSE Re-license charm as Apache-2.0 2016-07-03 16:37:37 +00:00
Makefile Sync helpers for 20.05 2020-05-18 14:47:34 +02:00
metadata.yaml Add yoga bundles and release-tool syncs 2021-11-12 11:18:58 -05:00
osci.yaml Switch to yoga unit jobs and add py39 2021-11-17 14:49:03 -05:00
pip.sh Add xena bundles 2021-10-04 15:18:26 +02:00
README.md Overhaul README 2021-09-13 13:32:38 -04:00
requirements.txt Sync release-tools 2021-07-22 14:59:02 +02:00
revision [hopem] added support for libvirt RBD imagebackend 2014-06-02 19:37:32 +01:00
test-requirements.txt Add yoga bundles and release-tool syncs 2021-11-12 11:18:58 -05:00
tox.ini Switch to yoga unit jobs and add py39 2021-11-17 14:49:03 -05:00

Overview

The nova-compute charm deploys Nova Compute, the core OpenStack service that provisions virtual instances (VMs) and baremetal servers (via Ironic). The charm works alongside other Juju-deployed OpenStack services.

Usage

Configuration

This section covers common and/or important configuration options. See file config.yaml for the full list of options, along with their descriptions and default values. See the Juju documentation for details on configuring applications.

config-flags

A comma-separated list of key=value configuration flags. These values will be placed in the [DEFAULT] section of the nova.conf file.

enable-live-migration

Allows the live migration of VMs.

enable-resize

Allows the resizing of VMs.

migration-auth-type

Selects the TCP authentication scheme to use for live migration. The only accepted value is 'ssh'.

customize-failure-domain

When MAAS is the backing cloud and this option is set to 'true' then all MAAS-defined zones will become available as Nova availability zones, and option default-availability-zone will be overridden. See section Availability Zones.

default-availability-zone

Sets a single default Nova availability zone. It is used when a VM is created without a Nova AZ being specified. The default value is 'nova'. A non-default Nova AZ must be created manually (i.e. openstack aggregate create). See section Availability Zones.

libvirt-image-backend

Specifies what image backend to use. Possible values are 'rbd', 'qcow2', 'raw', and 'flat'. The default behaviour is for Nova to use qcow2.

openstack-origin

States the software sources. A common value is an OpenStack UCA release (e.g. 'cloud:bionic-train' or 'cloud:focal-wallaby'). See Ubuntu Cloud Archive. The underlying host's existing apt sources will be used if this option is not specified (this behaviour can be explicitly chosen by using the value of 'distro').

pool-type

Dictates the Ceph storage pool type. See sections Ceph pool type and RBD Nova images for more information.

Ceph pool type

Ceph storage pools can be configured to ensure data resiliency either through replication or by erasure coding. This charm supports both types via the pool-type configuration option, which can take on the values of 'replicated' and 'erasure-coded'. The default value is 'replicated'.

For this charm, the pool type will be associated with Nova-managed images.

Note

: Erasure-coded pools are supported starting with Ceph Luminous.

Replicated pools

Replicated pools use a simple replication strategy in which each written object is copied, in full, to multiple OSDs within the cluster.

The ceph-osd-replication-count option sets the replica count for any object stored within the 'nova' rbd pool. Increasing this value increases data resilience at the cost of consuming more real storage in the Ceph cluster. The default value is '3'.

Important

: The ceph-osd-replication-count option must be set prior to adding the relation to the ceph-mon application. Otherwise, the pool's configuration will need to be set by interfacing with the cluster directly.

Erasure coded pools

Erasure coded pools use a technique that allows for the same resiliency as replicated pools, yet reduces the amount of space required. Written data is split into data chunks and error correction chunks, which are both distributed throughout the cluster.

Note

: Erasure coded pools require more memory and CPU cycles than replicated pools do.

When using erasure coding two pools will be created: a replicated pool (for storing RBD metadata) and an erasure coded pool (for storing the data written into the RBD). The ceph-osd-replication-count configuration option only applies to the metadata (replicated) pool.

Erasure coded pools can be configured via options whose names begin with the ec- prefix.

Important

: It is strongly recommended to tailor the ec-profile-k and ec-profile-m options to the needs of the given environment. These latter options have default values of '1' and '2' respectively, which result in the same space requirements as those of a replicated pool.

See Ceph Erasure Coding in the OpenStack Charms Deployment Guide for more information.

Ceph BlueStore compression

This charm supports BlueStore inline compression for its associated Ceph storage pool(s). The feature is enabled by assigning a compression mode via the bluestore-compression-mode configuration option. The default behaviour is to disable compression.

The efficiency of compression depends heavily on what type of data is stored in the pool and the charm provides a set of configuration options to fine tune the compression behaviour.

Note

: BlueStore compression is supported starting with Ceph Mimic.

Deployment

These deployment instructions assume that the following applications are present: glance, nova-cloud-controller, ovn-chassis, and rabbitmq-server. Storage backends used for VM disks and volumes are configured separately (see sections Ceph backed storage and Local Cinder storage.

Let file nova-compute.yaml contain the deployment configuration:

    nova-compute:
      config-flags: default_ephemeral_format=ext4
      enable-live-migration: true
      enable-resize: true
      migration-auth-type: ssh
      openstack-origin: cloud:focal-wallaby

To deploy nova-compute to machine '5':

juju deploy --to 5 --config nova-compute.yaml nova-compute
juju add-relation nova-compute:image-service glance:image-service
juju add-relation nova-compute:cloud-compute nova-cloud-controller:cloud-compute
juju add-relation nova-compute:neutron-plugin ovn-chassis:nova-compute
juju add-relation nova-compute:amqp rabbitmq-server:amqp

Ceph backed storage

Two concurrent Ceph backends are supported: RBD Nova images and RBD Cinder volumes. Each backend uses its own set of cephx credentials.

The steps below assume a pre-existing Ceph cluster (see the ceph-mon and ceph-osd charms).

RBD Nova images

RBD Nova images are enabled by setting option libvirt-image-backend to 'rbd' and by adding a relation to the Ceph cluster:

juju config nova-compute libvirt-image-backend=rbd
juju add-relation nova-compute:ceph ceph-mon:client

Warning

: Changing the value of option libvirt-image-backend will orphan any disks that were set up under a different setting. This will cause the restarting of associated VMs to fail.

This solution will place both root and ephemeral disks in Ceph.

Pro tip: An alternative is to selectively store just root disks in Ceph by using Cinder as an intermediary. See section RBD Cinder volumes as well as Launch an instance from a volume in the Nova documentation.

RBD Cinder volumes

RBD Cinder volumes are enabled by adding a relation to Cinder via the cinder-ceph application. Assuming Cinder is already backed by Ceph (see the cinder-ceph charm):

juju add-relation nova-compute:ceph-access cinder-ceph:ceph-access

Note

: The nova-compute:ceph-access relation is not needed for OpenStack releases older than Ocata.

Local Cinder storage

To use local storage, Cinder will need to be configured to use local block devices. See the cinder charm for details.

Availability Zones

Nova AZs can be matched with MAAS zones depending on how options default-availability-zone and customize-failure-domain are configured. See Availability Zones in the OpenStack Charms Deployment Guide for in-depth coverage of how this works.

SSH keys and VM migration

VM migration requires the sharing of public SSH keys (host and several select users) among the compute hosts. By design, only those hosts belonging to the same application group will get each other's keys. This means that VM migration cannot occur (without manual intervention) between hosts belonging to different groups.

Note: The policy of only sharing SSH keys amongst hosts of the same application group may be struck down. This is being tracked in bug LP #1468871.

NFV support

This charm (in conjunction with the nova-cloud-controller and neutron-api charms) supports NFV for Compute nodes that are deployed in Telco NFV environments.

For more information on NFV see the Network Functions Virtualization (NFV) page in the OpenStack Charms Deployment Guide.

Network spaces

This charm supports the use of Juju network spaces (Juju v.2.0). This feature optionally allows specific types of the application's network traffic to be bound to subnets that the underlying hardware is connected to.

Note

: Spaces must be configured in the backing cloud prior to deployment.

In addition this charm declares two extra-bindings:

  • internal: used to determine the network space to use for console access to instances.

  • migration: used to determine which network space should be used for live and cold migrations between hypervisors.

Note that the nova-cloud-controller application must have bindings to the same network spaces used for both 'internal' and 'migration' extra bindings.

Scaling back

Scaling back the nova-compute application implies the removal of one or more compute nodes. This is documented as a cloud operation in the OpenStack Charms Deployment Guide. See Remove a Compute node.

Actions

This section lists Juju actions supported by the charm. Actions allow specific operations to be performed on a per-unit basis. To display action descriptions run juju actions nova-compute. If the charm is not deployed then see file actions.yaml.

  • disable
  • enable
  • hugepagereport
  • instance-count
  • list-compute-nodes
  • node-name
  • openstack-upgrade
  • pause
  • register-to-cloud
  • remove-from-cloud
  • resume
  • security-checklist

Documentation

The OpenStack Charms project maintains two documentation guides:

Bugs

Please report bugs on Launchpad.