commit | 733504de81cc46ee1d653b509dcbbfc25080edf4 | [log] [tgz] |
---|---|---|
author | Max Chu <maxmchu@gmail.com> | Fri Aug 18 15:04:53 2017 -0700 |
committer | Max Chu <maxmchu@gmail.com> | Fri Aug 18 15:10:57 2017 -0700 |
tree | 40a71cf46290672f6665c75ae4d147a5186ea665 | |
parent | 0ecceded4a04790d9d44d16e0e99afcadb07b668 [diff] |
fixing ecord-global and juju-facts Change-Id: I849e2e30ab58211231bf2c9e8ea2413caef6ccec
This repository contains Ansible playbooks for installing and configuring software components that build a CORD POD: OpenStack, ONOS, and XOS.
It is used as a sub-module of the main CORD repository, but can independently bring up various CORD profiles for development work.
If you want to set up an entire CORD pod on physical hardware, or set up the Cord-in-a-Box deployment, you should start at the CORD repository.
There's a helper script, scripts/cord-bootstrap.sh. that will install development environment prerequisites on a Ubuntu 14.04 node. You can download it with:
curl -o ~/cord-bootstrap.sh https://raw.githubusercontent.com/opencord/platform-install/master/scripts/cord-bootstrap.sh
Running the script will install the repo tool, Ansible, and Docker, as well as make a checkout of the CORD manifest into /opt/cord
.
You can specify which gerrit changesets you would like repo to checkout using the -b
option on the script as documented here.
Once you have done this, if you're not already in the docker
group, you should logout and log back into your system to refresh your user account's group membership. If you don't do this, any docker command you run or ansible runs for you will fail. You can check your group membership by running groups
.
Once you log back in, you may want to run tmux
to maintain a server-side session you can reconnect to, in case of network trouble.
All of the commands below assume you're in the /opt/cord/build/platform-install
directory.
If you are doing work that does not involve openstack
you can create a development environment inside a VM on your local machine. This environment is mostly designed to do GUI
, APIs
and modeling
related work. It can also be useful to test a synchronizer
whose job is synchronizing data using REST APIs.
To do that we provided a Vagrant
VM. From this folder just execute vagrant up head-node
. We'll create an Ubuntu based VM with all the cord
code shared from your local machine, so that you can make your changes locally and quickly test the outcome.
Once the vm
is created you can connect to it with vagrant ssh head-node
and then from the /opt/cord/build/platform-install
execute the profile you are interested in. For instance you can spin up the frontend
configuration with: ansible-playbook -i inventory/frontend deploy-xos-playbook.yml
.
Note that the cord-bootstrap.sh
script is automatically invoked by the provisioning script and the vagrant
vm requires VirtualBox
Credentials will be autogenerated and placed in the credentials/
directory when the playbooks are run, where the credential name is the filename, and the contents of the file is the password.
For most profiles the XOS admin user is named xosadmin@opencord.org
.
Most profiles are run by specifying an inventory file when running ansible-playbook
. For most frontend or mock profiles, you'll want to run the deploy-xos-playbook.yml
playbook.
For example, to run the frontend
config, you would run:
ansible-playbook -i inventory/frontend deploy-xos-playbook.yml
Setting up the initial environment, and launching XOS for the first time, will take about 30 minutes. You can then explore the environment you've set up.
Suppose you've made a change to your code and want the changes reflected in the running XOS containers. To get there, run:
ansible-playbook -i inventory/frontend redeploy-xos-playbook.yml
Rebuilding and redeploying containers using this method will usually take much less time than the initial setup phase. All changes to the local environment's source tree should be reflected in the redeployed containers. Note that running the above playbook will preserve the existing database.
When you're ready to tear down your environment, for example to switch to another profile, or to launch the current profile again from a fresh database, run:
ansible-playbook -i inventory/frontend teardown-playbook.yml
This will destroy all the docker containers created. Note that you must run the teardown-playbook.yml
using the same inventory file it was created with, otherwise rogue docker containers may still be running after the teardown.
You can then make changes to code and re-run the same or a different profile.
To create a new CORD profile, you should:
inventory/
that defines the cord_profile
variable, with the name of your profile.[all:vars] cord_profile=my-profile
Create a .yaml variables file in profile_manifests/
with the name of your profile (ex: my-profile.yaml
), and populate it with your configuration.
To test the profile, run the deploy-xos-playbook.yml
playbook using your inventory profile: ansible-playbook -i inventory/my-profile deploy-xos-playbook.yml
Before commit, please run ./scripts/lintcheck.sh .
in the repo root, which will perform the same ansible-lint check that Jenkins performs when in review in gerrit.
This profile runs API tests for both the REST and TOSCA APIs. This can be done in an automated fashion:
ansible-playbook -i inventory/api-test api-test-playbook.yml
The XOS credentials in this config are padmin@vicci.org
and letmein
(until the tests are modified to support generated credentials).
Builds a basic XOS frontend installation, useful for UI testing and experimentation.
Builds a Mock R-CORD or Mock M-CORD pod, without running service synchronizers in a manner similar to frontend.
Used as a part of the OpenCloud deployment. Similar to rcord
.
This is a part of the R-CORD deployment - start using the steps specified in that repo.
This profile is designed to integrate XOS with physical infrastructure pieces like MaaS, OpenStack, and ONOS. See the CORD-in-a-Box Quick Start Guide for how to set up a virtual multi-node R-CORD pod on a single host.
If you've already built a CiaB and want to go through the dev loop described above on the head/production node, when running ansible-playbook
you must pass the path to the configuration file that was generated by Gradle during the build:
ansible-playbook -i inventory/head-localhost --extra-vars @../genconfig/config.yml playbook.yml
You may find the following shell aliases to be helpful during development:
alias xos-teardown="pushd /opt/cord/build/platform-install; ansible-playbook -i inventory/head-localhost --extra-vars @/opt/cord/build/genconfig/config.yml teardown-playbook.yml" alias xos-launch="pushd /opt/cord/build/platform-install; ansible-playbook -i inventory/head-localhost --extra-vars @/opt/cord/build/genconfig/config.yml launch-xos-playbook.yml" alias compute-node-refresh="pushd /opt/cord/build/platform-install; ansible-playbook -i /etc/maas/ansible/pod-inventory --extra-vars=@/opt/cord/build/genconfig/config.yml compute-node-refresh-playbook.yml"
With these aliases, tearing down XOS and rebuilding it from a fresh database becomes:
xos-teardown; xos-launch; compute-node-refresh
cord_profile
: name of the profile_manifest to use.
Paths on configuration node (where playbooks are run, may also be build node)
config_cord_dir
location on configuration node of cord dirconfig_cord_profile_dir
location on configuration node of cord_profile dirpki_dir
, ssh_pki_dir
: where SSL and SSH certificates are created on config nodecredentials_dir
- location where autogenerated passwords file are createdPaths on head node (target system operated on by playbooks)
head_cord_dir
- where the cord
directory is copied to on the head node (deprecated when we reach container-only deploys)head_cord_profile_dir
- location of the cord_profile
directory on the head nodehead_onos_cord_dir
- location of the onos-cord
directory on the head nodehead_onos_fabric_dir
- location of the onos-fabric
directory on the head nodePrefix every role file with the yaml start block and comment with name of file relative to role base.
--- # rolename/tasks/main.yml
When using templates, put the template filename and path within the role as a comment in the template so that it's easy to determine which template was used to create a file after it's been created.
If you use a variable that isn't created by the ansible setup task, define it in the role defaults file. The default value for anay variable must be the same across all role defaults.
Use the YAML style syntax for tasks, not the older =
syntax, as the former is more likely to indent and syntax highlight properly in most editors.
Roles should always have the same outcome, so avoid using conditionals or tags to change the behavior of a role.