solar/README.md
2015-10-02 15:34:23 +02:00

299 lines
9.7 KiB
Markdown

[![Build Status](https://travis-ci.org/Mirantis/solar.svg?branch=master)](https://travis-ci.org/Mirantis/solar) [![Coverage Status](https://coveralls.io/repos/Mirantis/solar/badge.svg?branch=master&service=github)](https://coveralls.io/github/Mirantis/solar?branch=master)
# Requirements
## Supported development platforms
Linux or MacOS
## Additional software
[VirtualBox](https://www.virtualbox.org/wiki/Downloads): 5.x
[Vagrant](http://www.vagrantup.com/downloads.html): 1.7.x
Note: Make sure that [Vagrant VirtualBox Guest plugin](https://github.com/dotless-de/vagrant-vbguest) is installed `vagrant plugin install vagrant-vbguest`
Note: If you are using VirtualBox 5.0 it's worth uncommenting paravirtprovider setting in `vagrant-settings.yaml` for speed improvements:
```bash
paravirtprovider: kvm
```
For details see Customizing `vagrant-settings.yaml` section.
# Setup development env
Setup environment:
```bash
cd solar
vagrant up
```
Login into vm, the code is available in /vagrant directory
```bash
vagrant ssh
solar --help
```
Get ssh details for running slave nodes (vagrant/vagrant):
```bash
vagrant ssh-config
```
You can make/restore snapshots of boxes (this is way faster than reprovisioning them)
with the `snapshotter.py` script:
```bash
./snapshotter.py take -n my-snapshot
./snapshotter.py show
./snapshotter.py restore -n my-snapshot
```
`snapshoter.py` to run requires python module `click`.
1. On debian based systems you can install it via `sudo aptitude install python-click-cli`,
2. On fedora 22 you can install it via `sudo dnf install python-click`,
3. If you use virtualenv or similar tool then you can install it just with `pip install click`,
4. If you don't have virtualenv and your operating system does not provide package for it then `sudo pip install click`.
5. If you don't have `pip` then [install it](https://pip.pypa.io/en/stable/installing/) and then execute command step 4.
# Solar usage
For now all commands should be executed from `solar-dev` machine from `/vagrant` directory.
Basic flow is:
1. Create some resources (look at `examples/openstack/openstack.py`) and connect
them between each other, and place them on nodes.
1. Run `solar changes stage` (this stages the changes)
1. Run `solar changes process` (this prepares orchestrator graph, returning
change UUID)
1. Run `solar orch run-once <change-uuid>` (or `solar orch run-once last`
to run the lastly created graph)
1. Observe progress of orch with `watch 'solar orch report <change-uuid>'`
(or `watch 'solar orch report last'`).
Some very simple cluster setup:
```bash
cd /vagrant
solar resource create nodes templates/nodes.yaml '{"count": 2}'
solar resource create mariadb_service resources/mariadb_service '{"image": "mariadb", "root_password": "mariadb", "port": 3306}'
solar resource create keystone_db resources/mariadb_db/ '{"db_name": "keystone_db", "login_user": "root"}'
solar resource create keystone_db_user resources/mariadb_user/ user_name=keystone user_password=keystone # another valid format
solar connect node1 mariadb_service
solar connect node1 keystone_db
solar connect mariadb_service keystone_db '{"root_password": "login_password", "port": "login_port", "ip": "db_host"}'
# solar connect mariadb_service keystone_db_user 'root_password->login_password port->login_port' # another valid format
solar connect keystone_db keystone_db_user
solar changes stage
solar changes process
# <uid>
solar orch run-once <uid> # or solar orch run-once last
watch 'solar orch report <uid>' # or solar orch report last
```
You can fiddle with the above configuration like this:
```bash
solar resource update keystone_db_user '{"user_password": "new_keystone_password"}'
solar resource update keystone_db_user user_password=new_keystone_password # another valid format
solar changes stage
solar changes process
<uid>
solar orch run-once <uid>
```
To get data for the resource `bar` (raw and pretty-JSON):
```bash
solar resource show --tag 'resources/bar'
solar resource show --json --tag 'resources/bar' | jq .
solar resource show --name 'resource_name'
solar resource show --name 'resource_name' --json | jq .
```
To clear all resources/connections:
```bash
solar resource clear_all
```
Show the connections/graph:
```bash
solar connections show
solar connections graph
```
You can also limit graph to show only specific resources:
```bash
solar connections graph --start-with mariadb_service --end-with keystone_db
```
You can make sure that all input values are correct and mapped without duplicating your values with this command:
```bash
solar resource validate
```
Disconnect
```bash
solar disconnect mariadb_service node1
```
Tag a resource:
```bash
solar resource tag node1 test-tags # Remove tags
solar resource tag node1 test-tag --delete
```
# Low level API
## Usage:
Creating resources:
```python
from solar.core.resource import virtual_resource as vr
node1 = vr.create('node1', 'resources/ro_node/', 'rs/', {'ip':'10.0.0.3', 'ssh_key' : '/vagrant/tmp/keys/ssh_private', 'ssh_user':'vagrant'})[0]
node2 = vr.create('node2', 'resources/ro_node/', 'rs/', {'ip':'10.0.0.4', 'ssh_key' : '/vagrant/tmp/keys/ssh_private', 'ssh_user':'vagrant'})[0]
keystone_db_data = vr.create('mariadb_keystone_data', 'resources/data_container/', 'rs/', {'image' : 'mariadb', 'export_volumes' : ['/var/lib/mysql'], 'ip': '', 'ssh_user': '', 'ssh_key': ''}, connections={'ip' : 'node2.ip', 'ssh_key':'node2.ssh_key', 'ssh_user':'node2.ssh_user'})[0]
nova_db_data = vr.create('mariadb_nova_data', 'resources/data_container/', 'rs/', {'image' : 'mariadb', 'export_volumes' : ['/var/lib/mysql'], 'ip': '', 'ssh_user': '', 'ssh_key': ''}, connections={'ip' : 'node1.ip', 'ssh_key':'node1.ssh_key', 'ssh_user':'node1.ssh_user'})[0]
```
To make connection after resource is created use `signal.connect`.
To test notifications:
```python
keystone_db_data.args # displays node2 IP
node2.update({'ip': '10.0.0.5'})
keystone_db_data.args # updated IP
```
If you close the Python shell you can load the resources like this:
```python
from solar.core import resource
node1 = resource.load('rs/node1')
node2 = resource.load('rs/node2')
keystone_db_data = resource.load('rs/mariadb_keystone_data')
nova_db_data = resource.load('rs/mariadb_nova_data')
```
Connections are loaded automatically.
You can also load all resources at once:
```python
from solar.core import resource
all_resources = resource.load_all('rs')
```
## Dry run
Solar CLI has possibility to show dry run of actions to be performed.
To see what will happen when you run Puppet action, for example, try this:
```bash
solar resource action keystone_puppet run -d
```
This should print out something like this:
```bash
EXECUTED:
73c6cb1cf7f6cdd38d04dd2d0a0729f8: (0, 'SSH RUN', ('sudo cat /tmp/puppet-modules/Puppetfile',), {})
3dd4d7773ce74187d5108ace0717ef29: (1, 'SSH SUDO', ('mv "1038cb062449340bdc4832138dca18cba75caaf8" "/tmp/puppet-modules/Puppetfile"',), {})
ae5ad2455fe2b02ba46b4b7727eff01a: (2, 'SSH RUN', ('sudo librarian-puppet install',), {})
208764fa257ed3159d1788f73c755f44: (3, 'SSH SUDO', ('puppet apply -vd /tmp/action.pp',), {})
```
By default every mocked command returns an empty string. If you want it to return
something else (to check how would dry run behave in different situation) you provide
a mapping (in JSON format), something along the lines of:
```bash
solar resource action keystone_puppet run -d -m "{\"73c\": \"mod 'openstack-keystone'\n\"}"
```
The above means the return string of first command (with hash `73c6c...`) will be
as specified in the mapping. Notice that in mapping you don't have to specify the
whole hash, just it's unique beginning. Also, you don't have to specify the whole
return string in mapping. Dry run executor can read file and return it's contents
instead, just use the `>` operator when specifying hash:
```bash
solar resource action keystone_puppet run -d -m "{\"73c>\": \"./Puppetlabs-file\"}"
```
# Resource compiling
You can compile all `meta.yaml` definitions into Python code with classes that
derive from `Resource`. To do this run
```bash
solar resource compile_all
```
This generates file `resources_compiled.py` in the main directory (do not commit
this file into the repo). Then you can import classes from that file, create
their instances and assign values just like these were normal properties.
If your editor supports Python static checking, you will have autocompletion
there too. An example on how to create a node with this:
```python
import resources_compiled
node1 = resources_compiled.RoNodeResource('node1', None, {})
node1.ip = '10.0.0.3'
node1.ssh_key = '/vagrant/.vagrant/machines/solar-dev1/virtualbox/private_key'
node1.ssh_user = 'vagrant'
```
# Higher-level API
There's also a higher-level API that allows to write resource instances in more
functional way, and in particular avoid `for` loops. Here's an example:
```python
from solar import template
nodes = template.nodes_from('templates/riak_nodes.yaml')
riak_services = nodes.on_each(
'resources/riak_node',
{
'riak_self_name': 'riak{num}',
'riak_hostname': 'riak_server{num}.solar',
'riak_name': 'riak{num}@riak_server{num}.solar',
}
)
riak_master_service = riak_services.take(0)
riak_slave_services = riak_services.tail()
riak_master_service.connect_list(
riak_slave_services,
{
'riak_name': 'join_to',
}
)
```
For full Riak example, please look at `examples/riak/riaks-template.py`.
Full documentation of individual functions is found in the `solar/template.py` file.
# Customizing vagrant-settings.yaml
Solar is shipped with sane defaults in `vagrant-setting.yaml_defaults`. If you need to adjust them for your needs, e.g. changing resource allocation for VirtualBox machines, you should just compy the file to `vagrant-setting.yaml` and make your modifications.