Just one-shade-of-openstack

Just one Shade of OpenStack
Cloud applications are simple!

Who? What? Why?
Roberto Polli - Solutions Architect @ par-tec.it. Loves writing in C,
Java and Python. RHC{E,VA}, MySQL|MongoDB Certified DBA.
Par-Tec – Proud sponsor of this talk ;) provides expertise in IT
Infrastructure & Services and Business Intelligence solutions +
Vertical Applications for the financial market. Contributes to various
FLOSS.
Manage OpenStack with python Shade library to simplify
automation.

Agenda
⬝ Intro to OpenStack
⬝ The day I met Shade (Quickstart)
⬝ Building the cloud
⬝ Shade in Ansible Modules
⬝ A contribution story

OpenStack is Programmable Infrastructure
Programmable Infrastructure made open
- Amazon Web Services like (IaaS)
- Deploy Virtual {Machines,Network,...} on the internet
- REST interface
API Specs + python implementation
Industry backed (Red Hat, Rackspace, HP, Cisco, Oracle, ..)

OpenStack Workflows
Web Interface
API Endpoint
Queues
Datastores

OpenStack but... what's a Stack?
Virtualized infrastructure made of multiple components:
- load balancers
- web servers
- networks and routers
- datastores
- ...

OpenStack but... what's a Stack?

Deploy OpenStack Applications - CLIent API
$ openstack server create --image centos-7 --flavor m1.tiny web-1
$ openstack network create my-dmz
$ openstack subnet create --network my-dmz ...
$ openstack volume create --size 50 my-disk
$ openstack server add volume web-1 my-disk
https://docs.openstack.org/user-guide/cli-cheat-sheet.html

Deploy OpenStack Applications: Heat Templates
Heat is the Orchestration component.
- group resources into a HOT
- Heat Orchestration Template
- reuse them
Stacks can be updated and deleted,
Heat takes care of them.
You can compose templates specifying
a filename or an URI
Stacks are *persistent*
heat_template_version: 2016-04-08
description: >
A Template is a yaml file describing resources
and getting parameters in input.
parameters:
service_name:
type: string
...
resources:
web-1:
type: OS::Nova::Server
properties:
name: { get_param: service_name }
...
network:
type: https://git.it/network_template.yaml
properties:
...

Deploy OpenStack Applications: Heat Templates

Deploy OpenStack Applications: Ansible
OpenStack Modules
- more flexible than HOT
- removing a stack won't remove
created resources
- requires an external orchestrator
- can invoke heat (eg. for testing
HOTs)
HOT Quota is planned only in Ocata.
Older versions can't do it
- name: Yes, you can
hosts: localhost
...
tasks:
- name: Create and configure
os_project:
state: present
name: a_project
description: with Ansible
domain: default
enabled: true
wait: yes
-name: Create a user
os_user: name="joe" default_project="a_project"
-name: Manage ACL
os_user_role:
user: joe
role: admin
project: a_project

Meet Shade
I was working with various OpenStack projects when at EP16 in Bilbao I followed
a training on Cloud Applications
Goals:
- simplify openstack usage
- manage multiple openstack clouds
- more interaction
JOIN
EUROPYTHON
2017

OpenStack Applications
Use Shade directly:
- flexibility
- custom workflow
Apache libcloud:
- multi-cloud
- more generic

Shade Quickstart - 1/3
$ pip install shade # Install Shade
$ vim clouds.yml # Store credentials
$ ipython # Run (i)python
import shade
# Connect to the first cloud
mycloud = shade.operator_cloud(cloud='mycloud')
# List virtual servers
servers = mycloud.list_servers()
# List server names
[x.name for x in servers]
[
u'os3-node1',
u'os3-master0',
u'os3-master1',
...
]
# clouds.yml is the credential file
identity_api_version: '3.0'
clouds:
# Insert here your credentials.
mycloud:
verify: true # or false
auth:
username: admin
password: secret
auth_url: 'https://mycloud.com/v3'
project_name: 'admin'
domain_id: 'default'
interface: public

import shade
# Connect to the first cloud
site_a = shade.operator_cloud(cloud='mycloud')
# Connect to the second cloud
site_b = shade.operator_cloud(cloud=rackspace)
get_servers=lambda site: {x.name: x
for x in site.list_servers() }
# List virtual servers
servers_a = get_servers(site_a)
servers_b = get_servers(site_b)
# Check servers not highly available.
set(servers_a.keys()) - set(servers_b.keys())
# Get doppelgangers ip ;)
for k,v in servers_a.items():
v.doppelganger_ip = servers_b[k].interface_ip
# Add many clouds to your config
#
# Known cloud providers can be
# referenced via eg. "profile: rackspace"
identity_api_version: '3.0'
clouds:
mycloud:
...
rackspace:
auth:
cloud: rackspace
project_id: 275610
username: openstack
password: xyzpdq!lazydog
region_name: DFW,ORD,IAD
interface: internal

# Use logging for troubleshooting issues, eventually dumping to a file.
import shade
import logging
logging.basicConfig(level=logging.DEBUG, filename="client.log")
shade.simple_logging(debug=100, http_debug=0)
# Trace python-requests logging here or with http_debug=1
import httplib as http_client
http_client.HTTPConnection.debuglevel = 1
requests_log=logging.getLogger("requests.packages.urllib3")
requests_log.setLevel(logging.DEBUG)
requests_log.propagate = True
# Cloud configs are read with os-client-config
mycloud = shade.operator_cloud(cloud='mycloud')

Some more Shade - 1/2
# You can access every entry as an
object
server.image
{u'id': u'42c26768-0edf-...'}
# or as a dict (eg for complex attrsa)
server['OS-SRV-USG:launched_at']
u'2017-02-18T09:42:19.000000'
# Shade retains custom properties
# in the properties attribute
server.properties # a dict!
{
'OS-EXT-AZ:availability_zone':
u'nova',
'OS-EXT-STS:power_state': 1,
...
}
# The model.rst document,
# describes shade entries
Server = dict(
location=Location(),
id=str(),
name=str(),
image=dict() or str(),
flavor=dict(),
volumes=list(), # Volume
interface_ip=str(),
...
)

Some more Shade - 2/2
# Pandas creates wonderful reports
import pandas as pd
# Get your data
servers_a = cloud_a.list_servers()
servers_b = cloud_c.list_servers()
all_servers = servers_a + servers_b
# dump them into a pandas.DataFrame
df = pd.DataFrame(data=all_servers)
report = df[['name', 'status',
'cloud', 'interface_ip']]
# then to Excel
writer = ExcelWriter('cloud.xlsx')
report.to_excel(writer,'servers')
writer.save()
#
# The report outcome!
#
name status cloud interface_ip
0 os3-node1 ACTIVE mycloud
10.168.176.124
1 os3-master0 ACTIVE mycloud
10.168.176.125
...
8 os3-master0 ACTIVE rackspace
172.23.176.125
9 os3-master2 ACTIVE rackspace
172.23.176.123

Shade and Ansible
Provisioning projects on a
multi-datacenter OpenStack
Create projects with Heat:
- the project remains as a stack
- accidental removal risk
- can't manage quotas & co with HOT*
- update a quota may trigger a full
stack-update
"Projects represent
the base unit of
ownership in
OpenStack, in that all
resources in OpenStack
should be owned by a
specific project."
docs.openstack.org

Implement an Ansible module with Shade
Many policies for creating and updating
OpenStack objects
Testing/checking differences between
expected and actual settings
Trigger custom actions depending on
os_project status
# developing the os_project_access module
# grant access on various resources to
# a given project
- name: ansible provisioning a project
os_project_access:
...
resource_type: nova_flavor
resource_name: x1.medium
project_id: 0000-ffff-0000
- name: with iterations
os_project_access:
resource_type: cinder_volume_type
state: "{{item.state}}"
resource_name: "{{item.name}}"
project_id: 000-fff-000
with_items:
- {name: "sas", state: present }
- {name: "ssd", state: absent }

Writing Ansible Modules
PLEASE, DO NOT
write complex modules
accessing directly {nova,cinder,..}-api

Writing Ansible modules
# Implement a general workflow!
resource = _get_resource(resource_name_or_id)
allowed_projects = _get_resource_access(resource_id) # get acls
if state == 'present':
_add_resource_access(resource_id, target_project_id)
elif state == 'absent' and target_project_id in allowed_projects:
_remove_resource_access(resource_id, target_project_id)

Contributing to Shade
# Implement missing methods in shade
if resource_type == 'nova_flavor':
_get_resource = cloud.get_flavor
_list_resource_access = cloud.list_flavor_access # <- write this!
_add_resource_access = cloud.add_flavor_access # <- write this!
_remove_resource_access = cloud.remove_flavor_access # <- write this!
...
elif resource_type == 'your_resource_type':
def __not_general_enough_to_be_in_shade(resource_id): # <- if patch refused ;)
...
_list_resource_access = __not_general_enough_to_be_in_shade
else: raise NotImplementedError("Resource %r not implemented" % resource_type)

⬝ Join #openstack-shade on irc
⬝ Sign Up https://launchpad.net/openstack
⬝ Install git-review
⬝ Write functional tests
⬝ Install devstack and test there before submitting for review

Check your progresses and give feedback on other's patches
https://review.openstack.org/#/q/is:watched is:open

The Hardest Part
Where is my git-review password? No, it's not your account one ;)
1
2

Questions?
Thank You
roberto.polli@par-tec.it

Just one-shade-of-openstack

More Related Content

What's hot (13)

Similar to Just one-shade-of-openstack (20)

More from Roberto Polli (20)

Recently uploaded (20)

Just one-shade-of-openstack