Long Quan Sha and Ethan Lynn from IBM and Tian Hua Huang from Huawei presented on best practices for Heat resource modules and deployment patterns. They discussed Heat introduction, software deployment options using cloud-init and software deployments, building custom images, and signal transport methods. They also covered creating resource modules based on business concepts to make templates easier to understand and compose common deployment patterns. Finally, they demonstrated resource modules and a load balancing autoscaling group template.

1. Best practices for heat resource
modules and patterns
Long Quan Sha, software engineer at IBM
Ethan Lynn, software engineer at IBM
Tian Hua Huang, software engineer at HuaWei

2. Agenda
• Heat introduction
• Software deployment introduction
• Hands-on 1
• Heat resource module/deployment pattern
• Demo
• Hands-on 2
• Best practices
• Q&A

3. Heat Introduction

4. Heat
• Heat’s mission is to orchestrate composite cloud applications
using a declarative template format through an OpenStack-
native REST API.
Heat
Template
Heat
Nova
Cinder
Neutron
Swift
Senlin

5. Heat Template
heat_template_version: xxxx-xx-xx
parameters:
…
resources:
...
outputs:
...
Template Sections
$ heat stack-create <stack_name>
$ -f your.template
$ -P ’Image=fedora;flavor=m1.large’
Create Stack
http://docs.openstack.org/developer/heat/template_guide/hot_spec.html`

6. Software Deployment Introduction

7. Software Configuration
• Two ways to run software configuration scripts in VM instances:
• User-data boot scripts and cloud-init
• Software deployment resources
Pros Cons
Cloud-init • Most images have cloud-
init built in.
• Only run once after
instance first boots.
• Heat cannot be notified
when the scripts finish.
Software Deployments • Triggered by stack
create/update and
managed by heat.
• Send a signal back to heat
when finished.
• The sequence of multiple
scripts can be controlled.
• Require os-*-config
services running in VM
instances.

8. Sample Template of cloud-init
cloud-init Configuration
Heat
Nova
Server
Nova Metadata
Communicate
Create Pull Scripts
Query Scripts

9. Sample Template of Software Deployment
Heat
Nova
Server
Create
Communicate
Pull Scripts
(via 3 ways)
Software
Deployment
http://docs.openstack.org/developer/heat/template_guide/openstack.html#OS::Heat::SoftwareConfig

10. Build Image with Software Config Tools
$ git clone https://git.openstack.org/openstack/diskimage-builder.git
$ git clone https://git.openstack.org/openstack/tripleo-image-elements.git
$ git clone https://git.openstack.org/openstack/heat-templates.git
$ export ELEMENTS_PATH=tripleo-image-elements/elements:heat-
templates/hot/software-config/elements
$ disk-image-create vm fedora
$ os-collect-config os-refresh-config os-apply-config
$ heat-config heat-config-script
$ -o fedora-software-config.qcow2
Build Custom Image
Full Guide:
http://docs.openstack.org/developer/heat/template_guide/software_deployment.html#custom-image-script

11. Signal Transport
Three ways for server to talk with heat:
• URL
• Zaqar Messaging Service
• Swift Object Storage Service
Using URL:
signal_transport: HEAT_SIGNAL
software_config_transport: POLL_SERVER_HEAT
Using Zaqar:
signal_transport: ZAQAR_SIGNAL
software_config_transport: ZAQAR_MESSAGE
Using Swift:
signal_transport: TEMP_URL_SIGNAL
software_config_transport: POLL_TEMP_URL

12. Hands on
• Openstack env
• IP: 159.122.194.136
• SSH password: demo/passw0rd20116
• Horizon password: admin/admin
• Templates
• Link: https://github.com/lynic/templates/tree/master/austin
• Location: /home/demo/hands-on/softwareconfig/
• example_cloudinit.yaml: sample template of cloudinit
• example_software_deployment.yaml: sample template of software deployment
resources
• Create stack
• cd hands-on/softwareconfig/
• source openrc.demo
• heat stack-create group1 –f example_cloudinit.yaml

13. Resource Modules and Deployment Patterns

14. Heat Template Example
• Create a VM:
1.OS::Neutron::SecurityGroup
2.OS::Neutron::Port
3.OS::Nova::KeyPair
4.OS::Nova::Server

15. Heat Resources – Pain Points
• Based on OpenStack components rather than business
concepts
• Too low-level to use and understand
• Too many properties of each resource
• Too much time invested learning resources before creating a
template
• Difficult to deploy a large scale application in production

16. Resource Module
• Based on regular Heat resources, abstract business concepts
• Easy to understand, without having to know underlying
technologies of OpenStack
• Extracted from real world deployment experiences
• Easy to adopt deployment patterns for typical scenarios
• Easy to compose a template

17. Resource module
• https://github.com/shalq/summit-demo/tree/master/resource-
module/heat

18. Resource Module
resource_registry:
"IBM::Alchemy::Member": member.yaml
"IBM::Alchemy::Group": group.yaml
"IBM::Alchemy::LBMember": lb_member.yaml
"IBM::Alchemy::LBGroup": lb_group.yaml
"IBM::Alchemy::HAGroup": haproxy_group.yaml
"IBM::Alchemy::Manager": manager.yaml
"IBM::Alchemy::Keypair": keypair.yaml
"IBM::Alchemy::Network": network.yaml

19. Resource Module - Example
Create a netIBM::Alchemy::Network
• Create a net
• Create a router
• Add net to router
• Set gateway for router

20. Deployment Patterns
• "IBM::Alchemy::Manager"
• Topology information about the stack
• Deployment scripts to setup other nodes
• Deployment tools: Salt, Ansible, etc.
salt-master
salt-minion
salt-minion
salt-minion
manager
Cluster-A
Cluster-A
Cluster-A
Cluster-ACluster-A
Cluster-B
Cluster-A
Cluster-A
Cluster-C

21. Autoscaling Group - Cluster
• Autoscalling group:
• "IBM::Alchemy::Member": member.yaml
• "IBM::Alchemy::Group": group.yaml
• Autosaclling group with load balance
• "IBM::Alchemy::LBMember": lb_member.yaml
• "IBM::Alchemy::LBGroup": lb_group.yaml
• HA group
• "IBM::Alchemy::HAGroup": haproxy_group.yaml

22. IBM::Alchemy::Group
IBM::Alchemy::Group IBM::Alchemy::Member

23. IBM::Alchemy::LBGroup
IBM::Alchemy::LBGroup
vip Pool member
Pool member
Pool member
OS::Neutron::Pool
• Scale up
• Scale down
• Neutron pool
• Load balance
client

24. IBM::Alchemy::HAGroup
vip
Front-end
(IBM::Alchemy::HAGroup)
Back-end
IBM::Alchemy::Group
haproxykeepalived
keepalived haproxy
primary node
secondary node
cluster
cluster
cluster
client

25. Demo for Resource Module
IBM::Alchemy::LBGroup
• Deploy salt on manager node and member nodes
• Deploy docker on member nodes
• Setup web server on member nodes
• Auto scaling group for a web server
• Scale up
• Scale down
• Load balance
IBM::Alchemy::Manager
manager
Salt-master Salt-minion web
vip client
deploy
request

26. Demo Stack

27. Hands-on

28. Hands-on
• Prerequisites
• OpenStack (devstack) env
• An image with os-collect-config installed is available in glance
• Create a flavor named “demo”, 2G ram, 6G disk
• Download templates from /home/demo/hands-on/heat/
• Specify network parameters when creating a stack
• OpenStack env
• ssh demo@159.122.194.136 password:passw0rd2016
• Templates
• Location: /home/demo/hands-on/heat/
• test-lb.yaml: create a cluster for webapp with loadbalance
• env.yaml: define resource modules

29. Hands-on Steps
• 1. Create a stack
• cd /home/demo/hands-on/ heat/
• . openrc.demo ;
• heat stack-create –e env.yaml –f test-lb.yaml <stack-name>
• 2. Check stack
• check resource-list of stack : heat resource-list <stack-name>
• check status of VM in the stack
nova list | grep <stack-name>
nova console-log <member-vm>
• 3. Logon the vm
• Get private key : echo -ne `heat output-show <stack-name> key 2>/dev/null `|tr -d '"‘
• Save the private key into a temprory file such as /tmp/group1.key
• Change its permission to 600 by “chmod 600 /tmp/group1.key”
• Logon the vm: ssh –i /tmp/group1.key fedora@<vm-ip>
• You can check stack progress by “heat resource-list <stack>” or nova console <server>

30. Hands-on Steps
• 4. Get cluster information when stack is in
CREATE_COMPLETE
• Get floatingip of vip: heat output-show <stack-name> vip_floatingip
• Check group size of the stack : heat outputshow <stack-name> asg_size
• Check member list of the group : heat outputshow <stack-name> member_list
• 5. Test webapp by vip when stack is in CREATE_COMPLETE
• Get floatingip of the vip : heat output-show <stack-name> vip_floatingip
• Test floatingip of the vip : curl <floatingip of vip> , it will output server’s hostname and current time
• Test vip: logon the vm, and run “curl <vip>”
• 6. Add a new member into cluster <optional>
• Get scale up url: heat output-show <stack-name> scale_up_url
• Scale up: curl –XPOST –i <scale_up_url>
• Test vip : curl <vip> for several times, the request will be redirected to member1 and member2

31. Tips for Heat
• Resource group status not reflected on stack-level, check group change by
“heat resource-list”
• Use --timeout if stack deployment is expected significantly time consuming
• Use template-validate to check your template with parameters or stack-
preview to preview
whether the stack can be created
• Use –dry-run of stack-update to see what would be changed
• Operations limited within project scope, stack-list provides –global-tenant
to get stacks from all projects
• Environment is powerful : stack-create/update –e
• If you don’t want to replace your resources when some parameters change,
set immutable to True for the parameter

32. Q&A

33. Thank You