Openstack Rally - Benchmark as a Service. Openstack Meetup India. Ananth/Rahul.

Benchmark as a Service
Openstack- Rally
Openstack Meetup Bangalore
01/03/2014

C B Ananth – Cisco Systems – cbpadman@cisco.com
Rahul Upadhyaya – Cisco Systems – rahuupad@cisco.com

Agenda
•
•
•
•
•
•

Introduction
Components
Configuration and Scenarios
Running benchmarking tests
Demo
Q&A

Rally - Why ?
"How does OpenStack work at scale?”

Rally vs/& Tempest
Rally Versus Tempest
Both have different Objectives
Rally checks behavior at scale and stress.
Tempest just checks functional sanity.

Rally + Tempest
However Rally can invoke tempest internally
and check functional sanity. (work in progress)

Rally Components
• Server Providers - provide servers (virtual servers), with ssh access, in one L3
network.
• Deploy Engines - deploy OpenStack cloud on servers that are presented by
Server Providers
• Verification - component that runs tempest (or another specific set of tests)
against a deployed cloud, collects results & presents them in human readable
form.
• Benchmark engine - allows to write parameterized benchmark scenarios &
run them against the cloud.

Installing Rally on Ubuntu.
Note : We would recommend using a separate machine for Rally. If the machine
that you use to run rally has some Openstack components running, we suggest
creating a virtual environment for running Rally because it may have conflicts with
version of the client python libraries.

Prerequisite
•
•
•

sudo apt-get update
sudo apt-get install libpq-dev git-core python-dev libevent-dev libssl-dev python-pip libffi-dev
sudo pip install pbr

Installing Rally on Ubuntu.
Installing Rally
Clone
• git clone https://github.com/stackforge/rally.git && cd rally
• python setup.py install
Configure
• sudo mkdir /etc/rally
• sudo cp etc/rally/rally.conf.sample /etc/rally/rally.conf
• sudo vim /etc/rally/rally.conf
• # Change the "connection" parameter,
• # e.g. to connection="sqlite://///home/<your_username>/.venv/rally-db/$sqlite_db" (or any other place)
Create Database
• rally-manage db recreate

Rally : Deploy Engines.
The task of a deploy engine is to control the process of deploying some OpenStack
distribution like DevStack or FUEL before any benchmarking procedures take place.
Every deploy engine should implement the following fairly simple interface:
• constuctor, which takes a deployment entity as its only parameter;
• deploy(), which should deploy the appropriate OpenStack distribution given the
cloud config from the deployment object the engine was initialized with (possibly
using one of available server providers)
• cleanup(), which should clean up the OpenStack deployment (again, possibly using
one of available server providers).

Rally. Deploy Engine . Example
If you already have a existing Openstack Deployment that you want to benchmark
Use a json file that looks like one below. (with value specific to your system):
{
"name": "DummyEngine",
"endpoint": {
"auth_url": "http://192.168.122.22:5000/v2.0/",
"username": "admin",
"password": "password",
"tenant_name": "admin"
}
}

For Devstack based deployment
{
"name": "DevstackEngine",
"localrc": {
"ADMIN_PASSWORD": "secret",
"NOVA_REPO": "git://example.com/nova/",
...
},
"devstack_repo": "git://example.com/devstack/",
"provider": {
"name": "${PROVIDER_NAME}",
...
}
}

Rally – Server Providers
Server providers in Rally are typically used by deploy engines to manage virtual
machines necessary for OpenStack deployment and its following benchmarking.

The key feature of server providers is that they provide a unified interface for
interacting with different virtualization technologies (LXS, Virsh etc.) and cloud
suppliers (like Amazon).
Every server provider should implement the following basic interface:
•
•
•

constructor, which should take the deployment entity the provider should bind to and a
config dictionary as its parameters;
create_servers(image_uuid, type_id, amount), which should create the requested number
of virtual machines of the given type using a specific image. The method should also
return the list of created servers wrapped in special Server entities.
destroy_servers(), which should destroy all virtual machines previously created by the
same server provider.

Rally – Server Providers –Ex
DummyProvider – If you already have an Openstack Deployment.
This provider does nothing, but returns endpoints from configuration. This may
be useful if you have specific software/hardware configuration ready to deploy
OpenStack.
{
"name": "ExampleEngine",
"provider": {
"name": "DummyProvider",
"credentials": [{"user": "root", "host": "host1.net"}, {"user": "root", "host": "host2.net"}]
}

}

Rally – How_to_run
Simple 
1. Initialize your Deployment
2. Create a json for Supported Benchmarking scenario
3. Run Benchmarking against a deployment above

Rally – Initialize Deployment
1. Create a Deployment configuration (json) file.
If you are running Rally against a existing Openstack Deployment your should look like
{
"name": "DummyEngine",
"endpoint": {
"auth_url": <KEYSTONE_AUTH_URL>,
"username": <ADMIN_USER_NAME>,
"password": <ADMIN_PASSWORD>,
"tenant_name": <ADMIN_TENANT>
}
}

2. Create a deployment using deployment create command
$ rally deployment create --filename=dummy_deployment.json --name=dummy

3. If you want to list deployments
$ rally deployment list
4. Switch to a different Deployment
$ rally use deployment --deploy-id=<Another deployment UUID>

Rally – Set Benchmark scenario.
Some sample configurations are provided at “rally/doc/samples/tasks/”.
Lets pick up a scenario boot-and-delete.json from Nova. It looks like
{
"NovaServers.boot_and_delete_server": [
{
"args": {"flavor_id": 1,
"image_id": "73257560-c59b-4275-a1ec-ab140e5b9979"},
"execution": "continuous",
"config": {"times": 10, "active_users": 2, "tenants": 3,
"users_per_tenant": 2}
}
]
}

Modify this to design your test-case. Similarly for all other cases other available json can be modified or
you can even write a new one for a custom case.
Rally by default uses the last created deployment. Use the switch deployment commands to run the
tests with a different deployment.

Lets dig deeper
•
•
•
•
•
•

Test name : “NovaServers.boot_and_delete_server”
Execution : either continuous / periodic
Times : Number of times the test needs to be run
Active_users : Number of parallel threads (concurrent users)
Tenants : Total number of tenants to be created
Users_per_tenant : Number of users per single tenant

Other parameters to be used only with supported tests
• Network : Name of network to be used
• Script : If a script is passed as input to the test
• Actions : soft_reboot / stop_start

Rally – Run Benchmark.
Run your benchmark scenario by pointing at the json you created in the previous step
$ rally --verbose task start --task=my-task.json
You can check the state of the task
$ rally task list
To check a complete task analysis
$ rally task detailed <Task UUID>

Rally Result –Example.
$ rally task detailed <Task UUID>
================================================================================
Task <Task UUID> is finished. Failed: False
-------------------------------------------------------------------------------test scenario NovaServers.boot_and_delete_server
args position 0
args values:
{u'args': {u'flavor_id': <Flavor UUID>,
u'image_id': u'<Image UUID>'},
u'config': {u'active_users': 1, u'times': 2}}
+--------------------+---------------+---------------+---------------+
|
action
| max (sec) | avg (sec) | min (sec) |
+--------------------+---------------+---------------+---------------+
| nova.boot_server | 9.22798299789 | 8.90022659302 | 8.57247018814 |
| nova.delete_server | 4.24928498268 | 3.26377093792 | 2.27825689316 |
+--------------------+---------------+---------------+---------------+
+---------------+---------------+---------------+---------------+-------------+
| max (sec)
| avg (sec)
| min (sec)
| success/total | total times |
+---------------+---------------+---------------+---------------+-------------+
| 13.4775559902 | 12.1641695499 | 10.8507831097 | 1.0
| 2
|
+---------------+---------------+---------------+---------------+-------------+
--------------------------------------------------------------------------------

Supported Benchmarks (as of 01/03/14)
Basic Use-Cases of
 Authentication
 Nova
 Glance
 Keystone
 Cinder

Disclaimer

Rally is currently undergoing heavy development,some things may vary with time.
Make sure to go through to check the wiki when you start at
https://wiki.openstack.org/wiki/Rally
However, “Keep Calm” as basics would be the same.

Benchmarking scenarios
• Glance
– Create-and-delete-image
– Create-image-and-boot-instance
• Cinder
– Create-and-delete-volume
• Keystone
– Create-and-delete-user
– Create-user

Benchmarking scenarios
• Authenticate
– Authenticate-keystone
• Nova
– Boot-and-delete-periodic
– Boot-and-delete
– Boot-runcommand-delete
– Boot-server-from-volume-and-delete
– Soft-reboot
Note: Other scenarios can be suggested to the community under the
section, “Benchmarking scenario wishlist”

Cisco Openstack Installer(COI)
If you are completely new to Openstack, you would love a easy and reliable way to
deploy Openstack before running Rally. Out of the many solutions freely available, you
may choose Cisco Openstack Installer (COI) to Install Openstack. COI is puppet based
open-source and freely available installer for Openstack. You can setup a Openstack
cloud in less than 10 commands which are fairly simple and listed below.
Below is a step-by-step guide which is an adaptation from the wiki to install an All-inOne Node for Openstack – Havana version.
Steps :
1. Install Ubuntu 12.04 server on a VM/physical machine.
2. Set the hostname so that you have a proper FQDN (for puppet certificates)
3. Sudo –su
4. apt-get install -y git
5. cd /root && git clone -b havana
https://github.com/CiscoSystems/puppet_openstack_builder && cd
puppet_openstack_builder && git checkout h.1
6. export vendor=cisco
7. export scenario=all_in_one

Cisco Openstack Installer(COI)
( Conditional ) If you are using any eth other than eth0 as a primary interface then edit
the file “~/puppet_openstack_builder/install-scripts/install.sh” and replace eth0 with
desired eth
8. (Fix to a Bug) edit “~/puppet_openstack_builder/manifests/setup.pp” and
completely replace the line 52 with this string -- host_aliases =>
"'$::hostname'.${::build_server_domain_name}”

9. Once done with above run “install.sh” in “~/puppet_openstack_builder/” folder and
wait for 20-30 mins till puppet configures openstack as a All-in-One setup.
10. You can log-onto horizon using password (admin/Cisco123)

Contribute to Rally













Discussions & RoadMap
https://etherpad.openstack.org/p/Rally_Main
Open and assigned tasks
https://trello.com/b/DoD8aeZy/rally
To get account ping Boris in IRC (boris-42) or email me (boris(at)pavlovic.me)
IRC chat
server: freenode.net
chanel: #openstack-rally
Weekly Meetings
The Rally project team holds weekly meetings on Tuesdays at 1700 UTC in IRC, at the #openstack-meeting
channel.
Source
https://github.com/stackforge/rally
Project space
http://launchpad.net/rally
Blueprints
active: http://blueprints.launchpad.net/rally
v1 base: https://blueprints.launchpad.net/rally/+spec/init
Bugs
https://bugs.launchpad.net/rally

References
• https://wiki.openstack.org/wiki/Rally
• http://www.slideshare.net/mirantis/rally-benchmarkingatscale
• http://thoughtsoncloud.com/2013/12/benchmarking-and-scalingopenstack-in-the-enterprise/
• http://docwiki.cisco.com/wiki/Openstack:Havana-Openstack-Installer

Openstack Rally - Benchmark as a Service. Openstack Meetup India. Ananth/Rahul.

More Related Content

Openstack Rally - Benchmark as a Service. Openstack Meetup India. Ananth/Rahul.