Jump to: navigation, search

Difference between revisions of "Neutron/LBaaS/HowToRun"

< Neutron‎ | LBaaS
(Ubuntu Packages Setup)
m (Validation)
Line 89: Line 89:
 
== Validation ==  
 
== Validation ==  
  
We now have two hosts with a load balancer pointed at them, but those hosts are serving up any HTTP content.  
+
We now have two hosts with a load balancer pointed at them, but those hosts are not serving up any HTTP content.  
  
 
A simple trick is to use netcat on the hosts to implement a simple webserver.  For example, run:  
 
A simple trick is to use netcat on the hosts to implement a simple webserver.  For example, run:  

Revision as of 16:39, 15 March 2014

Getting the Code

LBaaS introduces changes in the following modules: (currently all changes are in master branch)

  • neutron
  • python-neutronclient
  • horizon
  • devstack

Devstack Setup

Add the following lines to your localrc:

enable_service q-lbaas

Then run stack.sh

After stack.sh completes you'll be able to manage your Load Balancer via the CLI tools and within Horizon

Ubuntu Packages Setup

Install the neutron-lbaas-agent and haproxy:

sudo apt-get install neutron-lbaas-agent haproxy

Once packages are installed make the directory for the LBaaS service and copy the configuration into place:

sudo mkdir -p /etc/neutron/plugins/services/agent_loadbalancer/
sudo cp /etc/neutron/lbaas_agent.ini /etc/neutron/plugins/services/agent_loadbalancer/lbaas_agent.ini

And edit the service_plugins in [DEFAULT] section in neutron.conf to enable the service:

sudo sed -i.bak "s/\#\ service_plugins\ \=/service_plugins = neutron.plugins.services.agent_loadbalancer.plugin.LoadBalancerPlugin/g" /etc/neutron/neutron.conf
Finally enable the Load Balancer section in Horizon by editing
/etc/openstack-dashboard/local_settings.py
and changing:
OPENSTACK_NEUTRON_NETWORK = {
    'enable_lb': False
}

to

OPENSTACK_NEUTRON_NETWORK = {
    'enable_lb': True
}

Once done restart your Neutron services and Apache to start using.

Topology Setup

Spin up three VMs, two to be servers, and one to be a client.

nova boot --image <image-uuid> --flavor 1 server1
nova boot --image <image-uuid> --flavor 1 server2
nova boot --image <image-uuid> --flavor 1 client

Get the UUID of the private subnet.

neutron subnet-list

Create a Pool:

neutron lb-pool-create --lb-method ROUND_ROBIN --name mypool --protocol HTTP --subnet-id <subnet-id> 

Create Members (using the IPs of server1 and server2):


nova list 

neutron lb-member-create --address <server1-ip> --protocol-port 80 mypool
neutron lb-member-create --address <server2-ip> --protocol-port 80 mypool

Create a Healthmonitor and associated it with the pool:

neutron lb-healthmonitor-create --delay 3 --type HTTP --max-retries 3 --timeout 3
neutron lb-healthmonitor-associate <healthmonitor-uuid> mypool

Create a VIP

neutron lb-vip-create --name myvip --protocol-port 80 --protocol HTTP --subnet-id <subnet-id> mypool

note the address for use below.

Validation

We now have two hosts with a load balancer pointed at them, but those hosts are not serving up any HTTP content.

A simple trick is to use netcat on the hosts to implement a simple webserver. For example, run:

while true; do echo -e 'HTTP/1.0 200 OK\r\n\r\n<servername>' | sudo nc -l -p 80 ; done 

replacing <servername> with "server1" and "server2" as appropriate. Once the server is started, you'll see incoming HTTP GET requests.. that's the load balancer health check in action!

If you have python installed, you can also create an index.html with the text "server1" or "server2" then in the same directory run

sudo python -m SimpleHTTPServer 80

Finally, to test real load balancing, from your client, use wget to make sure your requests are load-balanced across server1 and server2 as expected.

wget -O - http://<server1-ip> 
wget -O - http://<server2-ip> 

Then use wget to hit the VIP IP several times in succession. You should bounce between seeing server1 and server2.

wget -O - http://<vip-ip> 
wget -O - http://<vip-ip> 
wget -O - http://<vip-ip> 
wget -O - http://<vip-ip> 

Full list of LBaaS CLI commands is available at Quantum/LBaaS/CLI

Troubleshooting

LBaas is implemented similar to L3 + DHCP using namespaces. You can use "ip netns list" to find the namespace named qlbaas-<pool_id>, and then test connectivity from that namespace.

Use "screen -x stack" to view the q-svc and q-lbaas tabs for errors.

Grep syslog for "haproxy" to see messages from Haproxy (though they are quite cryptic!)