Skip to content

Exercise 5. Restarting the Open vSwitch agent without cutting the network

Goal: restart the Neutron Open vSwitch agent and see that running instances keep their network, while new instances wait for it. Time: about 10 minutes. Runbook: I need to restart the Open vSwitch agent on a compute. Its instances keep their network

Before you start

Read before every exercise. Create an instance with a floating IP that answers ping, as in exercise 4, and call it ex5-test.

1. Do it

In a second SSH session, leave a ping running:

ping -i 0.5 <floating-ip>

Then, in the first session:

⚠️ State-changing. Restarts the Open vSwitch agent. Running instances keep their network; ports created or changed during the restart wait for the agent. Disposable lab only.

sudo systemctl restart --no-block kolla-neutron_openvswitch_agent-container.service

--no-block gives the prompt back at once while the restart goes on, about a minute. Without it, systemctl waits for the end of the restart, and the next command would come too late to see anything.

Right away, while the agent restarts, create a second instance and time how long it takes to become ACTIVE:

time openstack server create --image cirros --flavor m1.tiny --network demo-net --key-name mykey --wait ex5-second

2. What you should see

  • The ping keeps answering. On the runbook's lab, no request was lost.
  • The restart of the unit takes about a minute. During it, sudo systemctl status kolla-neutron_openvswitch_agent-container.service shows deactivating (stop), then active (running) again (runbook's lab: about 70 s; 62 s on a later lab).
  • The second instance waits for the agent. In a later lab review of the runbook, an instance requested during the restart became ACTIVE after about 80 s instead of 18 s, and the agent was reported not alive for a few seconds at the end of the restart.

Before reading on: if the agent restarts, what keeps forwarding the packets?

3. Read the runbook

Read its Diagnosis and Validity and limits: what the agent does, what ovs-vswitchd does, and the setting (drop_flows_on_start) that would make a restart cut the traffic.

4. Check

Follow the runbook's Verification: the agent is alive, the ping still answers, the flows are present.

5. Clean up

Stop the ping (Ctrl-C), then:

openstack server delete --wait ex5-test ex5-second
openstack floating ip delete <floating-ip>

Done with the lab for today? Delete everything: a running lab keeps costing.