Instances Lose External Connectivity After Detaching Floating IP
Problem
After removing a floating IP from an instance, external connectivity ceases.
Environment
- Platform9 Managed OpenStack - v4.5 - v5.2
- Neutron
Cause
The neutron-l3-agent service fails to update the routing rule corresponding to the instance's fixed IP within the router namespace on the L3 agent node(s) responsible for hosting the virtual router (in a DVR configuration, this would be the instance hypervisor). Subsequently, this causes a loss of external connectivity for the instance.
Resolution
- Identify the fixed IP, network, and host associated with the affected instance.
Screen_Shot_2020-01-02_at_3.38.20_PM.png
- Identify the router associated with the network.
Screen_Shot_2020-01-02_at_3.39.56_PM.png
- Record the UUID of the network and router identified in the previous two steps.
Note: In the UI, this may be done by navigating to the "Networks" tab, selecting either Networks or Routers and then subsequently clicking on the name of the resource. An OpenStack UUID column will be visible in the resource overview.
- In the
qrouternamespace (corresponding to the UUID of the router) on the instance's host, inspect the routing rules.
Note: In the following examples, a unique router UUID will be used which will differ from your own. Ensure you replace the UUID when executing the commands yourself.
- Inspect the route table (of which the ID may be identified based on the prior step) corresponding to the instance's fixed IP.
The default route with the rfp interface in the rule indicates all instance traffic is still going to the floating IP namespace.
- Delete the stale routing rule.