AZ aware router placement

[matelakat]

We know which network nodes live close to which AZs. We also know which routers belong to which AZs, We want to use this information to move routers to network nodes that are close to compute nodes.

neutron-ha-tool can:

• evacuate an agent (--l3-agent-evacuate)
• move a list of routers (--router-list-file)
• move to a specific agent (--target-agent-id)

The idea is to re-use neutron-ha-tool for this job.

QnA

• How do we know which routers belong to which AZs?
• How do we know which agents (network nodes) belong to which AZs?

[matelakat]

Having n network nodes, something like this would work:

Some pseudo code:

for source_agent in agents:
    source_routers = get_routers_on_agent(source_agent)
    for target_agent in agents:
        target_routers = get_routers_on_agent(target_agent)
        routers_to_move = source_routers && target_routers
        move_routers(routers_to_move, source_agent, target_agent)

[tpatzig]

for the agents i'd propose to set the description to the name of the AZ, e.g: neutron agent-update --description "rot_2" <agent>

To find out which router has instances in which AZ, we have to check the networks, having an interface on that router: neutron port-list --network-id <network_id> -c device_owner -f value | grep compute results an entry per instance port compute:AZ, like: compute:rot_2

If we find routers, where we have networks attached with instances in multiple AZs, the script should output a warning and do nothing. A dry-run, what the script will do is required.

[matelakat]

@tpatzig what shall be the value for ? I guess we need to iterate through all the networks?

[tpatzig]

This should give us all the information:

for router in `neutron router-list -c id -f value`; do
    echo "--------------------------------------------"
    echo "router: $router"
    for network in `neutron router-port-list $router --device_owner network:router_interface -c network_id -f value`; do
        echo "network: $network"
        neutron port-list --network-id $network -c device_owner -f value | grep compute | uniq -c
    done
done

[matelakat]

The first iteration if the tool is ready. It will print out some log messages stating what should be done. 2 scenarios are not clear:

• What should we do if a specific az has no associated L3 agent
• What should we do if multiple L3 agents could be serving the router (let's say the router serves compute nodes in different azs)
• How shall we keep the load of L3 agents in balance

To run the first iteration:

git clone https://github.com/matelakat/cookbook-openstack-network
cd cookbook-openstack-network/
git fetch origin router-optimizer-0 && git checkout FETCH_HEAD && python files/default/router-optimizer.py 2>&1 | tee -a router-optimizer.log

The code for the first iteration is here:

https://github.com/matelakat/cookbook-openstack-network/tree/router-optimizer-0