Assigning IP addresses based on topology
About IP address assignment
Calico can be configured to use specific IP pools for different topological areas. For example, you may want workloads in a particular rack, zone, or region to receive addresses from the same IP pool. This may be desirable either to reduce the number of routes required in the network or to meet requirements imposed by an external firewall device or policy.
There are three approaches to configuring IP address assignment behavior which the IPAM section of the cni-plugin configuration reference document explains in detail. For the purposes of topology, IP address assignment must be per-host (node) which disqualifies Kubernetes annotations as an option since it is only configurable on a per-namespace or per-pod level. Left between using CNI configuration and IP pool node selectors, the latter is preferred as it does not require making any changes within the host’s file system which the former does.
At a high level, node selection-based IP address assignment is exactly what it sounds like: node labels are set and then the appropriate node selectors on the desired IP pool resources are set. The remainder of this article goes into a detailed example of using this feature to configure IP address assignment based on a certain rack affinity.
Important: If Calico is unable to determine an IP pool for a workload based on the above order, or if there are no IP addresses left in the determined IP pools, then the workload will not be assigned an address and will fail to start. To prevent this, we recommend ensuring that all nodes are selected by at least one IP pool.
This feature requires Calico for networking in etcd mode.
In this example, we created a cluster with four nodes across two racks (two nodes/rack). Consider the following:
Using the pod IP range
192.168.0.0/16, we target the following setup: reserve
192.168.1.0/24 pools for
By installing Calico without setting the default IP pool to match,
calicoctl get ippool -o wide shows that Calico created its
default IP pool of
Delete the default IP pool.
default-ipv4-ippoolIP pool resource already exists and accounts for the entire
/16block, we will have to delete this first:
calicoctl delete ippools default-ipv4-ippool
Label the nodes.
To assign IP pools to specific nodes, these nodes must be labelled using kubectl label.
kubectl label nodes kube-node-0 rack=0 kubectl label nodes kube-node-1 rack=0 kubectl label nodes kube-node-2 rack=1 kubectl label nodes kube-node-3 rack=1
Create an IP pool for each rack.
calicoctl create -f -<<EOF apiVersion: projectcalico.org/v3 kind: IPPool metadata: name: rack-0-ippool spec: cidr: 192.168.0.0/24 ipipMode: Always natOutgoing: true nodeSelector: rack == "0" EOF
calicoctl create -f -<<EOF apiVersion: projectcalico.org/v3 kind: IPPool metadata: name: rack-1-ippool spec: cidr: 192.168.1.0/24 ipipMode: Always natOutgoing: true nodeSelector: rack == "1" EOF
We should now have two enabled IP pools, which we can see when running
calicoctl get ippool -o wide:
Verify that the IP pool node selectors are being respected.
We will create an nginx deployment with five replicas to get a workload running on each node.
kubectl run nginx --image nginx --replicas 5
Check that the new workloads now have an address in the proper IP pool allocated for the rack that the node is on with
kubectl get pods -owide.
The grouping of IP addresses assigned to the workloads differ based on what node that they were scheduled to. Additionally, the assigned address for each workload falls within the respective IP pool that selects the rack that they run on.
Note: Calico IPAM will not reassign IP addresses to workloads that are already running. To update running workloads with IP addresses from a newly configured IP pool, they must be recreated. We recommmend doing this before going into production or during a maintenance window.
For more information on the structure of the IP pool resource, see the IP pools reference.