Advanced AddressPool configuration

Controlling automatic address allocation

In some environments, you’ll have some large address pools of “cheap” private IPs (e.g. RFC1918), and some smaller pools of “expensive” IPs (e.g. leased public IPv4 addresses).

By default, MetalLB will allocate IPs from any configured address pool with free addresses. This might end up using “expensive” addresses for services that don’t require it.

To prevent this behaviour you can disable automatic allocation for a pool by setting the autoAssign flag to false:

apiVersion: metallb.io/v1beta1
kind: IPAddressPool
metadata:
  name: cheap
  namespace: metallb-system
spec:
  addresses:
  - 192.168.10.0/24
apiVersion: metallb.io/v1beta1
kind: IPAddressPool
metadata:
  name: expensive
  namespace: metallb-system
spec:
  addresses:
  - 42.176.25.64/30
  autoAssign: false

Addresses can still be specifically allocated from the “expensive” pool with the methods described in the usage section.

Note

To specify a single IP address in a pool, use /32 in the CIDR notation (e.g. 42.176.25.64/32).

Reduce scope of address allocation to specific Namespace and Service

This option can be used to reduce the scope of particular IPAddressPool to set of namespaces and services, by adding an optional namespace and/or service selectors. This is useful for mutitenant context in which there is a need for pinning IPAddressPool to specific namespace/service.

apiVersion: metallb.io/v1beta1
kind: IPAddressPool
metadata:
  name: ippool-ns-service-alloc-sample
  namespace: metallb-system
spec:
  addresses:
    - 192.168.20.0/24
  avoidBuggyIPs: true
  serviceAllocation:
    priority: 50
    namespaces:
      - namespace-a
      - namespace-b
    namespaceSelectors:
      - matchLabels:
          foo: bar
    serviceSelectors:
      - matchExpressions:
          - {key: app, operator: In, values: [bar]}

The above IPAddressPool example is pinned to Service(s) which has a label matching with expression key: app, operator: In, values: [bar] created either in namespace-a or namespace-b or any namespace has a label foo:bar.

If multiple IPAddressPool objects are available to a Service, MetalLB will check for the availability of IPs by sorting the matching IPAddressPool objects by priority. It will first select the IPAddressPool with the lowest priority number (i.e., priority=1 is the highest priority). If the priority field is unset or set to 0, it will have the lowest priority (i.e., it will be the last pool to be used). If multiple IPAddressPool objects have the same priority, the choice will be random.

Note

When a service explicitly chooses an IPAddressPool via metallb.universe.tf/address-pool annotation or an IP address via spec.loadBalancerIP or metallb.universe.tf/loadBalancerIPs annotation which doesn’t match the service will stay in pending.

Handling buggy networks

Some old consumer network equipment mistakenly blocks IP addresses ending in .0 and .255, because of misguided smurf protection.

If you encounter this issue with your users or networks, you can set the AvoidBuggyIPs flag of the IPAddressPool CR. By doing so, the .0 and the .255 addresses will be avoided.

Changing the IP of a service

The current behaviour of MetalLB is to try to preserve the connectivity despite a change of configuration that might disrupt a service happens. For example, removing an IPAddressPool that contains IPs currently assigned to services.

If that happens, instead of reallocating (if possible) a new IP to the service, the configuration change is marked as stale and MetalLB keeps running with the last valid configuration.

In order to re-assign a new IP to the services, there are two options:

  • restarting the MetalLB’s controller pod
  • deleting and re-creating the service
Note

This behaviour is subject to change making MetalLB observe any state requested by the user, regardless of the fact that it may cause service disruptions or not.

PreferDualStack IP Family Policy

MetalLB supports PreferDualStack ip policy, which allows services to prefer dual-stack IP allocations while gracefully falling back to single-stack if dual-stack is unavailable.

Consider this service

apiVersion: v1
kind: Service
metadata:
  name: my-app
spec:
  type: LoadBalancer
  ipFamilyPolicy: PreferDualStack
  ipFamilies:
  - ipv6
  - ipv4

Suppose that the advertisement is as followed:

apiVersion: metallb.io/v1beta1
kind: BGPAdvertisement
metadata:
  name: external
  namespace: metallb-system
spec:
  ipAddressPools:
  - ipv4-pool
  - ipv6-pool
  - dualstack-pool

The above Service will be assigned 1 ipv4 and 1 ipv6 from the dualstack-pool.

In case the dualstack-pool is not added to ipAddressPools, svc my-app will be assigned an ipv6 from the pool ipv6-pool, as ipv6 is the first ip family listed in the service’s spec.ipFamilies field.

In case the field spec.ipFamilies is empty, ipv4 will be preferred.

Metallb’s behaviours in svc with PreferDualStack family policy, in case of pool change:

  • If an svc was assigned with an ip from a single-stack pool, and the said pool is extended to be dualstack, the originally assigned ip will stay assigned, and the svc will be assigned an additional ip from the newly available stack.
  • If an svc was assigned with ips from a dual-stack pool, and the said pool becomes single-stack, the assigned ips are considered malfunctioning, both of them are cleared from the svc. Metallb will then perform a re-assign.
  • When an svc was assigned with an ip from a single-stack pool, and a dual-stack pool becomes available, the svc will keep its current pool.