Skip to content
A CNI IPAM plugin that assigns IP addresses cluster-wide
Go Shell Dockerfile
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
cmd add kubernetes CRD backend Oct 9, 2019
doc [doc][daemonset] Adds IPPool CRD to daemonset installation, updates r… Nov 26, 2019
hack
pkg specify namespace on list operation Oct 21, 2019
script
vendor add kubernetes CRD backend Oct 9, 2019
.gitignore
.travis.yml [doc][daemonset] Adds IPPool CRD to daemonset installation, updates r… Nov 26, 2019
Dockerfile add kubernetes CRD backend Oct 9, 2019
LICENSE Initial commit Apr 8, 2019
README.md [minor] Fix kubectl apply command in readme Nov 27, 2019
go.mod
go.sum use scripts in travis for kubebuilder steps Oct 10, 2019

README.md

whereabouts

Travis CI status Go report card

whereabouts-logo

An IP Address Management (IPAM) CNI plugin that assigns IP addresses cluster-wide.

If you need a way to assign IP addresses dynamically across your cluster -- Whereabouts is the tool for you. If you've found that you like how the host-local CNI plugin works, but, you need something that works across all the nodes in your cluster (host-local only knows how to assign IPs to pods on the same node) -- Whereabouts is just what you're looking for.

Whereabouts can be used for both IPv4 & IPv6 addressing.

Introduction

CNI (Container Network Interface) plugins typically have a configuration element named ipam. CNI IPAM plugins can assign IP addresses, and Whereabouts assigns IP addresses within a range -- without having to use a DHCP server.

Whereabouts takes an address range, like 192.168.2.0/24 in CIDR notation, and will assign IP addresses within that range. In this case, it will assign IP addresses from 192.168.2.1 to 192.168.2.255. When an IP address is assigned to a pod, Whereabouts tracks that IP address in a data store for the lifetime of that pod. When the pod is removed, Whereabouts then frees the address and makes it available to assign on subsequent requests. Whereabouts always assigns the lowest value address that's available in the range.

You can also specify ranges to exclude from assignment, so if for example you'd like to assign IP addresses within the range 192.168.2.0/24, you can exclude IP addresses within it by adding them to an exclude list. For example, if you decide to exclude the range 192.168.2.0/28, the first IP address assigned in the range will be 192.168.2.16.

In respect to the old equipment out there that doesn't think that IP addresses that end in .0 are valid -- Whereabouts will not assign addresses that end in .0.

The original inspiration for Whereabouts comes from when users have tried to use the samples from Multus CNI (a CNI plugin that attaches multiple network interfaces to your pods), which includes examples that use the host-local plugin, and they find that it's... Almost the right thing. Sometimes people even assume it'll work across nodes -- and then wind up with IP address collisions.

Whereabouts is designed with Kubernetes in mind, but, isn't limited to use in just Kubernetes.

To track which IP addresses are in use between nodes, Whereabouts uses etcd or a Kubernetes Custom Resource as a backend. The goal is to make Whereabouts more flexible and to use additional storage backends, we welcome any contributions towards this goal.

Please note that Whereabouts is very new. Any issues and PRs are welcome, some of the known limitations are found at the bottom of the README.

Installation

There's two steps to installing Whereabouts:

  • Installing Whereabouts itself (it's just a binary on disk).
  • Creating IPAM CNI configurations.

Optionally, you can use etcd directly as a backend for data storage. In which case, you'll want to install etcd, a suggested method is provided later in the README.

Installing Whereabouts.

You can install this plugin with a Daemonset, using:

git clone https://github.com/dougbtv/whereabouts && cd whereabouts
kubectl apply -f ./doc/daemonset-install.yaml -f ./doc/whereabouts.cni.k8s.io_ippools.yaml

NOTE: This daemonset is for use with Kubernetes version 1.16 and later. It may also be useful with previous versions, however you'll need to change the apiVersion of the daemonset in the provided yaml, see the deprecation notice.

You can compile from this repo (with ./hack/build-go.sh) and copy the resulting binary onto each node in the /opt/cni/bin directory (by default).

Not that we're also including a Custom Resource Definition (CRD) to use the kubernetes datastore option. This installs the kubernetes CRD specification for the ippools.whereabouts.cni.k8s.io/v1alpha1 type.

Example Config

Included here is an entire CNI configuration. Whereabouts only cares about the ipam section of the CNI config. In particular this example uses the macvlan CNI plugin. (If you decide to copy this block and try it too, make sure that the master setting is set to a network interface name that exists on your nodes). Typically, you'll already have a CNI configuration for an existing CNI plugin in your cluster, and you'll just copy the ipam section and modify the values there.

{
      "cniVersion": "0.3.0",
      "name": "whereaboutsexample",
      "type": "macvlan",
      "master": "eth0",
      "mode": "bridge",
      "ipam": {
        "type": "whereabouts",
        "datastore": "kubernetes",
        "kubernetes": { "kubeconfig": "/etc/cni/net.d/whereabouts.d/whereabouts.kubeconfig" },
        "range": "192.168.2.225/28",
        "exclude": [
           "192.168.2.229/30",
           "192.168.2.236/32"
        ],
        "log_file" : "/tmp/whereabouts.log",
        "log_level" : "debug",
        "gateway": "192.168.2.1"
      }
}

Example etcd datastore configuration

If you'll use the etcd datastore option, you'll likely want to install etcd first. Follow the instructions to do so in a later section of the README.

NOTE: You'll almost certainly want to change etcd_host.

{
      "cniVersion": "0.3.0",
      "name": "whereaboutsexample",
      "type": "macvlan",
      "master": "eth0",
      "mode": "bridge",
      "ipam": {
        "type": "whereabouts",
        "etcd_host": "example-etcd-cluster-client.cluster.local:2379",
        "range": "192.168.2.225/28",
        "exclude": [
           "192.168.2.229/30",
           "192.168.2.236/32"
        ],
        "log_file" : "/tmp/whereabouts.log",
        "log_level" : "debug",
        "gateway": "192.168.2.1"
      }
}

Example IPv6 Config

The same applies for the usage of IPv6:

{
      "cniVersion": "0.3.0",
      "name": "whereaboutsexample",
      "type": "macvlan",
      "master": "eth0",
      "mode": "bridge",
      "ipam": {
        "type": "whereabouts",
        "log_file" : "/tmp/whereabouts.log",
                "log_level" : "debug",
        "etcd_host": "example-etcd-cluster-client.cluster.local:2379",
        "range": "2001::0/116",
        "gateway": "2001::f:1"
      }
}

Core Parameters

Required

Three parameters are required:

  • type: This should be set to whereabouts.
  • range: This specifies the range in which IP addresses will be allocated.

In this case the range is set to 192.168.2.225/28, this will allocate IP addresses in the range

If you need a tool to figure out the range of a given CIDR address, try this online tool, subnet-calculator.com.

Range end syntax

Additionally, the range parameter can support a CIDR notation that includes the last IP to use. Example: range: "192.168.2.225-192.168.2.230/28".

Optional

The following parameters are optional:

  • range_start : First IP to use when allocating from the range. Optional, if unset is inferred from the range.
  • range_end : Last IP to use when allocating from the range. Optional, if unset the last ip within the range is determined.
  • exclude: This is a list of CIDRs to be excluded from being allocated.

In the example, we exclude IP addresses in the range 192.168.2.229/30 from being allocated (in this case it's 3 addresses, .229, .230, .231), as well as 192.168.2.236/32 (just a single address).

Note: It's up to you to properly set exclusion ranges that are within your subnet, there's no double checking for you (other than that the CIDR notation parses).

Additionally -- you can set the route, gateway and DNS using anything from the configurations for the static IPAM plugin (as well as additional static IP addresses).

etcd Parameters

Required:

  • etcd_host: This is a connection string for your etcd hosts. It can take a single address or a list, or any other valid etcd connection string.

Optional:

  • etcd_username: Basic Auth username to use when accessing the etcd API.
  • etcd_password: Basic Auth password to use when accessing the etcd API.
  • etcd_key_file: Path to the file containing the etcd private key matching the CNI plugin’s client certificate.
  • etcd_cert_file: Path to the file containing the etcd client certificate issued to the CNI plugin.
  • etcd_ca_cert_file: Path to the file containing the root certificate of the certificate authority (CA) that issued the etcd server certificate.

Logging Parameters

There are two optional parameters for logging, they are:

  • log_file: A file path to a logfile to log to.
  • log_level: Set the logging verbosity, from most to least: debug,error,panic

Installing etcd. (optional)

etcd installation is optional. By default, we recommend the custom resource backend (given in the first example configuration).

We recommend that you if you're trying it out in a lab, that you use the etcd-operator, the installation guide is just a few steps.

Once you've got etcd running -- all you'll need to provide Whereabouts is the endpoint(s) for it. In the etcd-operator style installation, you'd find those with:

kubectl get svc | grep "etcd-cluster-client"

This will give you the service name and the port to use, in this case you'll specify it in the configuration in a service-name:port format, the default port for etcd clients is 2379.

Note: It's important to remember that CNI plugins (typically) run directly on the host and not inside pods. This means that if you use the DNS name (which might look something like example-etcd-cluster-client.default.svc.cluster.local) for the service (recommended) make sure that you can resolve those hostnames directly from your hosts. You may find some tips regarding that here.

Building

Run the build command from the ./hack directory:

./hack/build-go.sh

Acknowledgements

Thanks big time to Tomofumi Hayashi, I utilized his static CNI IPAM plugin as a basis for this project to give me a head start!

The typeface used in the logo is AZONIX, by MixoFX.

Known limitations

  • It has read/write locking to prevent race conditions, but, it's not optimized. It's write locked for all ranges.
  • If you specify overlapping ranges -- you're almost certain to have collisions, so if you specify one config with 192.168.0.0/16 and another with 192.168.0.0/24, you'll have collisions.
    • This could be fixed with an admission controller.
    • And admission controller could also prevent you from starting a pod in a given range if you were out of addresses within that range.
  • There's approximately a cap of 18,500 possible addresses in a given range before you'll have to configure etcd to allow more than 1.5 megs in a value.
  • There's probably a lot of comparison of IP addresses that could be optimized, lots of string conversion.
  • The etcd method that I use is all ASCII. If this was binary, it could probably store more and have more efficient IP address comparison.
  • Unlikely to work in Canada, apparently it would have to be "where aboots?" for Canadians to be able to operate it.
You can’t perform that action at this time.