Data / Kubernetes / Open Source / Sponsored / Contributed

KOps Adds Support for Calico’s eBPF Dataplane

17 May 2021 10:25am, by

Reza Ramezanpour
Reza is a developer advocate at Tigera, working to promote adoption of Project Calico. Before joining Tigera, Reza worked as a systems engineer and network administrator.

Kubernetes operations (kOps) is one of the official Kubernetes (K8s) projects. The kOps project allows for the rapid deployment of production-grade K8s clusters on multiple cloud platforms. By leveraging YAML manifests, kOps delivers a familiar experience to users who have worked with kubectl. Similar to K8s clusters in popular cloud platforms, kOps helps set up self-managed clusters to easily deliver high availability. Given its ease of use, it is a popular choice when users want to deploy self-hosted Kubernetes clusters.

With the recent release of kOps v1.19, support for the Calico eBPF dataplane was added to the utility. In addition to the above-mentioned features, the latest kOps update offers an effortless way to autodeploy K8s clusters using Project Calico for networking and the Calico eBPF dataplane. Calico eBPF dataplane implementation replaces kube-proxy and delivers equivalent functionality. It also leverages the most optimal datapath for traffic. These changes deliver a network performance boost and source IP preservation to your cluster.

In this blog post, we will showcase the steps required to deploy a cluster that uses these newly available features.

What is eBPF?

eBPF is a virtual machine embedded within the Linux kernel. It allows small programs to be loaded into the kernel and attached to hooks, which are triggered when some event occurs. This allows the behavior of the kernel to be (sometimes heavily) customized. While the eBPF virtual machine is the same for each type of hook, the capabilities of the hooks vary considerably. Since loading programs into the kernel would otherwise be a security risk, the kernel runs all programs through a very strict static verifier. The verifier sandboxes the program, ensuring it can only access allowed parts of memory and ensuring that it must terminate quickly.

Calico’s eBPF dataplane makes use of eBPF functionality to allow source IP preservation, direct server return (DSR) and even better performance. Let’s explore this in more detail.

Source IP Preservation

In a default networking setup, kube-proxy pods are in charge of routing inbound (ingress) traffic to target pods using network address translation (NAT) technology. NAT might be a good option to deliver packets. However, in the process of delivery, NAT replaces the original client source IP address, making troubleshooting and security implementation difficult since packets do not represent their original source.

Kube-proxy cluster IP

As an alternative to using Kubernetes standard kube-proxy, Calico’s eBPF dataplane also supports native service handling. This preserves source IP to simplify network policy, offers DSR to reduce the number of network hops for return traffic and provides even load-balancing independent of topology, with reduced CPU and latency compared to kube-proxy.

Calico native service handling

Performance

Calico eBPF dataplane mode is an alternative to our standard Linux dataplane mode, based on iptables, that pushes the performance limits further. In eBPF dataplane mode, Calico leverages eBPF programs to process packets rapidly and efficiently without ever leaving the packet-processing context of the Linux kernel. The efficiency achieved by this method is close to natively compiled code in the kernel.

*Performance data generated using the kubernetes network benchmark project.

Calico’s eBPF dataplane attaches eBPF programs to each Calico interface as well as the data and tunnel interfaces. This allows Calico to spot workload packets early and handle them through a fast path that bypasses iptables and other packet processing that the kernel would normally do.

Replacing kube-proxy

Calico’s eBPF dataplane includes native service handling, so you no longer need to run kube-proxy. Calico’s native service handling outperforms kube-proxy both in terms of networking and control plane performance, and supports features such as source IP preservation.

The differences in performance are most noticeable if you have workloads that are particularly sensitive to network latency or if you are running large numbers of services. You can read more about these performance advantages and find a range of different benchmark charts comparing kube-proxy performance with Calico’s native service handling in this blog from the Calico team.

CPU usage with 5K services (smaller is better)

Demo

Before We Begin: Install aws-cli and kubectl

Note: This section demonstrates how to deploy a K8s cluster equipped with Calico’s eBPF dataplane, using kOps in AWS cloud platform. If you are using a different cloud provider, feel free to change the settings to meet your needs. kOps supports multiple cloud providers such as Google, Microsoft Azure and many more.

To populate a cluster in the AWS public cloud, kOps uses aws-cli command line capabilities. Therefore, it is a requirement to install and configure aws-cli on your machine before following the below how-to section. You will also need kubectl to interact with your K8s API server. If you are not sure how to install these requirements, check out the following links:

How-to

kOps can create self-managed K8s clusters without hassle in a matter of seconds. There are a few configurations required to successfully deploy a kOps cluster.

S3 bucket (AWS-specific)

Amazon S3 bucket is a storage server used to store objects in AWS cloud. kOps takes advantage of this storage system to store cluster configuration and deployment status.

Note: AWS S3 bucket names are globally unique, which means when an S3 is created by a user, that name will be unavailable to other customers. Please make sure you change the calico-kops-demo to a unique favorite name before executing the following command.

First, execute the following command to create a new S3 bucket:

You should see a result similar to:

Note: kOps recommends that you turn on the S3 versioning feature using the --versioning-configuration Status=Enabled argument. In a production environment, this can help if you ever need to revert or recover your cluster.

After creating the S3 bucket, we will need to store the name in an environment variable for kOps to use, using the following command:

If you are following this demo in Windows, please use the following PowerShell command:

If you would prefer not to create an environment variable, use the --state argument to define your S3 store name.

Then, use the following command to create the required configuration files in your S3 bucket:

Note: No resources will be created at this point (EC2, routes, etc.). The following command only creates necessary config files in your S3 bucket.

Let’s go through some of the parameters that were used to create the cluster.

By adding the following line to our configuration, we disabled kubeproxy pods and prevented them from being installed on our cluster, given that we are going to implement Calico in eBPF mode to take on the responsibility of directing our network traffic flow.

The networking section in kOps declares which container network interface (CNI) should be used in the cluster.

We can configure Calico using parameters available in kOps. For example, turning on the eBPF dataplane is as easy as setting the bpfEnabled parameter with a true variable and appending it to the network section of the manifest.

We can also customize our installation further with parameters such as:

Now that we have created the cluster, we need to provide an SSH public key. Note that kOps and AWS can only import RSA keys, not other key types:

Note: AWS EC2 resources created by kOps require an SSH key to be imported for SSH authentication. Make sure you have a valid SSH key.

On Mac and Linux, creating an SSH key is as simple as running ssh-keygen.

If you are using Windows, SSH keys can be created using programs like PuTTYgen.

Use the following command to create the cluster resources:

After issuing this command, EC2 and other required resources will be created in AWS. This will take some time.

Since cluster creation can be time-consuming, you can use the following command to find out when creation is complete:

After some time, you should see a result similar to:

That’s it; now you have a self-managed K8s cluster equipped with a Calico eBPF dataplane.

Clean Up

If you would like to remove the resources we have just created, use the following commands:

Conclusion

As we have established in this blog post, running a self-managed K8s cluster equipped with Calico using an eBPF dataplane is an easy task using the kOps deployment tool. kOps abstracts away much of the complexity, but still supports eBPF and Calico for excellent performance. The resulting cluster not only outperforms vanilla cluster networking but offers additional features.

Did you know you can become a certified Calico operator? Learn Kubernetes networking and security fundamentals using Calico in this free, self-paced certification course.

If you enjoyed this post, you might also like:

The New Stack is a wholly owned subsidiary of Insight Partners. TNS owner Insight Partners is an investor in the following companies: Docker, Tigera.

Featured image via Pixabay.

A newsletter digest of the week’s most important stories & analyses.