Pivotal Container Service 1.1, Now GA, Helps You Run Kubernetes Without Complexity. Why PKS Just Works.

June 28, 2018 Fred Melo

Pivotal Container Service (PKS) 1.1 is now GA. Download the bits and read the docs.

The theme of the release: operability. IT teams want Kubernetes, and they want day-to-day management to be simple. That was the overwhelming sentiment of our customer interactions since we launched PKS.

How exactly does PKS improve Kubernetes operability? Let’s take a deeper look.

PKS Includes Kubernetes 1.10. And It Gives You Seamless, No Downtime Upgrades

When you use Kubernetes, you want the latest stable release that has been battle-tested in production environments. So PKS comes with the most recent and stable open-source release, Kubernetes 1.10. (Google Container Engine recently upgraded to 1.10 as well.) Check out the release notes for the full rundown on what’s new. PKS 1.1 is also Kubernetes 1.10 certified!

Of course, new features are no fun if upgrades are a pain. But that doesn’t happen with PKS. The service offers you seamless, no downtime updates, just like the rest of Pivotal Cloud Foundry. Upgrade during business hours at your leisure. There’s no need to wait for nights or weekends!

Deploy Kubernetes Across Multiple Availability Zones, and Gain a Fourth Level of HA

PKS 1.0 shipped with three levels of high availability (HA). In PKS 1.1, we ship the fourth level: HA across Availability Zones (AZs).

Now, you can deploy Kubernetes clusters across multiple Availability Zones. This configuration delivers the ultimate resiliency for Kubernetes deployments. You can now tolerate failures at four layers:

  1. Pod level

  2. Process level

  3. VM level

  4. AZ level

If an AZ goes down, your Kubernetes clusters will still be online serving traffic. We added this fourth level for Kubernetes to give customers the level of operational excellence required by complex distributed systems. (In fact, long-time Pivotal customers have come to depend on the four layers of high availability in Cloud Foundry.) Watch this new video to see the new feature in action!

 

Highly Available Clusters Now Beta in PKS 1.1

So we added the fourth layer of HA with multi-AZs to PKS. What else can be done to keep your k8s clusters online? How about HA master nodes across AZs? We ship this feature as a beta in PKS 1.1. Kubernetes documentation calls this “HA clusters.” You may also hear it referred to as “multi-masters”.

Here’s how this feature keeps your k8s clusters available.

The nodes responsible for handling workloads in Kubernetes (“worker nodes”) can be horizontally scaled to support increased load. However, the state of the cluster, API endpoints, and cluster management are responsibilities of the k8s master node and the accompanying etcd key-value store.

In general, there can only be ONE active master/etcd pair at any given time within a Kubernetes cluster. If that master node is down, bad things happen. You can’t deploy additional containers, scale the environment, or even get telemetry on your cluster.

Enter the multi-master feature. With PKS 1.1, you can deploy multiple master etcd nodes per cluster, across multiple Availability Zones. There’s still going to be a single master active at any given time. But if you happen to lose a VM, or if an AZ goes down, PKS will instantly elect one of the masters deployed in a different zone to become active. Your cluster is immediately back to the last good known state!

If you lose a master node, you’ve lost the cluster state. But at least your workers are processing traffic. If you're on a single AZ - and you lose that entire AZ - your workloads are gone. That’s why we’re releasing multi-AZ first, with HA clusters as a fast follow.

Pivotal doesn’t recommend you use the multi-master feature in production yet. But you should enable it for test and dev scenarios. Try out the beta, and tell us what you think!

Full Multi-Tenant Network Automation for Your K8s Clusters

PKS already provides the ability to segregate a single Kubernetes cluster into multiple tenants via namespaces. The service creates one separate network (with a separate logical switch) per namespace. Best of all, it’s completely automated, thanks to VMware NSX-T.

With PKS 1.1, we take the same approach to onboarding new tenants for multiple clusters. This feature allows the highest level of isolation between tenants. Each tenant can have its own entire cluster...which also makes it straightforward to create completely isolated environments.

Let’s say you want to create a new Kubernetes cluster with complete isolation. Just add a few details to your pks create cluster command in the API or CLI. PKS will then automatically deploy a new NSX-T logical switch and accompanying separate network.

Do you prefer to use a single cluster or multiple K8s clusters for multiple tenants? Take your pick! PKS gives you the power to adjust the level of tenancy control and isolation to suit your requirements.

Best of all, you gain this flexibility without adding operational complexity. Cloud Foundry BOSH handles all of this for you!

A New Deployment Option for PKS: Dell EMC’s Pivotal Ready Architecture

Why is hyperconverged infrastructure so popular? Well, you simply procure the gear, drop an appliance in your data center, plug it in, and you’re online. This convenience is now available for PKS deployments, with the new release of Dell EMC’s Pivotal Ready Architecture (PRA).

PRA provides a tested, validated reference architecture for PKS and PAS, Pivotal’s app platform. It features hyper‑converged VxRail from Dell EMC, along with VMware vSphere and NSX-T.

More infrastructure options are a good thing, and PRA will be especially attractive if you want to automate the lifecycle management of your infrastructure.

Handy Add-Ons for PKS Deployments atop VMware: VMware vRealize LI and Wavefront

Running PKS in your data center? Then you may want to consider two useful modules, sold separately.

Log Streaming Aggregation and Analytics with VMware vRealize LI

The built-in integration with VMware vRealize Log Insight allows system and container logs to be easily managed and analyzed. It provides deep operational visibility and faster troubleshooting for PKS and your workloads running on the service.

Container Monitoring Using Wavefront

Wavefront helps you uncover “leading problem indicators of containerized applications” with real-time analytics. How exactly? With slick customizable dashboards that show you what’s happening with your Kubernetes deployments:

Want to see more? Check out this demo video.

Why You Need Cloud Foundry *and* Kubernetes

Every modern business is going to have an app platform, Kubernetes, and serverless functions as part of their IT portfolio. And you’re going to want each of these abstractions to be secure, stable, and scalable. That’s the promise of Pivotal Cloud Foundry: continuously deliver all your apps to every major cloud with a single platform.

Consider the story of West Corp , a global communications and network infrastructure services firm that handles 70% of all 911 calls in North America. According to Thomas Squeo, SVP of Digital Transformation and Enterprise Architecture, Pivotal Application Service and PKS are crucial to its business:

“We’re optimizing around the operational model that comes with the PAS platform ...For the remainder [of apps] that won’t land on PAS, PKS means we can address an even greater percentage of the application portfolio, giving us the versatility of Kubernetes managed containers, but managed with the same operational platform.”

The interest in Kubernetes is massive - and for good reason! For the enterprise, the challenge remains how to run it in production, securely at scale. PKS is purpose-built to solve these challenges.

Ready to get started? Download the bits and read the docs. Then, start deploying containers to PKS and see how easy operability for Kubernetes can be! Want to read more about PKS 1.1? Check out VMware’s product release blog.

Kubernetes and containers will be a hot topic at the cloud event of the year: SpringOne Platform, Sept 24-27 in Washington, DC. Register now! Use discount code S1P200_FMelo for $200 off the registration fee!

About the Author

Fred Melo

Fred Melo is a Technical Director at Pivotal. He spends most of his time helping customers build better software by adopting cloud-native architectures.

Previous
An Inside Look at the Open Service Broker API: Easily Deliver Services to Cloud Foundry and Kubernetes
An Inside Look at the Open Service Broker API: Easily Deliver Services to Cloud Foundry and Kubernetes

Next
Beyond the 12 Factor App: Exploring the DNA of Highly Scalable, Resilient Cloud Applications
Beyond the 12 Factor App: Exploring the DNA of Highly Scalable, Resilient Cloud Applications