Mirantis Kubernetes Engine 4 Brings Composability to Enterprise Kubernetes
New MKE 4 Delivers Game-Changing Flexibility and Simplicity for Platform Engineers, enabling composable Kubernetes platforms, tuned to use-case requirements, that are autonomously self-correcting, visualized and lifecycle-managed from a single point of control
On November 20th, Mirantis will launch Mirantis Kubernetes Engine 4 – the latest in our MKE line of enterprise Kubernetes solutions. MKE 4 delivers the excellent performance and simple and smooth operator experience of its predecessors, along with an improved, but still familiar web UI. But it also delivers game-changing flexibility and simplicity for platform engineers, including a highly agile Kubernetes cluster architecture, streamlined operations with declarative configuration and built-in operators, and more. If you’re at KubeCon + CloudNativeCon North America this week, our team will be demonstrating MKE 4 along with open source k0s Kubernetes, k0smotron, and Lens Desktop.
A new core Kubernetes cluster architecture based on k0s
Open source k0s Kubernetes, originated by Mirantis, is a CNCF-certified ‘lightweight’ Kubernetes that distributes as a single binary with zero host environment dependencies beyond a recent Linux kernel. It even runs on AMD and ARM CPUs with extremely-sparing resource requirements (a k0s worker can run in as little as 0.5GB RAM). And it installs with its own local ‘k0s’ client, making on-node setup activities a breeze. There’s also a mature open source cluster configuration utility, k0sctl, which MKE 4’s CLI (called mkectl) leverages. Making MKE 4 ‘powered by k0s’ delivers huge benefits to MKE 4 users, including:
Standardization - k0s (and MKE 4) are CNCF-certified standard Kubernetes, so can be expected to ‘just work’ with CNCF Kubernetes components, services, and applications.
Flexibility - The core of k0s is very flexible and very un-opinionated – designed for declarative configuration, and supporting numerous CNCF component options integrated with Kubernetes-standard CXX APIs. This pays big dividends in MKE 4, helping to make Mirantis composability and on-cluster automation work hassle-free (see below).
Agility - k0s (and MKE 4) run on basically any contemporary Linux, with zero manual toil for host setup: no tweaks to configuration, no installation of dependencies, no friction or hassle. One command installs. Another makes the node active for a role. A third joins the node to a cluster as a controller+worker or worker. So MKE 4’s mkectl utility can build and manage clusters very fast (single-digit minutes), reliably, and repeatably. k0s and k0sctl users will feel right at home.
A complete Kubernetes platform composition system
By default, MKE 4 delivers a complete ‘secure enterprise Kubernetes’ platform, with Calico container networking, NGINX ingress, Velero cluster backup, minio S3-compatible object storage, and many more features for security, operational agility, and quality of life. But that’s just the beginning: MKE 4’s declarative platform composition system lets you craft whole multi-service Kubernetes platforms optimized for your use cases. Just use the mkectl client to generate a model deployment template, then add and change composable open source components easily: adding observability, logging/monitoring/alerting, cost analytics, CI/CD, developer meta-platforms for VM hosting, serverless, FaaS, and more!
Composability changes the game for Kubernetes platform engineers. No more hassles figuring out how to implement service X in context Y (guardrails in MKE 4 steer you away from platform compositions that don’t work). No more software supply chain anxiety about whether a component or service is secure. And no more need to deliver half-baked clusters and solve real-world technical and business problems later (with risky manual labor).
Mirantis stands behind MKE 4 composability with automation and expertise. We continually consume upstream code for key CNCF open source components and solutions, validate and harden them, make them composable for MKE 4, then package and provide the artifacts online (for dynamic consumption from our OCI registries or locally from yours for air-gapped installation). You just add the ‘blueprints’ for services to your deployment template, and ‘mkectl apply.’
With MKE 4, the CNCF ecosystem is yours to explore, with Mirantis on hand to guarantee that things work and issues are patched on strict SLAs.
Add-Ons to swiftly enable platforms
Leveraging MKE 4’s composability framework (and the expanding set of pipelines for continuously consuming upstream CNCF innovation, validating it, and making it composable), Mirantis will shortly (and thereafter, regularly) release new component subsystems for MKE 4 that are easy to consume and make part of platforms, adapting them to specific technical and business use-cases.
Already released in MKE 3.7.15 (and soon for MKE 4) is a Helm-enabled ‘blueprint’ called MKE Virtualization, which onboards KubeVirt – enabling an MKE cluster to host VM workloads alongside containers on workers with an installed KVM hypervisor. This creates a single, Kubernetes-native operating environment for containers and VMs, which can be much more cost-effective and operationally agile than expensive proprietary IaaS (we’re looking at you, VMware), as well as less locked-in and less complex than leveraging standard public cloud features (e.g., cloud IaaS and cloud Kubernetes) separately to host the same diverse workload mix. Among other benefits, it gives developers and operators brand new options for managing VMs in cloud native ways: like automated restarts, autoscaling, and so forth. All via the Kubernetes standard API, versus a cloud provider’s proprietary multi-service automation API.
MKE Virtualization will come at no extra charge with an enterprise support subscription for MKE 4. And it’s only the first MKE 4 Add On in the pipeline. Mirantis has a whole roadmap of optional composable solutions coming for MKE 4, including FinOps, LMA, policy enforcement, and more. All just a cut-and-paste and ‘mkectl apply’ away from being usable by your organization.
MKE 4’s web UI will be familiar to users of earlier MKE releases. Basic cluster metrics via Grafana are enabled by default. The web UI provides a host of easy-to-use point and click affordances for defining ingresses, namespaces, and editing other artifacts to control the Kubernetes environment.
MKE 4’s web UI will be familiar to users of earlier MKE releases. Basic cluster metrics via Grafana are enabled by default. The web UI provides a host of easy-to-use point and click affordances for defining ingresses, namespaces, and editing other artifacts to control the Kubernetes environment.
Powerful declarative operations from any Linux, Mac, or Windows desktop or server
The new mkectl client runs on any laptop or jump server as a native application, and the installer brings along k0sctl and kubectl to create a whole MKE 4 management environment. Defining an MKE 4 platform starts by using the mkectl client to generate a default cluster configuration (‘mkectl config > mke.yaml’), then using Mirantis-provided tools to add composable components. Use Git or any standard CI system (e.g., ArgoCD) to version-control platform definitions. Run preflight checks to clear errors, then simply ‘mkectl apply’ your config to deploy on any kind of bare metal or virtual infrastructure (like MKE 3, MKE 4 is ‘bring your own infra’).
Declarative LCM gives you agility, simplicity, and prevents disasters that can occur when you modify running clusters. Once deployed, you can scale, add new composable services, or update your cluster from the mkectl client by modifying the config and reapplying. Changes are made with a workload-sparing rolling-update procedure, and are idempotent and can be rolled back if you encounter issues. Bonus: the mkectl client is non-critical – as long as you have your config (which can be retrieved from the running MKE 4 cluster) you can just install mkectl and its companion apps again, hand it the config, and you’ve regained control of your cluster.
Automated drift correction and zero touch updates
Erroneously executed manual changes endanger cluster stability, and can expose organizations to existential levels of risk by accidentally opening up security holes and causing deviations from required behaviors essential to compliance. Changes can also be implemented by rogue workloads, exposing clusters to further exploits and data exfiltration. To prevent this, MKE 4 deploys with an MKE Operator onboard that does automatic drift correction: any change made manually that differs from declared cluster state is automatically (and quickly) reversed.
For fully-automated, zero touch updates, you can deploy the k0s-native AutoPilot operator for autonomous rolling updates on a schedule. The mkectl operator also has an ‘update’ verb that lets you controllably update an entire deployed MKE 4 platform.
Onramp to Multi-Cluster
With k0s at core, composable platform orientation, declarative configuration, and automation, MKE 4 can serve multiple roles:
For orgs that are in the early stages of Kubernetes adoption
MKE 4 delivers an ultra-modern take on the classic ‘opinionated’ enterprise Kubernetes architecture, composed of open source building blocks. With best-of-breed open source solutions for container networking, ingress, storage, backup, and other critical options pre-integrated, orgs can stop worrying about the difficulty and risks of open source Do-it-Yourself and the labor involved in keeping a complex open source enterprise cluster updated, given its huge number of internal dependencies. MKE 4 lets you deploy reliable, highly secure, performant enterprise Kubernetes in minutes, scale and update using a GitOps-like process, stay free of worries over security and compliance, and stay focused on application development (which, after all, is where the business value lies). Mirantis can even fully manage MKE 4 on your behalf. MKE 4 is supremely flexible -- setting you free from the lock-in factors and prescriptive requirements of proprietary Kubernetes platforms, and the unbounded usage-based costs of cloud Kubernetes.
For orgs that need to maintain large, complex, shared Kubernetes platforms
MKE 4 vastly simplifies the platform engineering required to configure, implement, scale, periodically enhance, and lifecycle manage multi-service clusters serving multiple teams. Composable components and declarative management makes it simple to assemble clusters, add services (prior to or after deployment), version control configurations. Automated drift correction prevents errors from manual changes and encourages adoption of infra-as-code principles. And Mirantis-validated MKE 4 Add Ons for FinOps, LMA, VM hosting and other critical services keeps your large clusters cost-efficient, versatile, and optimized.
For orgs getting ready to adopt Kubernetes multi-cluster and multi-cloud
MKE 4 makes this much simpler. Instead of large, potentially fragile, hard to secure, hard to resource-manage shared clusters, use MKE 4 to define and operate simpler platforms, each tuned to a particular use-case or workload type. While MKE 4 isn’t (by itself, anyway) a complete multi-cluster solution, it still delivers real operational benefits to ensure that the labor involved in creating and operating multiple clusters doesn’t grow linearly with the number of clusters and nodes under management.
Please contact us to learn more about MKE 4, and visit us for a demo at KubeCon + CloudNativeCon, Booth R22, in Salt Lake City this week!
What about MKE 3?
MKE 4 is a new architecture that shares a web UI and essential functionality with earlier MKE versions (the current version of MKE 3.7). MKE 4 will offer new levels of flexibility and operational agility to many current MKE users. But – one thing MKE 4 does not do is support Swarm and Kubernetes dual-mode orchestration – a unique feature of MKE 3 that some Mirantis customers depend on.
For this reason, Mirantis will continue developing and evolving the MKE 3 dual-orchestrator architecture, which now has its own long-term roadmap. MKE 3.8 will be delivered towards the year’s end (watch this space for news). P.S. MKE Virtualization (KubeVirt) is already available free for MKE 3.7.15.
We’re also announcing a non-disruptive upgrade path from MKE 3.7 or 3.8 to MKE 4, for users who don’t require Swarm. Initiated from the MKE 3.x web UI or CLI, the upgrade process is engineered to let workloads keep running as updates are performed node by node. For more information, please contact your Mirantis Customer Success Manager or Account Executive.
