The rate at which Kubernetes clusters are likely to be deployed in enterprise IT environments is on the verge of accelerating greatly. IT organizations now are being challenged with how to monitor all those Kubernetes clusters as they are being spun up. To address that issue, Rancher Labs has added a multi-cluster capability for the open source Prometheus monitoring tools developed under the auspices of the Cloud Native Computing Foundation (CNCF).
At the same time, Rancher Labs has announced that Rancher Kubernetes Engine (RKE), a Kubernetes installer, and RancherOS, a lightweight operating system for deploying Docker containers, have been ported to Arm processors. Shannon Williams, vice president of sales and marketing for Rancher Labs, says the goal is to make it possible to deploy instances of Kubernetes on both Arm platforms running in the data center as well as at the network edge as part of, for example, an internet of things (IoT) project. Williams concedes demand for Kubernetes on Arm processors is relatively light at present. But over time, Rancher Labs expects to see IT organizations deploying Kubernetes on a mix of processor types and classes from both Arm and Intel.
While there is no shortage of monitoring options when it comes to Kubernetes, Prometheus is frequently employed by developers because it is being developed alongside Kubernetes by the CNCF. The challenge is that each Kubernetes cluster needs to have its own instance of Prometheus running. To simplify the deployment of Kubernetes, Rancher Labs now installs a Prometheus Operator, developed by CoreOS, into the Kubernetes cluster to make it possible to automate the deployment of Prometheus. Rancher also adds a secure proxy through which instances of Prometheus are then linked back to an instance of Grafana, open source visualization software. That secure proxy also serves to ensure PromQL statements are limited to namespaces belonging to a specific project running on a Kubernetes cluster.
While it’s still early days as far as Kubernetes being deployed in production environments, the probability is high that many organizations will find themselves trying to contain Kubernetes cluster sprawl. Not only does each development team usually prefer to employ a single cluster for every application project, multiple teams are likely to favor different distributions of Kubernetes. There may be some IT organizations that will be able to enforce a standard Kubernetes distribution across the enterprise, but most IT organizations today don’t have any such standard in place.
The good news is that long term, Kubernetes not only provides a common construct around which to build DevOps processes, but it also provides a framework that makes it easier to isolate tasks that should be managed by developers from those that should be managed by an IT operations team.
Adoption of Kubernetes will require IT organizations to acquire new tools or hope their existing ones can be upgraded to support a new platform. Rancher Labs is betting IT organizations will prefer to deploy management tools that are specifically designed for Kubernetes. The challenge for organizations will be not so much figuring out how to provision Kubernetes clusters, but rather adapting their processes to manage Kubernetes clusters at true enterprise scale.