DEVOPSdigest

In the past few years, Kubernetes (open sourced by Google in 2014) has moved from the bleeding edge of technology to critical mass adoption. Kubernetes is a container orchestration system that has democratized distributed, microservice-based architectures for at-scale deployments, thrusting businesses into a new generation of cloud native computing.

Kubernetes usage is clearly on the rise. Yet more and more companies using Kubernetes with applications in production are facing challenges managing and operating the systems. This begged the question: Which challenges are commonplace, and what patterns can we observe from Kubernetes usage?

To answer this question, over 1,000 software engineers, DevOps practitioners, and IT architects were surveyed. The report, Managing and Scaling Applications in Kubernetes, was published last year and revealed three key insights:

1. While everyone is using Kubernetes, not everything runs in Kubernetes

Kubernetes adoption is high — a whopping 70% of survey respondents reported using Kubernetes-based container orchestrators for at least one business project. Yet 58% of respondents were running less than half of their applications in Kubernetes. IT departments everywhere have adopted Kubernetes, but their deployments are often limited to non-critical applications. As Kubernetes continues to evolve, new challenges are emerging as businesses strive to scale existing deployments. 

2. Increased adoption leads to increased complexity

As Kubernetes grows in popularity, it fragments and becomes exponentially complex. While the use of Kubernetes is prevalent, not all deployments are uniform. DevOps teams are deploying Kubernetes in a multitude of ways. Most companies operate heterogeneous environments where no two clusters are alike. This has its advantages, as it grants organizations the chance to put each workload in the environment that suits it most. It also creates complexity, as Kubernetes environments become increasingly multi-cluster, multi-cloud, and multi-ingress.

As businesses scale the size of their data footprint, it rarely makes sense to deploy all their infrastructure in just one Kubernetes cluster. Multi-cluster is a strategy for deploying an application across multiple Kubernetes clusters, treating clusters as disposable commodities. This brings availability, improved latency, better performance, isolation, and operational readiness. According to the survey, more than 60% of Kubernetes users are running two or more clusters in production. They are doing so to separate services across tiers, locales, teams, or providers - and adding to the complexity of their infrastructure in the process.

The future is multi-cloud. More and more organizations are using multiple public clouds at the same time, different clouds for different workloads, and combining public clouds with traditional on-prem infrastructures (ie. hybrid cloud). Multi-cloud strategies increase agility and flexibility, minimize vendor lock-in, benefit from best-of-breed features, and improve cost efficiency. Crucially, they allow businesses to control the geography of applications and adhere to stringent data regulations as a result. More than half of survey respondents were using multi-cloud (AWS and GCP were the primary clouds) and valuing the flexibility to run applications where needed.

With time, a vast and fragmented cloud native ecosystem with hundreds of competing and overlapping vendors, tools, and platforms has grown around Kubernetes. Dozens of tools are required to manage today's distributed environments. Over 60% of survey respondents use multi-ingress solutions to manage access to Kubernetes clusters, indicating the use of disparate toolsets to manage the growing array of technologies as end-users navigate a landscape filled with similar vendors and offerings. 

3. Heterogeneous in nature, today's Kubernetes environments are difficult to manage and operate

Today's world is multi-everything and heterogeneous in nature. The growing array of clusters, clouds, and ingresses being used all at once is compounding the innate complexity of Kubernetes. This creates a number of challenges for the management and operations of Kubernetes environments.

Troubleshooting was highlighted as a top concern for DevOps teams. It is difficult to identify the root cause of problems when applications are distributed and deployments are diverse.

As companies add more and more technologies to their stacks, observability becomes an uphill battle. Each solution comes with its tools and can be quite siloed in its own right. Without a single pane of glass across heterogeneous environments, visibility into clusters and ingresses can be opaque.

And even though most companies today already use an observability platform (the survey showed 74% use Grafana and 68% use Prometheus), there is evidence that these tools are not sufficient. Setup and configuration is often difficult and time-consuming, and when incidents occur, developers prefer to log directly into clusters to manually review logs.

Observability is crucial for managing and operating Kubernetes environments, but the heterogeneity of the technology makes it a key challenge for DevOps teams. 

Kubernetes has become the de facto standard for container orchestration and has been adopted en masse across all industries. Yet new emerging challenges must now be overcome before businesses can scale existing deployments. As Kubernetes and its surrounding ecosystem continues to evolve, applications become increasingly fragmented, distributed, and heterogeneous — a key trend that will likely continue for the foreseeable future.