A survey of 1,200 IT professionals conducted by Canonical that was published today finds close to half (46%) are employing Kubernetes in a production environment, but from an overall adoption perspective, many challenges remain.
The most commonly cited hurdles are lack of in-house skills (55%), IT infrastructure (37%), incompatibility with legacy systems (33%), difficulty training users (30%), compliance concerns (35%) and integrating cloud-native applications (19%).
The most important goals in using Kubernetes and cloud-native technologies are improved maintenance, monitoring and automation (65%), modernizing infrastructure (46%) and faster time to market (27%). Just over two-fifths of respondents said they are now managing more than 500 machines.
More than three quarters of respondents (78%) reported at least one hybrid or multi-cloud use case in production in their organization. The most common use cases for hybrid and multi-cloud are acceleration of development and increased automation (21%), expanding cloud backup options to cut costs (13%), disaster recovery (13%) and clustering of mission-critical databases (6%).
David Booth, vice president of cloud-native applications at Canonical, says the results make it clear that there is greater need to rely on automation to manage Kubernetes environments. In fact, the bulk of survey respondents are either using operators to automate tasks in production environments (14%) or are experimenting with them (17%). Just under 30% of survey respondents are investigating them. Only 28% use Helm charts of varying maturity to manage software in Kubernetes environments. A full 40%, however, are reusing the same tools, such as bash scripting, that they employ to manage virtual machines.
Overall, only 16% of survey respondents said they are running Kubernetes exclusively, which Booth says suggests that Kubernetes is being deployed alongside a wide range of legacy platforms. In most case, Kubernetes is initially deployed on top of virtual machines. However, use of Kubernetes on bare-metal platforms continues to increase.
Regardless of where Kubernetes is deployed, it’s apparent that many organizations need to upskill to manage Kubernetes environments, adds Booth. The issue, of course, is the degree to which organizations prefer to hire or train site reliability engineers (SREs) to manage fleets of Kubernetes clusters versus traditional IT administrators. A single SRE is capable of managing fleets of Kubernetes clusters at scale, but they are often difficult to find and retain, given the level of competition for IT professionals with those skills. IT administrators not only cost less, but there are now a wider range of graphical tools that enable IT administrators to manage Kubernetes clusters at scale more easily.
In some ways, organizations that adopted Kubernetes early are at more of disadvantage than those that came to it later, simply because the tooling surrounding the platform today is that much more advanced, notes Booth. Kubernetes is simultaneously one of the most powerful and complex IT platforms to ever find its way into a mainstream IT environment. The challenge is finding the right layer of abstraction to apply over it depending on the skills and expertise of the IT team being asked to manage it.