Enterprises Are Moving Beyond the “Databases Don’t Belong on Kubernetes” Debate
For years, conventional wisdom held that databases and Kubernetes simply did not belong together.
Kubernetes was designed for stateless workloads. Databases are inherently stateful. The operational risks surrounding storage persistence, failover, performance consistency, and data protection led many organizations to keep their databases outside of containerized environments.
That narrative is changing. Today, more than 60% of enterprises run at least some stateful workloads on Kubernetes, reflecting growing confidence in the platform’s maturity. Improvements in storage architectures, Kubernetes operators, observability platforms, and Day 2 operational tooling have significantly reduced many of the concerns that once prevented database adoption.
At the same time, AI is beginning to influence database operations, creating new opportunities for automation while raising important questions about trust, governance, and operational control.
In this episode of AppDevANGLE, I spoke with Peter Farkas, CEO of Percona, about the evolution of database operations, the growing role of Kubernetes in stateful environments, AI-assisted database management, and why open-source flexibility remains strategically important as enterprises modernize their data infrastructure.
Our conversation explored why database adoption on Kubernetes has become a practical reality, where AI can deliver meaningful operational value, and why infrastructure freedom is becoming a critical long-term business requirement.
Kubernetes Has Become a Practical Choice for Database Operations
One of the most notable shifts in the market is how dramatically attitudes toward Kubernetes databases have evolved. As Farkas explained, the conversation has moved from an outright rejection of stateful workloads on Kubernetes to a more nuanced understanding of where the platform provides operational advantages.
“Ten years ago, the mantra was stateless on Kubernetes, Kubernetes stateful elsewhere,” Farkas explained. “A DBA would tell you they would not be responsible for any data loss or performance issues if you ran your database in a container.”
That position has softened significantly. The biggest driver has been the maturation of Kubernetes operators.
“These operators are not just deployment tools,” said Farkas. “They really encode operational knowledge. They handle backups, scaling, and failover with a level of consistency that a human administrator would probably not be able to match at three o’clock in the morning.”
Rather than treating Kubernetes as simply an orchestration layer, organizations increasingly view it as a platform for standardizing database operations across multiple environments. The result is greater consistency, improved automation, and simplified operational management for platform teams supporting diverse application portfolios.
Kubernetes Is Not a Universal Answer
While Kubernetes adoption continues to expand, Farkas emphasized that organizations should avoid treating it as a universal solution.
“There isn’t a silver bullet,” he said. “That is true for databases and it is true for Kubernetes.”
Certain workloads still benefit from traditional deployment models. Ultra-low-latency environments such as high-frequency trading platforms, real-time bidding systems, and highly specialized performance-sensitive applications may continue to favor bare-metal architectures. Similarly, organizations operating under strict regulatory frameworks may encounter compliance requirements that necessitate physical isolation of infrastructure resources.
“If the compliance framework requires physical hardware isolation, then you will need to stick to bare metal,” Farkas noted.
However, these remain exceptions rather than the norm. For most enterprise workloads, the operational consistency delivered by Kubernetes often outweighs the modest performance overhead associated with containerization. This reflects a broader industry trend where operational simplicity is increasingly viewed as a competitive advantage.
AI Is Becoming an Operations Assistant, Not an Autonomous Operator
The conversation also addressed the growing role of AI in database operations. While vendor messaging often focuses on autonomous databases and self-healing systems, Farkas offered a more pragmatic perspective on where AI is delivering value today.
“A good analogy would be self-driving cars,” he explained. According to Farkas, AI is already proving highly effective in observability, diagnostics, and anomaly detection. Database teams can leverage AI to analyze thousands of performance metrics simultaneously, identify unusual patterns, and surface likely root causes far faster than traditional monitoring systems.
“AI is incredible at looking at thousands and thousands of metrics and saying, ‘This specific query pattern caused the spike,’” he said.
However, that does not mean organizations should relinquish operational control.
“The hype is that all DBAs and system administrators will be replaced by self-healing databases,” Farkas said. “I don’t think we are going to do that.”
Instead, Percona advocates for what Farkas describes as augmented intelligence rather than autonomous intelligence.
“At Percona, our stance is that we would like to focus on augmented intelligence where we provide insights to the human who can make the final call.”
This distinction is increasingly important as organizations balance automation benefits with governance requirements and operational accountability.
Outcome-Based Database Operations Are Gaining Momentum
Another key theme from the discussion was the growing demand for outcome-oriented infrastructure services.
Historically, organizations purchased support contracts, consulting engagements, or platform subscriptions. Increasingly, however, enterprises are looking for measurable business outcomes rather than collections of technical services.
Farkas explained that this trend influenced Percona’s recent focus on solution bundles designed around specific operational goals.
“Most enterprises don’t just want a specific service,” he said. “They want an outcome.”
As database environments become more complex, many internal teams struggle to manage multiple database technologies, hybrid cloud architectures, Kubernetes deployments, and emerging AI workloads simultaneously.
Organizations increasingly value partners who can help deliver targeted outcomes such as AI readiness, modernization initiatives, platform migration, or operational efficiency improvements without requiring lengthy infrastructure transformation projects.
The shift mirrors broader enterprise technology buying behavior, where outcomes are becoming more important than products.
Open Source Simplicity Is Becoming a Strategic Priority
Looking ahead, Percona sees Kubernetes and open source converging around a common objective: making enterprise-grade data infrastructure easier to operate. Farkas believes one of the largest opportunities for the open-source ecosystem is improving usability.
“There is still a big gap for open source to fill when it comes to ease of use,” he explained.
Cloud providers have established expectations around self-service provisioning, simplified management, and operational convenience. Open-source technologies increasingly need to deliver similar user experiences without sacrificing flexibility or control.
Percona’s long-term strategy centers on building Kubernetes-native data platforms that simplify database operations while preserving the freedoms associated with open-source software.
“Our aim is for open source to be as easy to use and as convenient to use as a proprietary solution,” Farkas said.
For platform teams seeking alternatives to vendor lock-in, this represents an important evolution in the open-source ecosystem.
Analyst Take
The conversation around databases on Kubernetes has finally matured. The industry has moved beyond ideological debates about whether databases belong on Kubernetes and shifted toward practical discussions about operational tradeoffs, workload suitability, and platform standardization.
What makes this transition important is that Kubernetes is no longer primarily a deployment technology. It is becoming an operational abstraction layer that simplifies how organizations manage increasingly complex infrastructure environments.
At the same time, AI is following a similarly pragmatic trajectory. The most valuable near-term AI use cases are not fully autonomous databases or self-managing platforms. Instead, organizations are finding success with AI-assisted observability, diagnostics, and operational intelligence that help humans make better decisions faster.
Perhaps the most important takeaway from the discussion is Farkas’ emphasis on freedom. As enterprises evaluate Kubernetes, AI platforms, cloud providers, and database technologies, flexibility is becoming just as important as performance. Organizations increasingly recognize that infrastructure decisions made today will shape their ability to innovate tomorrow.
Open source, Kubernetes, and AI may all be part of that future.The organizations that succeed will be the ones that use these technologies to increase operational agility without sacrificing control over their data, platforms, and long-term strategic direction.
