Do You Really Need Kubernetes? Real‑World Opinions and Practical Tips

Lin Ying’s answer

Many projects combine Spring Cloud with Kubernetes, ending up with dozens of backend services and a lot of servers. For low‑traffic apps (daily active users under ten thousand, a few thousand new records per day), a simple Spring Boot deployment is cheaper and easier; the hype about decoupling is unnecessary when concurrency is low.

Tai Ku La’s answer

The author has worked in companies that push full containerization and Kubernetes regardless of workload. Even with only three servers, they would still install a Kubernetes cluster. If you have a single physical machine, they would even virtualize three nodes just to run K8s. The author believes that if someone says K8s isn’t needed, they probably don’t understand it well.

Reasons for being unable to live without K8s:

Cloud providers offer managed master clusters; you only pay for nodes, which costs about the same as a regular ECS instance.

Physical servers are still required, but K8s lets you use a server directly as a node, avoiding extra virtualization layers.

With kubectl, most operations no longer require SSH.

Typical kubectl commands:

kubectl create deploy xx --image=xxx

kubectl expose deploy xx --type=NodePort --target-port=80 --port=80

kubectl scale deploy xx --replicas=3

kubectl edit ing ...

kubectl set image xxx app=image=xxx</code> or <code>helm upgrade ...

These one‑liners replace lengthy shell scripts for rolling updates.

Anonymous user’s answer

In a team without an ops department, they only used Docker and an SLB for reverse proxy. When a container went down for two days, nobody noticed until a colleague checked the logs. Scaling required manually adding machines and editing Nginx/SLB configurations.

W’s answer

Ask yourself why you want K8s, whether it solves your problems, and if you’re ready for long‑term maintenance, CI/CD migration, and sufficient staffing. Without proper planning, K8s can become a source of hidden failures.

Yunqian Tech’s answer

A past project adopted microservices with about a dozen controllers and tens of thousands of daily page views. After six months the team disbanded, but developers gained microservice experience on their resumes.

runzhliu’s answer

Kubernetes is popular mainly because of its service‑orchestration capabilities. Its community hype mirrors early big‑data frameworks like Hadoop and Spark, often overlooking the operational burden it introduces.

Liu Xin’s answer

Often, it’s the ops department that pushes complex architectures, not the developers.

Net Addict’s answer

Rotating 200+ nodes without K8s would take weeks and risk P0 incidents; with K8s the same operation took only two days.

RLLvmd’s answer

K8s’ self‑healing automatically restarted a failing container over three thousand times in an hour, keeping the service uninterrupted for users.

Chen Moore’s answer

For a small setup of 3‑5 EC2 instances, using Spring Cloud or K8s feels overkill; a simple Nginx front‑end for a Spring Boot app is sufficient.