My Services Got Old! Can Kubernetes Handle the Aging of Microservices?

The exploding popularity of microservice based applications is taking companies to adopt them along with cloud services to support them. Containers are the common deployment infrastructures that currently serve millions of customers daily, being managed using orchestration platforms that monitor, manage, and automate most of the work. However, there are multiple concerns with the claims put forward by the developers of such tools. In this paper, we study the effects of aging in microservices and the utilization of faults to accelerate aging effects while evaluating the capacity of Kubernetes to detect microservice aging. We consider three operation scenarios for a representative microservice-based system through the utilization of stress testing and fault injection as a manner to potentiate aging in the services composing the system to evaluate the capacity of Kubernetes mechanisms to detect it. The results demonstrate that even though some services tend to accumulate aging effects, with increasing resource consumption, Kubernetes does not detect them nor acts on them, which indicates that the probe mechanisms may be insufficient for aging scenarios. This factor may indicate the necessity for more effective mechanisms, capable of detecting aging early on and act on it in a more proactive manner without requiring the services to become unresponsive.