Kubernetes 1.24: The Revolution in Storage Management is Here!

The release of Kubernetes 1.24 introduces Storage Capacity Tracking as a generally available feature. This innovation enables developers and DevOps teams to better manage and monitor the capacity of their storage resources.

What specifically changes for developers/DevOps teams?

With the new Storage Capacity Tracking, CSI drivers (Container Storage Interface) can provide information about the remaining storage capacity. The kube-scheduler uses this information to select appropriate nodes for Pods that still require volumes.

Without this information, a Pod might end up in a waiting loop because the kube-scheduler cannot find suitable nodes. In such cases, a node might be selected where the volume cannot be provisioned because the underlying storage lacks sufficient capacity.

Storage Capacity Tracking ensures that Pods are scheduled more efficiently by directing them to nodes with adequate available storage capacity. This not only improves efficiency but also minimizes the risk of failures or long wait times.

Practical examples or use cases

A practical example of applying Storage Capacity Tracking could be in an e-commerce system running on Kubernetes. For instance, when multiple Pods are created to handle high user load, tracking ensures that the Pods are scheduled on nodes with sufficient storage for their volumes. This ensures the application remains stable and can quickly respond to user requests.

Another example is data processing in the cloud. Here, data processing Pods can be scheduled to be provisioned on nodes with enough storage capacity for processing large data volumes, enhancing the overall application’s performance.

It is also important to mention that ayedo, as a Kubernetes partner, supports companies in optimally leveraging this new functionality to enhance the efficiency of their Kubernetes environments.

Problems we haven’t solved

Despite the many advantages, there are some limitations. If a Pod uses multiple volumes and only one can be successfully provisioned, future scheduling decisions may be constrained by the already provisioned volume. This can lead to the Pod not being scheduled on an appropriate node if the other volume cannot be provisioned there.

A proposal to solve this problem was presented in a KEP draft (Kubernetes Enhancement Proposal), suggesting that volumes that are provisioned but not yet used could be released and reprovisioned elsewhere. This could further increase efficiency, but there is a need for interested developers to pursue this idea further.

Additionally, there is no support in the Cluster Autoscaler for Pods with volumes yet. A prototype was developed but not implemented due to the complexity and challenges in configuration. Further development is needed here to create a closer coupling between the Autoscaler and CSI drivers.

Conclusion

The introduction of Storage Capacity Tracking in Kubernetes 1.24 represents a significant advancement in storage capacity management. This functionality will not only improve the efficiency of deployments but also the stability and reliability of applications running in Kubernetes. Companies using Kubernetes should integrate this innovation into their development processes to fully exploit the potential of their container orchestration.

Source: Kubernetes Blog