Optimizing Pod Assignment: QueueingHint in Kubernetes v1.32
The Kubernetes Scheduler is the core component that determines which nodes will run new pods. It …
With Kubernetes 1.32, the storage manager has officially reached General Availability (GA) status. This marks a significant step towards more efficient and predictable storage allocation for containerized applications. Since version 1.22, when it moved to beta status, the storage manager has proven to be reliable and stable, establishing itself as a useful complementary feature to the CPU Manager.
The storage manager contributes to optimizing storage allocation and alignment during the workload admission process by the Kubelet. This allows users to allocate exclusive memory for Pods in the Guaranteed QoS class. Changes introduced during the beta phase focused primarily on bug fixes, internal restructuring, and observability improvements, such as new metrics and enhanced logging.
The new observability metrics are particularly interesting:
These metrics help developers gain a better understanding of memory allocation in their applications and identify potential issues early.
An important update concerns the reliability of the storage manager. The Kubelet does not guarantee the order of Pods when they are admitted after a restart or reboot. In certain edge cases, this could lead to the storage manager rejecting some Pods, potentially causing a Kubelet failure. The beta implementation previously lacked some checks and logic to prevent these issues. To stabilize the storage manager’s reliability for general availability, small but crucial adjustments were made to the algorithm, enhancing its robustness and handling of edge cases.
The future of the Topology Manager in general, and the storage manager in particular, holds further exciting developments. Notably, ongoing efforts are being made to expand the storage manager’s support for Windows, enabling CPU and memory affinity on Windows operating systems.
This feature is driven by the SIG Node community. We invite you to join us, engage with the community, and share your ideas and feedback on this and other topics. We look forward to hearing from you!
Ayedo is your partner for Kubernetes solutions, helping you make the most of these new features.
Source: Kubernetes Blog
The Kubernetes Scheduler is the core component that determines which nodes will run new pods. It …
Forensic container checkpointing is based on Checkpoint/Restore In Userspace (CRIU) and allows for …
Introduction In the world of cloud-native technologies, OpenTelemetry (OTEL) is often cited as the …