April 2025 - v1.10.0

🚀 Highlights

Excluding Specific Nodes from Optimization

CloudPilot AI now supports the ability to exclude specific nodes from optimization processes. This is useful in scenarios such as gradually rolling out optimizations to test stability or when certain nodes are already covered by Savings Plans or Reserved Instances.

To exclude nodes from optimization, simply add the annotation karpenter.sh/do-not-disrupt: “true” to the nodes you want to preserve. For example:


$ kubectl get node
NAME                       STATUS   ROLES    AGE    VERSION
node-01.172.16.0.167       Ready    <none>   60m    v1.31.1-k8s.1
node-02.172.16.3.212       Ready    <none>   2m8s   v1.31.1-k8s.1
 
$ kubectl annotate node node-01.172.16.0.167 karpenter.sh/do-not-disrupt=true
node/node-01.172.16.0.167 annotated

Once the annotation is applied, the node will be excluded from all optimization actions. This behavior can be verified in the CloudPilot AI console.

Enhanced Spot Instance Prediction Algorithm (v2)

CloudPilot AI introduces a new version of its spot instance prediction algorithm. It leverages more contextual data to anticipate upcoming spot interruptions, allowing your workloads to migrate ahead of time with greater reliability.

This enhancement helps you use spot instances more confidently, even for production workloads.

The figure below demonstrates how the v2 algorithm proactively migrates workloads ahead of upcoming spot events, resulting in a significant reduction in spot-related disruptions.

Real-world Production data: fewer interruptions with predictive eviction

⚙️ Enhancements

Ensure replicas are scheduled across at least two nodes for unknown workload types, especially for custom resources (CRs).
Display real-time optimization plans whenever workload configurations are updated.

🛠️ Bug Fixes

Fixed an issue in the onboarding process where multiple node groups shared the same name in CloudPilot AI.
Corrected AWS node capacity extraction to improve pricing accuracy.
Resolved an issue where nodes were repeatedly recreated or deleted, improving overall system stability.

We believe these updates further enhance CloudPilot AI’s performance, efficiency, and usability. For any questions or support, feel free to connect with us via our Slack channel .

Stay tuned for more updates! 🚀