kubeelasti.dev

What It Is

KubeElasti is a CNCF Sandbox project positioned as a Kubernetes-native scale-to-zero Operator for HTTP services. Its core goal is to reduce replicas to 0 when a service is idle to save costs, then automatically restore the service to its minimum replica count when traffic arrives, while using a proxy queue to avoid request loss. The project was originally created by TrueFoundry and is developed openly on GitHub.

Core Capabilities and Integrations

KubeElasti integrates with existing Services through the ElastiService CRD, referencing a Deployment, StatefulSet, or Argo Rollouts resource as the scaling target. It continuously evaluates trigger conditions such as Prometheus queries and thresholds. When idle conditions are met, it enters Proxy Mode, scales replicas down to 0, and has the resolver receive and queue requests. When a valid request arrives or a trigger requires recovery, it scales the service up to minTargetReplicas and enters Serve Mode, allowing traffic to flow directly to Pods again. It can also respond directly to health checks while at 0 replicas, preventing load balancer probes from causing unnecessary wake-ups.

Relationship with the Ecosystem

KubeElasti does not try to replace HPA or KEDA. Instead, it handles the 0↔1 stage, while scaling from 1 to N is still left to HPA/KEDA. The documentation also mentions optional KEDA pause support to avoid conflicts between ScaledObject behavior and idle scale-down. For observability, KubeElasti exports Prometheus metrics, can be integrated into monitoring via ServiceMonitor, and supports importing Grafana dashboards.

Pricing and Deployment

The source material does not mention commercial pricing or fees for a hosted service. As a CNCF Sandbox project with open collaboration on GitHub, it is better understood as a self-hostable open-source infrastructure component. Deployment involves installing it into a Kubernetes cluster and creating an ElastiService via YAML.

Pros, Cons, and Who It Is For

Its advantages include no need to modify application code, compatibility with existing Ingress and Service Mesh setups, lower costs for low-traffic services, and a better cold-start experience through request queuing. The limitations are also clear: it is mainly designed for HTTP traffic; it depends on Kubernetes, Prometheus, and the ability to configure CRDs; scale-to-zero inherently introduces wake-up latency; and poorly configured thresholds may cause flapping. It is well suited to platform engineering, SRE, and cloud-native teams working on internal tools, low-traffic APIs, test environments, and cost-sensitive microservices.

Access from China

The collected material does not provide information about access from mainland China, mirrors, payments, or commercial support, so china_access can only be marked as unknown. Alternatives worth considering include KEDA, Knative, and HPA when scaling down to 0 is not required.