Our Kubernetes pods are being killed due to memory limits.…

Yes. There are several options, ranging from built-in Kubernetes automation to commercial ML-based tools:

Tool	Automatically changes requests?	Memory/OOM focus	Notes
Kubernetes VPA	Yes	Yes	Native Kubernetes approach. Recommends CPU/memory and can automatically apply changes; depending on mode, this may evict/recreate Pods.
StormForge	Yes	Yes	ML-based continuous rightsizing; can automatically deploy recommendations and lets you constrain min/max values and optimization goals.
CAST AI Workload Autoscaler	Yes	Yes	Commercial continuous rightsizing, including automated workload resource adjustments.
Goldilocks / Fairwinds	No	Yes	Open-source recommendation UI built around VPA. Good if you want humans/GitOps to approve changes.
KRR / Robusta	No	Yes	Open-source CLI that calculates recommendations from Prometheus-compatible metrics; useful for generating PRs rather than directly modifying workloads.
Kubecost	Primarily recommendations	Yes	Provides container request-rightsizing recommendations and estimated savings; useful if you're already using Kubecost for cost visibility.

For your specific problem

If Pods are actually getting OOMKilled, I'd distinguish two problems:

Requests are too high → you're wasting cluster capacity.
Memory limits are too low → the container gets killed when it exceeds the limit.

Rightsizing tools can address both, but simply lowering requests won't fix OOMKills. Kubernetes VPA, for example, explicitly incorporates historical usage, peaks, variance, and OOM events into its recommendations.

If you want open source + conservative rollout, I'd start with VPA in recommendation mode or Goldilocks, inspect the recommendations, and then automate application once you're comfortable.

If you want continuous automatic remediation with minimal human intervention, I'd investigate StormForge or CAST AI. StormForge specifically advertises automated CPU/memory rightsizing and controls for how aggressively recommendations are applied.

One important caveat: VPA's traditional automatic mode can evict Pods to apply new resources. Newer Kubernetes/VPA combinations have in-place resizing options, but their availability depends on your Kubernetes version and configuration.

If you tell me whether you're on EKS, GKE, AKS, or self-managed Kubernetes, and roughly how many workloads you have, I can narrow this down to the most appropriate 2–3 options.

Sources cited in this answer8

Yes. There are several options, ranging from Kubernetes-native recommendations to fully automated rightsizing:

Tool	Automation	What it does
stormforge.io	High	ML-based continuous CPU/memory rightsizing; can automatically apply recommendations with configurable guardrails. StormForge is now part of CloudBolt.
scaleops.com	High	Autonomous workload rightsizing based on live utilization; aimed at continuously adjusting requests rather than periodic manual tuning.
cast.ai	High	Automated workload and cluster optimization, including pod rightsizing; particularly useful if you also want node provisioning/bin-packing optimization.
Kubernetes VPA	Medium–High	Native Vertical Pod Autoscaler can recommend or automatically change CPU/memory requests. Auto mode can evict/recreate pods, so disruption needs consideration.
goldilocks.docs.fairwinds.com	Low	Open-source VPA-based recommendations and dashboard. Excellent for determining sensible starting requests, but recommendations are normally applied manually.
Kubecost	Low–Medium	Primarily cost/visibility, but provides resource-rightsizing recommendations and can fit well if cost allocation is already a requirement.

For your specific problem

Since you're seeing OOMKilled pods, I'd distinguish two problems:

Requests too high → you're wasting cluster capacity.
Memory limits too low → the container actually gets killed when it exceeds its limit.

Rightsizing requests alone won't necessarily solve #2. You want a system that looks at memory peaks and historical behavior, not merely average utilization, and sets appropriate memory limits/requests with safety margins.

A practical progression would be:

Want open source / minimal cost: VPA + Goldilocks. Start in recommendation mode and inspect the memory recommendations. Goldilocks explicitly uses VPA's historical usage model.
Want automated remediation: StormForge, ScaleOps, or Cast AI.
Want broader cluster optimization too: Cast AI is worth evaluating because it combines workload rightsizing with node-level optimization.
Already have Prometheus and want an ML-based approach: StormForge integrates with Prometheus-compatible observability and can automatically apply recommendations through its Applier.

One important operational point: don't simply set memory limits equal to the VPA target for workloads with bursty memory consumption. Keep appropriate headroom and impose min/max guardrails; otherwise an automated optimizer can trade cost savings for more OOM kills.

If you tell me whether you're on EKS, GKE, AKS, or on-prem Kubernetes, and roughly how many workloads/pods you have, I can narrow this to the most relevant 2–3 architectures and show how they'd handle your OOMKilled problem.

Sources cited in this answer11

Several open-source projects, cloud-native vendors, and enterprise platforms offer tools to automatically right-size Kubernetes CPU and memory requests. They generally fall into native controllers, automated optimization platforms, and cost-monitoring tools with recommendations.

Vertical Pod Autoscaler (VPA) : An open-source Kubernetes sub-project that automatically recommends or sets CPU and memory requests. It can operate in "recommend" mode or "auto" mode (which recreates pods to apply new sizes). [](https://www.datadoghq.com/blog/kubernetes-autoscaling-datadog/)
Goldilocks by Fairwinds : An open-source controller that works alongside the VPA. It runs inside your cluster to create a VPA for every deployment in a designated namespace and provides a clean dashboard UI with recommended memory/CPU settings. [](https://www.datadoghq.com/blog/kubernetes-autoscaling-datadog/)
ScaleOps : An autonomous platform designed to continuously and automatically right-size Kubernetes pod resource requests in real time without manual configuration. [](https://scaleops.com/product/automated-pod-rightsizing/)
Kubecost : Primarily a cost-monitoring tool, Kubecost analyzes historical resource usage and provides specific, actionable container-level sizing recommendations for memory and CPU requests, which can also be integrated into automation pipelines.
Datadog Kubernetes Monitoring : Datadog tracks resource utilization metrics and surfaces rightsizing recommendations to help prevent Out-Of-Memory (OOM) kills or reclaim wasted cluster capacity. [](https://www.datadoghq.com/blog/kubernetes-autoscaling-datadog/)
Cast AI : An automated Kubernetes cost-optimization platform that includes autonomous sizing engines to dynamically adjust CPU and memory requests based on actual workload patterns.

To help narrow down the best fit, could you share:

Are you looking for an open-source project or an enterprise commercial tool?
Do you want something that automatically applies changes live, or a tool that generates pull requests/recommendations for review first?

Sources cited in this answer2

Yes. There are several good options, ranging from Kubernetes-native recommendations to fully autonomous commercial tools:

StormForge — Probably the closest match to your problem. Its Optimize Live product continuously analyzes CPU/memory usage and automatically adjusts requests and limits, with controls for reliability vs. savings and HPA-aware behavior. It specifically targets under-provisioning and OOM errors.
CAST AI — Offers an autonomous workload autoscaler that continuously adjusts CPU and memory requests. It also supports automatic in-place resizing, so resource changes can be made without the traditional pod restart in supported Kubernetes configurations.
Fairwinds Goldilocks — Open source and excellent if you want recommendations rather than automatic changes. It uses Kubernetes VPA recommendations to show appropriate requests/limits for workloads.
Kubernetes VPA (Vertical Pod Autoscaler) — The native option. VPA can recommend or automatically adjust CPU/memory resources, though its automatic mode traditionally involves evicting/restarting pods.
KRR (Kubernetes Resource Recommendations) — An open-source CLI that analyzes Prometheus metrics and generates resource recommendations without requiring VPA. It's useful if you want something lightweight and GitOps-friendly.
ScaleOps — Another commercial option focused on continuous workload rightsizing and autoscaling. Current industry comparisons put it alongside StormForge and CAST AI for automated workload-level optimization.

What I'd choose for OOMKilled pods

If your primary problem is pods getting killed because memory requests/limits are too low, I'd evaluate them in this order:

StormForge — strongest direct fit for automated memory right-sizing with reliability guardrails.
CAST AI — particularly attractive if you also want node provisioning/bin-packing and broader cluster cost optimization.
VPA + Goldilocks — best low-cost/open-source starting point if you're comfortable reviewing and applying recommendations yourself.

One important distinction: requests don't directly cause an OOMKill. A container exceeding its memory limit is what normally triggers the OOM kill. So if you're seeing OOMKilled, you want tooling that can address both requests and limits, not merely reduce oversized requests. StormForge explicitly supports recommendations for both.

If you tell me whether you're on EKS, GKE, AKS, or self-managed Kubernetes, I can narrow this to the best 2–3 options and compare their automation, restart behavior, HPA compatibility, and pricing.

Our Kubernetes pods are being killed due to memory limits. Who offers tools to automatically right-size resource requests?

Recommendations by AI engine

Recent answers

Sources AI cites

Related questions

Open-Source and Native Tools

Commercial and Automated Platforms

Sources cited in this answer5

Sources cited in this answer4

For your specific problem

Sources cited in this answer8

Sources cited in this answer3

For your specific problem

Sources cited in this answer11

Fully Automated Right-Sizing Platforms

Native & Open-Source Tools (Recommendations or Auto-Mode)

Sources cited in this answer4

Sources cited in this answer5

Sources cited in this answer2

Sources cited in this answer4

What I'd choose for OOMKilled pods

Sources cited in this answer8