PricingPricingDocsDocsBlogBlogBook a demoBook a demoCompanyCompany (opens in a new tab)
Product
DocsAPI ReferencePricingConsole DemoBlogCompany
Start Freev3.4.7
Wave Karpenter

Node Warmup

Starts a node before the next Pod needs one. In one Wave benchmark run, the first new Pod was ready in 3 seconds instead of 31.

The challenge

The new Pod sits in Pending for 31 seconds while a node starts

Wave 3.1.0 release post (one benchmark run, first new Pod); Wave Node Warmup docs; karpenter.sh v1.14, NodeClaims and CapacityBuffers.

By default, Karpenter launches a node only after a Pod is Pending and no node has room for it. The instance has to launch, register as a Node, and initialize before the Pod can start (karpenter.sh).

In one Wave benchmark run on EKS, the first new Pod waited 31 seconds. The Node Warmup docs estimate the split: about 15s to launch the node, 5s for kubelet registration, 10s to pull the image, and 1s to schedule.

Karpenter can keep spare room too, with CapacityBuffers (an alpha feature) or a static NodePool. A CapacityBuffer holds room with placeholder Pods that are never created as real Pods, so the workload images are not pulled in advance (karpenter.sh).

The first new Pod: 31 seconds, then 3 seconds, in one Wave benchmark runWithout WaveWith Wavefirst new Pod: 31sfirst new Pod: 3s
Without Wave
31 seconds until the first new Pod is readyprovisioning ~15skubelet ~5simage pull ~10sscheduling ~1sPending0s10s20s30s
In one Wave benchmark run, the first new Pod was ready after 31 seconds. By default, Karpenter launches a node only after a Pod is Pending. The docs estimate the split: launch ~15s, kubelet ~5s, image pull ~10s, scheduling ~1s.
With Wave
3 seconds until the first new Pod is readyfirst new Pod in 3sthe node has already joined0s10s20s30s
In one Wave benchmark run with Node Warmup, the first new Pod was ready after 3 seconds. A warmup Pod made Karpenter launch the node before it was needed. When image pre-pull is on, the images are already on the node.

31s and 3s: the first new Pod in one Wave benchmark run (Wave 3.1.0 release post; EKS, c5n.xlarge and c6i.xlarge Spot, ap-northeast-1). The split of the 31s is an estimate from the Wave Node Warmup docs.

How Wave does it

From a Pending Pod to a running one
Karpenter by defaultFirst new Pod: 31s, one benchmark run
  1. Pod
    Pending: no node has room for it
  2. ↓launches→
    NodeClaim
    Launched: the instance is created
  3. ↓kubelet joins→
    Node
    Registered and initialized as a Node
  4. ↓places→
    kubelet
    pulls the image, then starts the container
Wave Node WarmupFirst new Pod: 3s, one benchmark run
  1. Wave Core
    finds no node that fits the largest Pod plus a 10% buffer
  2. ↓no room→
    Warmup Pod
    makes Karpenter launch a node now
  3. ↓node joins→
    Warm node
    consolidation paused 10 min; images pulled when pre-pull is on
  4. ↓places→
    New Pod
    starts on a node that has already joined
Pending, and NodeClaim Launched, Registered, Initialized: karpenter.sh v1.14 docs. 31s and 3s: the first new Pod in one Wave benchmark run (Wave 3.1.0 release post). Largest Pod, 10% buffer, 10-minute consolidation pause, and image pre-pull (off by default): Wave Node Warmup docs.
1

Find the gap

Wave Core tracks the largest Pod in each NodePool. When no node can fit it plus a 10% buffer (the default), Node Warmup starts a node.

2

Start a node early

A warmup Pod makes Karpenter launch a node now. Max Pod mode runs always or on a cron schedule. Trigger mode starts a set number of nodes of a given size on a cron schedule.

3

Keep the node ready

While warmup runs, Karpenter consolidation is paused for the NodePool, 10 minutes by default, and warmed nodes carry the do-not-disrupt annotation. When image pre-pull is on, Wave pulls the workload images onto the node in advance. Pre-pull is off by default.

Results

3s
Wave Node Warmup
31s
Karpenter by default

until the first new Pod was ready, in one Wave benchmark run

Wave 3.1.0 release post: one benchmark run on EKS, first new Pod only

Related in Wave Karpenter

Karpenter Dashboard →Spot Workload Placement →Read the docs →
Book a demoStart free
Product
Wave GPUAI Infrastructure Professional ServiceWaveWave AI Agent2026 Q4
Resources
DocsAPI ReferenceBlogConsole DemoPricing
Company
STCLabBook a demoContact
STCLab© 2026 Wave All rights reserved.
Terms of ServicePrivacy PolicyCookies