Migrate to Karpenter without the risk
Sedai translates your EKS Managed Node Groups into Karpenter NodePools, then cuts over one pool at a time. Every step is health-checked, and if validation fails, your workloads roll back automatically.
Animation: three EKS Managed Node Groups, running on nine identical fixed-size nodes, are migrated to Karpenter one node pool at a time. Each pool is cordoned, drained onto right-sized Karpenter nodes, and validated before the next one starts.
Karpenter is the better autoscaler. Getting there is the hard part.
Karpenter picks the cheapest instance that fits from any type or size, and consolidates onto fewer nodes on your schedule. So why do most production EKS clusters still run Cluster Autoscaler?
Two different languages
Node groups and NodePools don’t map one-to-one. Every setting gets translated by hand, and anything missed shows up in production.
Failures land on you
If a cutover goes sideways, someone has to notice, diagnose, and restore a safe state, under pressure, often at 2 a.m.
It’s always next quarter
A risky, hands-on migration keeps losing to feature work. So it slips, sprint after sprint, and the savings Karpenter promises never arrive.
A migration that checks its own work
Sedai's Karpenter Migration Assistant handles the two things teams underestimate: translating the configuration, and cutting over safely.
Read and translate
Sedai reads each node group’s configuration and generates the equivalent Karpenter NodePool.
Surface every difference
Settings with no safe Karpenter equivalent block the migration instead of being silently dropped. Minor changes are logged, so nothing surprises you later.
Approve and schedule
Run it in your maintenance window. Reschedule or cancel any time before it starts.
Cut over in reversible phases
Cordon, drain, validate, with cluster health checked at every step. If validation fails, workloads land back on your original nodes automatically.
Move one node pool at a time
No big-bang cutover. Migrate your lowest-risk pools first, keep the rest on Cluster Autoscaler, and finish when you’re ready. Both run side by side until then.
Doing it yourself vs. migrating with Sedai
Same destination. Very different odds of a quiet weekend.
| Doing it yourself | Sedai |
|---|---|
Hand-map every node group setting to Karpenter | Generates NodePools, surfaces every difference |
Unsupported settings surface after cutover | Unsupported settings block the migration up front |
Scripted cutovers, watched manually | Phased cordon, drain, and validate with health checks |
Manual diagnosis and rollback under pressure | Rolls back automatically if validation fails |
Migrates whenever the team can staff it | Runs in your scheduled maintenance window |
Then make Karpenter work harder
Karpenter decides how many nodes you need and which type is cheapest. Sedai decides which pod goes where, and keeps the whole cluster tuned to real workload behavior.
Bin Packing & Pod Placement
Smart Scheduler places pods on the nodes that fit them best, layered on top of Karpenter's provisioning
Pod Rightsizing
Smaller, accurate requests mean Karpenter provisions fewer, smaller nodes.
Karpenter Optimization
Keep NodePools tuned to how your workloads actually behave after the move
See it on your own clusters
Walk through a migration plan for your EKS node groups with a Sedai engineer.
Frequently asked questions
Currently, Sedai's Karpenter Migration Assistant supports Amazon EKS, migrating from EKS Managed Node Groups. Karpenter’s ecosystem is most mature on AWS, so that’s where the automation starts.
No. You migrate node pool by node pool. Pools you’ve migrated run on Karpenter while the rest stay on Cluster Autoscaler until you’re ready.
Sedai rolls back automatically, returning your workloads to your original infrastructure. No one has to diagnose and recover by hand.
Anything without a safe Karpenter equivalent stops the migration before it starts, so nothing is silently dropped. Cosmetic or non-essential differences are recorded so you know exactly what changed.
Yes. Schedule it for your maintenance window, and reschedule or cancel any time before it starts.
Sedai right-sizes pod requests so Karpenter provisions for real demand, and Sedai Scheduler packs pods densely at placement, so you run fewer, cheaper nodes that are also better packed.