EKS Fargate capacity is unavailable at this time.

Question

We have been running Fargate EKS for quite some months. Recently, We've been encountering with the errors "Fargate capacity is unavailable at this time.", and pods remains in the pending state. And, this is on everyday basis and very frustrating. Any idea why this is happening (other than no nodes available in that az).

The cluster is in eu-central-1 ( Frankfurt). The EKS Fargate profile is associated with one subnet (eu-central-1a). We are running like 20-40 pods in that profile. Acc. to amazon fargate quota it is capable of running 1000 tasks (max 4000 vCPUs). But with us Fargate even cannot handle this 20 pods at a time. If we would have known this issue, I'd have thought of migrating to the Fargate and Serverless.

Also, I have also tested if I was not running out of IPs for that subnet (256 in total). I tested with replica set of with around 200 replicas. I see that 50 IPs are still available but still some pods are giving the "Fargate out of capacity error".

![Enter image description here](/media/postImages/original/IMa0UtPDzGSImCYQ9wZWZHQg)

Answer

Hi Rajat,

It seems you are being affected by Insufficient Capacity Errors(ICE) for Fargate in eu-central-1 region and specifically in one Availability Zone, eu-central-1a where your subnet resides.

Since you are getting these errors frequently could you follow the below steps in order to help you reduce the risk of Insufficient Capacity Errors:

1.If possible, spread your workloads across all available Availability Zones in a Region.
2. Sharp scale-up waves should be smoothed down where possible/appropriate (for example, a scheduled start/stop workflow which scales all the staging pods at 9AM every day) by spreading requests over a multi-hour time window.
3. Check the [recommendation on this link for troubleshooting EKS pods on fargate](https://repost.aws/knowledge-center/eks-resolve-pending-fargate-pods).

Thanks,
Manish

EKS Fargate capacity is unavailable at this time.

Relevant content