Keda operator does not install on AWS EKS Fargate

[gildas]

Report

Deploying KEDA on AWS EKS Fargate fails with Keda pods complaining anout untolerated taint (eks.amazonaws.com/compute-type: fargate).

Expected Behavior

Keda pods to run.

Actual Behavior

Keda pods stay at Pending with this kind of events:

   Type     Reason            Age   From               Message                                                                                                                                                                                                                                                      
   ----     ------            ----  ----               -------                                                                                                                                                                                                                                                      
   Warning  FailedScheduling  8s    default-scheduler  0/X nodes are available: X node(s) had untolerated taint {eks.amazonaws.com/compute-type: fargate}. preemption: 0/X nodes are available: X Preemption is not helpful for scheduling.                                                                     │

Steps to Reproduce the Problem

1. Run an AWS EKS Fargate cluster
2. install Keda with helm install --namespace keda --ceate-namespace --set clusterName=<my-cluster> keda kedacore/keda
3. Watch the pods

Logs from KEDA operator

No logs available from Keda as it does not install

KEDA Version

2.16.0

Kubernetes Version

1.30

Platform

Amazon Web Services

Scaler Details

No response

Anything else?

No response

[omerap12]

I think you just need to add toleration https://github.com/kedacore/charts/blob/main/keda/values.yaml#L508

[gildas]

I think you just need to add toleration https://github.com/kedacore/charts/blob/main/keda/values.yaml#L508

I already tried that while getting some help from the community on slack (https://kubernetes.slack.com/archives/CKZJ36A5D/p1730281617416029?thread_ts=1729833857.363939&cid=CKZJ36A5D)

in short, I get another error:

Too many pods. preemption: 0/X nodes are available: X No preemption victims found for incoming pod

Which makes sense, since on AWS EKS Fargate, 1 pod == 1 node. This is what allows them to scale "magically".