Intermittent "http connection broke" in K8s despite stable Pods and high resources

[dylan-do-vjppro]

I am encountering frequent, random job cancellations in a Kubernetes environment. The job status changes to Canceled with the error message:
Canceled: Job canceled: http connection broke by

this happen usually occurs in less than 1 second of execution time.

Environment Context

• Deployment: Kubernetes (Helm/Manifest)

• Worker Type: TypeScript (Deno)

• Resource Status: * No OOMKilled events in kubectl describe pod.

• High CPU/Memory limits provided (resource starvation is not the cause).

• No Pod restarts or liveness probe failures observed during the error.

• Frequency: Intermittent, but frequent on short-lived flows.

Has anybody faced this issue before?

[rubenfiszel]

Most likely this is either coming from your ingress or the client and http connection is indeed being broken at some point after reaching windmill which result in cancelling the flow