I use Traefik as an Ingress Controller on the RPI cluster. I have set it using Helm Chart. It's been chugging along fine for the last few months for all my use cases. I'm experiencing a strange issue that I've been unable to resolve, and I hope someone here can shed some light on it.
I have MinIO (Object store similar to AWS S3) sitting behind Traefik Ingress Controller. I have two endpoints.
UI → console.minio.example.com
API → minio.example.com
If I upload a file less than 100KB, the upload works fine on both endpoints. Anything more than 100KB upload gets stuck and gets a Network Error in UI, and read: operation timed out on command line tool (mc). On the command line, I can see it upload 100%, but it fails to complete.
I have seen these configuration options - Options List - Traefik , and have tried most of them to no avail. By the way If I change the MinIO to NodePort, even 1GB files upload fine. So it's definately issue with comms between Traefik & MinIO.
I have tried setting these options, and the issue remains the same.
values:
additionalArguments:
# Timeouts
- "--serverstransport.forwardingTimeouts.responseHeaderTimeout=0"
- "--serverstransport.forwardingTimeouts.idleConnTimeout=0"
- "--entrypoints.websecure.transport.respondingTimeouts.readTimeout=0"
- "--entrypoints.websecure.transport.respondingTimeouts.writeTimeout=0"
# Connection pool
- "--serversTransport.maxIdleConnsPerHost=0" # Unlimited connection pool
- "--entrypoints.websecure.transport.lifeCycle.graceTimeOut=30s"
- "--providers.kubernetesIngress.throttleDuration=0s"
# Disable HTTP/2 which can cause buffering issues
- "--entrypoints.websecure.http.http2.maxConcurrentStreams=0"