Error with k8sPortForwardDialer dial: Upgrade request required

[mausch]

Describe the bug

$ telepresence connect

Launching Telepresence Root Daemon
Launching Telepresence User Daemon
telepresence connect: error: connector.Connect: the port-forward connection to the traffic manager timed out.  The current timeout 1m0s can be configured as "timeouts.trafficManagerConnect" in "/home/mauricio/.config/telepresence/config.yml"

$ telepresence loglevel debug
telepresence loglevel: error: connector.Connect: the port-forward connection to the traffic manager timed out.  The current timeout 1m0s can be configured as "timeouts.trafficManagerConnect" in "/home/mauricio/.config/telepresence/config.yml"

$ telepresence gather-logs
error getting logs from kubernetes components: rpc error: code = Unavailable desc = no active session

These are all the logs I have:

connector.log:

2023-11-07 09:21:24.0528 info    ---
2023-11-07 09:21:24.0528 info    Telepresence Connector v2.10.4 (api v3) starting...
2023-11-07 09:21:24.0528 info    PID is 3617605
2023-11-07 09:21:24.0528 info    
2023-11-07 09:21:24.0537 info    connector/server-grpc : gRPC server started
2023-11-07 09:21:24.2810 info    connector/session : -- Starting new session
2023-11-07 09:21:24.2811 info    connector/session : Connecting to k8s cluster...
2023-11-07 09:21:24.4800 info    connector/session : Server version v1.28.2+k3s1
2023-11-07 09:21:24.4801 info    connector/session : Context: loft-test-user
2023-11-07 09:21:24.4801 info    connector/session : Server: https://vcluster.elevate.direct/kubernetes/project/default/virtualcluster/vcluster-test-user
2023-11-07 09:21:24.4802 info    connector/session : Connected to context loft-test-user (https://vcluster.elevate.direct/kubernetes/project/default/virtualcluster/vcluster-test-user)
2023-11-07 09:21:24.6941 info    connector/session : Connecting to traffic manager...
2023-11-07 09:21:25.1745 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:21:26.5251 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:21:28.1803 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:21:30.7111 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:21:34.3111 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:21:41.8341 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:21:54.6083 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:22:09.9002 error   connector/session : Error with k8sPortForwardDialer dial: Upgrade request required
2023-11-07 09:22:24.6948 error   connector/session : Unable to connect to session: the port-forward connection to the traffic manager timed out.  The current timeout 1m0s can be configured as "timeouts.trafficManagerConnect" in "/home/mauricio/.config/telepresence/config.yml"

daemon.log:

2023-11-07 09:21:23.8039 info    ---
2023-11-07 09:21:23.8040 info    Telepresence daemon v2.10.4 (api v3) starting...
2023-11-07 09:21:23.8040 info    PID is 3617592
2023-11-07 09:21:23.8040 info    
2023-11-07 09:21:23.8045 info    daemon/server-grpc : gRPC server started

Versions (please complete the following information):

• Output of telepresence version

OSS Client     : v2.10.4
OSS Root Daemon: v2.10.4
OSS User Daemon: v2.10.4
Traffic Manager: not connected

• Operating system of workstation running telepresence commands: Ubuntu 22.04.03
• Kubernetes environment and Version [e.g. Minikube, bare metal, Google Kubernetes Engine]: v1.28.2+k3s1

[mausch]

This was caused by a AWS ELB that was configured on HTTP instead of TCP.
i.e. set "service.beta.kubernetes.io/aws-load-balancer-backend-protocol": "tcp" in the nginx ingress controller config.

Telepresence should catch this "Upgrade request required" error and show a more specific message e.g. at least "Could not establish a websocket connection"

[thallgren]

Yeah, maybe. Then again, a google search for "Upgrade request required" very quickly reveals what the problem is. A search for "Could not establish a websocket connection", not so much. This is why we, in general, try to convey the errors that the Kubernetes tooling gives us.

[mausch]

Telepresence could wrap the error message. This would make the error both googlable and more immediately actionable.