Exec-ing into pod not working - "Server lost session"

[dagobertdocker]

What happened?

When trying to initiate new terminal session in pod from dashboard, the terminal screen stays black, and a message is displayed "Server lost session".
In the logs, I see

[kubernetes-dashboard-api] 2024/03/11 13:59:01 handleTerminalSession: can't Recv: sockjs: session not in open state

What did you expect to happen?

Interactive terminal session established and shell prompt visible.

How can we reproduce it (as minimally and precisely as possible)?

1. Deploy Kubernetes Dashboard via helm chart
2. Navigate to a pod details page
3. Click "Exec into pod"
4. Observe "Server lost session" error message

Anything else we need to know?

No response

What browsers are you seeing the problem on?

Firefox

Kubernetes Dashboard version

7.1.2

Kubernetes version

1.26.9

Dev environment

No response

[floreks]

I could not reproduce it based on the provided steps. Tested on Firefox/Chrome and exec works just fine.

[dagobertdocker]

I upgraded to k8s 1.28.7, and since then, the problem is gone.

[dagobertdocker]

For future reference:

I am using a setup with HAProxy -> NGINX Inc's Ingress -> gogatekeeper (OAuth2 proxy) -> Kong service of Dashboard.
In the end, I had to whitelist the /api/sockjs path in gogatekeeper (it seems to remove auth headers) to get it to work consistently..

[shnigam2]

I have this issue on 1.28.5 EKS as well

[dagobertdocker]

Correction, it does not work consistently, but randomly. So, this issue is open and being tracked in #8838 I assume..