kubectl logs -f immediately returns instead of waits for container death

[antoinetran]

Short Description of the issue

When doing kubectl logs [pod] -f, it returns the logs and dies immediately, even if the main container is still running. It should not return and continues to show logs until it dies.

Environment

• Operating System:
• Other related components versions:

Steps to reproduce

Logs, stacktrace, or other symptoms

output

Summary of proposed changes

[antoinetran]

I don't know if there is a more elegant solution than doing a regular sleep of 2s and check if the container died, in follow mode. Or with inotify , we can avoid sleep.

[antoinetran]

It took me a bit of time to understand how the logs are returned to virtual kubelet. I created an issue at virtual-kubelet because something needs to be done there first.

[dciangot]

@antoinetran as anticipated via chat, I think it’s matter of implementing a stream response in LogRetrieval function in execute.go file in interlink

can you cross check and see if that makes sense to you?

[antoinetran]

@antoinetran as anticipated via chat, I think it’s matter of implementing a stream response in LogRetrieval function in execute.go file in interlink

can you cross check and see if that makes sense to you?

I am trying a few things, it will take some time, I am a bit slow in GO :)

[antoinetran]

I tested a bit the GO http stream, and I think it is a good solution. The HTTP server can send the HTTP status code , and write the body continuously. The client can get the HTTP status, and read the HTTP body, it hangs until the server stops. Then the client gets EOF at the read.