is there a way to make a version of this but which works?

doctorpangloss commented 1 month ago

I wonder how to simplify things so that there are fewer surprising interactions

the idea of reuse versus not reuse is very buggy
the messaging around port forwarding is very vague, since there is the ports that the deployment uses versus the ports that the service uses versus the ports that you are ostensibly trying to forward versus the ingress the user typically deploys manually
there are issues with specifying a namespace when it differs from the namespace the KUBECONFIG defaults to
Ctrl+C rarely successfully closes ktunnel
lots of dangling resources when things crash, which prevents gracefully starting ktunnel again
no release for the latest code
vague errors like invalid uuids occur from time to time

doctorpangloss commented 1 month ago

variety of error messages observed:

(1) trying with no deployment or service existing

./bin/windows/amd64/ktunnel.exe expose proxy --node-selector-tags kubernetes.io/os=linux -r 3000:3000 --namespace default
time="2024-08-07T11:24:42-07:00" level=info msg="Exposed service's cluster ip is: 10.152.184.144"
time="2024-08-07T11:24:42-07:00" level=info msg="waiting for deployment to be ready"
time="2024-08-07T11:24:42-07:00" level=info msg="ProgressDeadlineInSeconds is currently 600s. It may take this long to detect a deployment failure."
time="2024-08-07T11:24:43-07:00" level=info msg="deployment \"proxy\" successfully rolled out\n"
time="2024-08-07T11:24:43-07:00" level=info msg="port forwarding to https://appmana-cluster-03-k.appmana.com/api/v1/namespaces/default/pods/proxy-59dd6ff9b7-bq5dz/portforward"
time="2024-08-07T11:24:43-07:00" level=info msg="Waiting for port forward to finish"
time="2024-08-07T11:24:43-07:00" level=info msg="Forwarding from 127.0.0.1:28688 -> 28688\nForwarding from [::1]:28688 -> 28688\n"
INFO[2024-08-07 11:24:43.385] starting tcp tunnel from source 3000 to target localhost:3000
E0807 11:24:43.399449   34272 portforward.go:409] an error occurred forwarding 28688 -> 28688: error forwarding port 28688 to pod f73624df91aebbb4a4c272f7f6cef6ca7d892303d064a994c73f9444bc1140e1, uid : failed to execute portforward in network namespace "/var/run/netns/cni-d22a23d3-c16b-5bb2-b322-9cd147d88a21": failed to connect to localhost:28688 inside namespace "f73624df91aebbb4a4c272f7f6cef6ca7d892303d064a994c73f9444bc1140e1", IPv4: dial tcp4 127.0.0.1:28688: connect: connection refused IPv6 dial tcp6: address localhost: no suitable address found
ERRO[2024-08-07 11:24:43.399] Error sending init tunnel request: rpc error: code = Unavailable desc = connection error: desc = "error reading server preface: EOF"
time="2024-08-07T11:25:06-07:00" level=info msg="Got exit signal, closing client tunnels"

(2) trying after failure observed above

$ ./bin/windows/amd64/ktunnel.exe expose proxy --node-selector-tags kubernetes.io/os=linux -r 3000:3000 --namespace default
time="2024-08-07T11:25:09-07:00" level=info msg="Exposed service's cluster ip is: 10.152.184.144"
time="2024-08-07T11:25:09-07:00" level=info msg="waiting for deployment to be ready"
time="2024-08-07T11:25:09-07:00" level=info msg="ProgressDeadlineInSeconds is currently 600s. It may take this long to detect a deployment failure."

at this stage, hitting Ctrl+C shows time="2024-08-07T11:25:55-07:00" level=info msg="Got exit signal, closing client tunnels" with a hang

$ kubectl logs deployments/proxy
INFO[2024-08-07 18:24:43.406] Starting to listen on port 28688

everything appears to be healthy

(3) trying to recover from failure observed above

$ kubectl delete deployments/proxy services/proxy
deployment.apps "proxy" deleted
service "proxy" deleted

great, everything is deleted. let's rerun

 $ ./bin/windows/amd64/ktunnel.exe expose proxy --node-selector-tags kubernetes.io/os=linux -r 3000:3000 --namespace default
time="2024-08-07T11:27:26-07:00" level=info msg="Exposed service's cluster ip is: 10.152.184.199"
time="2024-08-07T11:27:26-07:00" level=info msg="waiting for deployment to be ready"
time="2024-08-07T11:27:26-07:00" level=info msg="ProgressDeadlineInSeconds is currently 600s. It may take this long to detect a deployment failure."

hmm, but it's healthy:

$ kubectl logs deployments/proxy
INFO[2024-08-07 18:27:27.572] Starting to listen on port 28688

okay, let's Ctrl+C again

time="2024-08-07T11:27:51-07:00" level=info msg="Got exit signal, closing client tunnels"

and now we hang

omrikiei commented 1 week ago

It would require some focus and attention span that I currently don't have, but I have been wanting to refactor and improve this for a while now..

omrikiei / ktunnel