0

Empirically, it looks like kubectl exec and ssh behave differently on disconnects of the client.

I performed the following experiments.

SSH

# Terminal 1
ssh <random_host_in_aws> 
bash -c 'sleep 600000'

Terminal 2
kill -9 <pid_of_ssh_process>
ssh <random_host_in_aws> 
ps aux | grep sleep
# No sleep processes

Kubectl Exec

# Terminal 1
kubectl exec ...
bash -c 'sleep 600000'

# Terminal 2
kill -9 <pid_of_kubectl_process>
kubectl exec ...
ps aux | grep sleep
# Sleep process is still around.

Is there some magic configuration in the Kubernetes cluster or client that will force it to behave the same way as ssh in the face of client disconnects?

0

You must log in to answer this question.

Browse other questions tagged .