问题
We noticed when we exec -it to connect into a pod, after certain idle time the connection get destroyed. Is there any option to leave the connection open longer?
I see there is an open PR, but wondering if there is any workaround for this issue.
回答1:
The short answer is no. And that's why:
Enabling TCP keepalive for console connections
TCP keepalive is a TCP option that causes packets to be exchanged over a connection even if there is no traffic to transport. It should be enabled on both ends of the connection. TCP keepalive must be enabled at the operating-system level and by the application/program opening TCP connections.
On Linux, edit the "/etc/sysctl.conf" file and add these lines:
net.ipv4.tcp_keepalive_time = 200 net.ipv4.tcp_keepalive_probes = 9 net.ipv4.tcp_keepalive_intvl = 50(feel free to adapt the values as you see fit). When done editing, you must make the new values known to the kernel:
# sysctl --load=/etc/sysctl.conf
Custom Configuration of TCP Socket Keep-Alive Timeouts
Default values for these properties is:
tcp_keepalive_time = 7200 seconds tcp_keepalive_probes = 9 tcp_keepalive_intvl = 75 seconds
The other possible way is to start some kind of proxy server on the client side and connect to Kubernetes apiserver through it. I haven’t tested it myself and it could be tricky, but here is an example of how to enable keepalives to backend for Nginx.
来源：https://stackoverflow.com/questions/50691104/kubectl-command-disconnects-after-few-minutes-of-idle-time