(Via Slack conversation in K8s #tilt channel)
@epotex suggested a feature in the Web UI where you can easily jump into a console to debug issues in your running containers, equivalent to kubectl exec:
Usually when I jump into a containers it would be to debug process or network related issues, for example I need to check connectivity to aother service(ping/telnet/nmap/etc) or maybe a response from api with a quick and dirty curl, a real life example from today I had a Django service that was hanging after starting where
ps aux | grep pythonwill show that it listen to the port but netstat/nmap confirm that the service actually is not listening(hang) so I needed to kubectl exec into the container to find all of that
A command-line shorthand, like tilt exec <resource name> would not be helpful to address this workflow:
to me it鈥檚 all boiled down to developers efficiency, that鈥檚 the focus for me as I build our dev tooling - less context switching is better - i鈥檓 not sure that the
tilt execwill be useful since if I鈥檓 already switching to the console i鈥檒l just use thekubectl execso it鈥檚 just going to be a redundant command if it makes sense, thanks for considering it!
Thanks for documenting it!
It would be neat if the command-line UI had this feature as well. I could foresee being frustrated by a web console experience as they always have some issue or another. But being able to get an embedded shell right from within my terminal using my actual terminal to run it would be very nice.
I personally would see value in a tilt exec command as well. A lot of my potential users would definitely prefer to have to look up the full k8s name if I can provide them a consistent shorthand name consistent with what they see in tilt.
I think the possibility to jump from console UI is better than a separate command.
Also, would be great to have possibility to run a shell with a built image instead of jumping inside pod.
Similar to this what could be awesome would be to have the ability to pass in arguments to a specific local_resource from the UI. See comment in slack here
eg ( In the scenario where we wanted to pass a filter into a test runner command we could do that )
Most helpful comment
I think the possibility to jump from console UI is better than a separate command.
Also, would be great to have possibility to run a shell with a built image instead of jumping inside pod.