oc debug should not use AttachablePodForObject

[smarterclayton]

Debug doesn't work against attachable pod for object because a scaled to zero Deployment (required for us to be able to debug and mount an RWO) hangs forever.

We should be using a different method MostAccuratePodTemplateForObject which should try to go as far down as it can and then return back a pod template, and if there are zero pods return the RS template

[smarterclayton]

Also, oc get deploy/prometheus -o yaml | oc debug --one-container -c prometheus -f - fails.

Basically, I need to be able to debug from an approximate pod template.

[smarterclayton]

It is not appropriate to change AttachablePodForObject (a template is not attachable).

[smarterclayton]

And this is actually a pretty common use case - you want to debug something with RWO. You scale it to zero (because otherwise you have to fight for the RWO). You then want to run debug on it.

[wanghaoran1988]

@smarterclayton @fabianofranz How do you think about just remove the AttachablePodForObject logic here and use the t.spec.Template directly? I am not very clear why we need carbon copy an existed pod instead of using the t.spec.Template, would you please tell me the logic behind this ?

[smarterclayton]

Debug wants to get the closest possible pod template to what is actually running. It does not require a running pod, it just prefers one if possible.

…

On Tue, Jul 11, 2017 at 5:53 AM, Haoran Wang ***@***.***> wrote: @smarterclayton <https://github.com/smarterclayton> @fabianofranz <https://github.com/fabianofranz> How do you think about just remove the AttachablePodForObject logic here <https://github.com/openshift/origin/blob/master/pkg/cmd/util/clientcmd/factory.go#L208-L214> and use the t.spec.Template directly? I am not very clear why we need carbon copy an existed pod instead of using the t.spec.Template, would you please tell me the logic behind this ? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14915 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABG_p4geiOae4kAee0cVAJkxyJsnbXERks5sM0YXgaJpZM4OG5sF> .

[wanghaoran1988]

Add MostAccuratePodTemplateForObject to upstream ObjectMappingFactory, so we can always get the pod template of the upstream resources.

Also, oc get deploy/prometheus -o yaml | oc debug --one-container -c prometheus -f - fails.
Basically, I need to be able to debug from an approximate pod template.

For this kind debug, template from stdin, I am try to get it work, but I found if we read the template from stdin, the debug pod can be created successfully but the tty tunnel cannot be established, as the code here don't think stdin is a terminal.

[smarterclayton]

Debug with piped input should only succeed if the user passed a command *AND* set --no-tty

…

On Thu, Jul 13, 2017 at 5:54 AM, Haoran Wang ***@***.***> wrote: Add MostAccuratePodTemplateForObject to upstream ObjectMappingFactory, so we can always get the pod template of the upstream resources. Also, oc get deploy/prometheus -o yaml | oc debug --one-container -c prometheus -f - fails. Basically, I need to be able to debug from an approximate pod template. For this kind debug, template from stdin, I am try to fix, but I found if we read the template from stdin, the pod can be created successfully but the tty tunnel cannot be established, as the code here <https://github.com/kubernetes/kubernetes/blob/master/pkg/kubectl/cmd/exec.go#L235> don't think stdin is a terminal. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14915 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABG_p6OMsLow9DF0ybMNw6cY32OW3dX_ks5sNelBgaJpZM4OG5sF> .