[BUG] return to node view results in replaying all the history states in the passed time range

[thuzhf]

Describe the bug When I am in nodes view at time t1, and create a debug pod and exec into that pod, and after some time when I exit the pod and return back to the nodes view at time t2, k9s will replay every state change within [t1, t2] in the nodes view. I expect k9s just render the latest state in the nodes view.

To Reproduce See above description.

Expected behavior I expect k9s just render the latest state in the nodes view.

Versions (please complete the following information):

• OS: Centos8
• K9s: 0.50.4
• K8s: 1.23.6

[derailed]

@thuzhf @sumansuhag Can you more details here and exact repro steps. I've tried to repro but I am seeing it. Here is what I did:

1. run k9s
2. no
3. Select a node say node1 and press enter to view pods
4. In pod view select pod1 via enter to see the containers.
5. Using the container debug plugin in the repo start a debug container
6. Hang out on the debug container and run a few commands.
7. Exit the shell via exit
8. K9s returns to the containers view
9. Press esc twice to return to node view.

=> Result: node view resume from t2 on.

Q: Are you saying k9s tries to replay all nodes related updates from t1->t2? How did you determine that?

When users shell out of k9s via either pod shell or a plugin. The cluster updates completely stops. The updates resumes once the user returns to k9s (aka at t2->on per your scenario and thus I am surprised that it will try to replay all updates since K9s is no longer tracking those.

What am I missing?

[derailed]

@sumansuhag Thank you for the update! I think I have a fair understanding on how my software works, so this is not what I was looking for.

I am still not grasping how to repro this and what are the symptoms? i.e when you say it replays t1-t2 how do you assert this? Please include all the exact steps to trigger the behavior and what you are seeing in the ui. From what I understand k9s is replaying all the api server state changes from t1 to t2 but based on your description can't seem to find the correct incantation.

[thuzhf]

@derailed I developed a node debugger plugin and can directly open a debug pod from node view, and when I exit the debug pod, I will directly return to the node view. Maybe this is one difference from yours. Another difference is that the debug pod exists for example several days. Also I am not sure whether the k8s version is relevant. I will try to reproduce this in the next days.

[thuzhf]

@derailed I have reproduced the situation. The result is:

1. K8S version is not related: 1.23.6 and 1.33.0 both can reproduce.
2. Long or short duration is not related: 30min or 1day both can reproduce.
3. The key is: return directly from debug pod to node view.

How to reproduce: you can run a debug pod on the selected node directly from node's view (you can use k9s plugin to implement this), and let it last for 10 mins, and then exit from the debug pod. Now you will return directly to the node's view (no intermediate steps). And you can see the node's view changes very quickly replaying all changes within the last 10mins.

[github-actions[bot]]

This issue is stale because it has been open for 30 days with no activity.

[thuzhf]

@derailed Any updates?

[github-actions[bot]]

This issue is stale because it has been open for 30 days with no activity.