Skip to content

Conversation

@effi-ofer
Copy link

The context.ctx file is copied only if it does not exist. But it could be stale. Why not always copy it?

@maugustosilva
Copy link
Collaborator

I have faced weird timing issues when copying ~/.kube/config to $LLMDBENCH_CONTROL_WORK_DIR/environment/context.ctx and that is why I didn't adopt the "always copy" approach. That being said, you are absolutely correct that we cannot afford to have a stored context.ctx which points to the "incorrect" cluster.

@effi-ofer
Copy link
Author

effi-ofer commented Nov 11, 2025

Both my team mate and I independently encountered a stale context file yesterday. So it's a real problem. In my case it happened because I switched LLMDBENCH_CONTROL_WORK_DIR to a directory that was previously used as the work dir and contained a stale context file, so you could argue it was my fault. But I think the change avoids the problem all together.

@maugustosilva
Copy link
Collaborator

You're right, we need to attempt to minimize it. Please take a look at the state of the code now (after #505 was merged) and let me know if it fixes your issues.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants