Hello WG Batch,
I’m looking at ways to improve operating Kueue in large multi-cluster environments. The current MultiKueue design has some limitations that make it more likely to unnecessarily disrupt workloads or fragment resources across clusters with how quotas can currently
be defined.
I’m building a list of specific challenges to help determine whether those are accepted limitations of MultiKueue, feasible to address in the current MultiKueue implementation, or cause to revisit how Kueue handles multi-cluster environments.
If you’ve experienced these or any similar problems, please comment or add directly to the doc. I plan to bring this up at the next WG Batch community meeting too if anyone would like to ask questions or leave feedback there.
Thanks,
Jon Huhn