I’m wondering if using a bursty limit like requests: cpu: 100m limits: cpu: 200m Would work better? If the problem is that you are getting

I’m wondering if using a bursty limit like requests: cpu: 100m limits: cpu: 200m Would work better? If the problem is that you are getting throttled at a lower rate than your specified limit, maybe bumping that would help. But you still get to use your target “request” for bin packing / node resource tracking. This would depend on the throttle level being proportional to the specified limit and not something orthogonal like number of processes - but if you don’t want to turn off limits entirely it might at least help.