wmgroot

Here are some notes I've taken as I've explored this particular problem space. Hopefully this is helpful as background to anyone else trying to understand the problem. I'd ultimately prefer...

Making it configurable just covers the case where someone might need to increase it in the future. I agree that 30s is likely much longer than it needs to be.

To check my understanding of `disruption.costPercent: 10`, you're effectively giving users a way to say that "I'm losing 10% of what I'm paying for this node, so make sure the...

Additionally, it could potentially support fixed or percent values via `IntOrString`, such as: ``` NodeClaim.spec.disruptionCost: 10% NodeClaim.spec.disruptionCost: 20 (effectively saying a one-time disruption of this node would cost $20) ```...

> Given this, how would you expect it to interact with drift? I expect drift to not take this feature into account at all. If the node needs to be...

We've also run into this problem and have been considering what could be done to address it. One alternative to consider to updating the allocation calculation dynamically is to assume...

No worries. I definitely wasn't hoping that the default behavior would change, and definitely understand that there are probably users out there who care a lot about maximizing the number...

This is stalled waiting for maintainer input. At my company we think * This is a serious issue that could impact a wide berth of cluster-autoscaler users (including CA +...

We eventually narrowed this issue down to our use of the `rootVolume: true` flag for these EC2NodeClasses. The example recommends using a 4GB volume for the OS and a second...

This can be closed as resolved, though updating the documentation to be clear about this might be helpful for users transitioning their NodePools to bottlerocket.