Savings plan support
**Tell us about your request** Build knowledge of AWS Savings Plans into Karpenter, so that it uses OnDemand capacity up to a certain limit, then tries spot as appropriate from there **Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?** We "reserve" a certain amount of capacity in our cluster as OnDemand, in order to ensure a baseline amount of capacity in our clusters, without the risk of spot terminations causing a lack of capacity. The rest of our capacity we'd like to at least attempt to run on spots, then fallback to OnDemand if those aren't available. **Are you currently working around this issue?** We're not dealing with this directly at the moment, but to workaround we'd probably just have two deployments - one that's OnDemand-only to use up the savings plan, and another that allows both Spot and OD. But then we need to manage multiple deployments, plus with different instance types and prices it's quite a manual process to keep a track of the utilisation of the SP by the "OnDemand" deployment <!-- Please keep this note for the community --> ### Community Note * Please vote on this issue by adding a 👍 [reaction](https://blog.github.com/2016-03-10-add-reactions-to-pull-requests-issues-and-comments/) to the original issue to help the community and maintainers prioritize this request * Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request *
Posted by spmason on Aug 8, 2022, 10:06 AM
Tags
Related issues
- The only thing keeping me on VSCode these days is the remote SSH development.
- Something similar was done in a sheep model: "A bioactive supramolecular and covalent polymer scaffold for cartilage repair in a sheep m...
- Proposal for OpenHarmony Adaptation of [ ThreadPool ]
- Proposal for a pipe-like operator to chain constructor/method invocations without nesting
- Implement support for Node Stashing
- RSC and CDN interaction makes next.js inefficient for highload projects
Raise real-world problems, find people facing the same issue, and discuss solutions.2,500 issues0 comments
support (E-commerce)
The only thing keeping me on VSCode these days is the remote SSH development.
Something similar was done in a sheep model: "A bioactive supramolecular and covalent polymer scaffold for cartilage repair in a sheep m...
Proposal for OpenHarmony Adaptation of [ ThreadPool ]
Proposal for a pipe-like operator to chain constructor/method invocations without nesting
Implement support for Node Stashing
RSC and CDN interaction makes next.js inefficient for highload projects
Extract Action Notifier framework for push notifications
Improve support for monorepo by supporting git options such as depth, and sparse checkout
Savings plan support
**Tell us about your request** Build knowledge of AWS Savings Plans into Karpenter, so that it uses OnDemand capacity up to a certain limit, then tries spot as appropriate from there **Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?** We "reserve" a certain amount of capacity in our cluster as OnDemand, in order to ensure a baseline amount of capacity in our clusters, without the risk of spot terminations causing a lack of capacity. The rest of our capacity we'd like to at least attempt to run on spots, then fallback to OnDemand if those aren't available. **Are you currently working around this issue?** We're not dealing with this directly at the moment, but to workaround we'd probably just have two deployments - one that's OnDemand-only to use up the savings plan, and another that allows both Spot and OD. But then we need to manage multiple deployments, plus with different instance types and prices it's quite a manual process to keep a track of the utilisation of the SP by the "OnDemand" deployment <!-- Please keep this note for the community --> ### Community Note * Please vote on this issue by adding a 👍 [reaction](https://blog.github.com/2016-03-10-add-reactions-to-pull-requests-issues-and-comments/) to the original issue to help the community and maintainers prioritize this request * Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request *
Comments (0)
No comments yet. Start the discussion.