-
Notifications
You must be signed in to change notification settings - Fork 1.7k
DRA: Native Resource Requests #5517
Copy link
Copy link
Open
Labels
lead-opted-inDenotes that an issue has been opted in to a releaseDenotes that an issue has been opted in to a releasesig/nodeCategorizes an issue or PR as relevant to SIG Node.Categorizes an issue or PR as relevant to SIG Node.sig/schedulingCategorizes an issue or PR as relevant to SIG Scheduling.Categorizes an issue or PR as relevant to SIG Scheduling.stage/alphaDenotes an issue tracking an enhancement targeted for Alpha statusDenotes an issue tracking an enhancement targeted for Alpha statustracked/yesDenotes an enhancement issue is actively being tracked by the Release TeamDenotes an enhancement issue is actively being tracked by the Release Teamwg/device-managementCategorizes an issue or PR as relevant to WG Device Management.Categorizes an issue or PR as relevant to WG Device Management.
Milestone
Metadata
Metadata
Assignees
Labels
lead-opted-inDenotes that an issue has been opted in to a releaseDenotes that an issue has been opted in to a releasesig/nodeCategorizes an issue or PR as relevant to SIG Node.Categorizes an issue or PR as relevant to SIG Node.sig/schedulingCategorizes an issue or PR as relevant to SIG Scheduling.Categorizes an issue or PR as relevant to SIG Scheduling.stage/alphaDenotes an issue tracking an enhancement targeted for Alpha statusDenotes an issue tracking an enhancement targeted for Alpha statustracked/yesDenotes an enhancement issue is actively being tracked by the Release TeamDenotes an enhancement issue is actively being tracked by the Release Teamwg/device-managementCategorizes an issue or PR as relevant to WG Device Management.Categorizes an issue or PR as relevant to WG Device Management.
Type
Projects
Status
📋 Backlog
Status
KEP Merged
Status
Sig Node Consulting
Status
Tracked for Docs Freeze
Enhancement Description
Using DRA drivers for core resources (like dra-driver-cpu), enables fine-grained resource management but introduces a critical accounting problem. The standard kube-scheduler resource accounting model is unaware of allocations made by the
DynamicResourcesplugin, and vice-versa, leading to a potential resource oversubscription on nodes. While a synchronization model similar to the one used for DRA Extended Resources is needed, the same solution cannot be applied directly. The key difference is that extended resources are advertised either through the node object (node.status.allocatable) or viaResourceSlice, but not both simultaneously for the same node. In contrast, core resources like CPU are always present innode.status.allocatable, and a DRA driver would also manage them viaResourceSlice. This simultaneous representation creates a unique challenge that requires a unified accounting model.cc: @johnbelamaric, @mortent , @ffromani , @yujuhong , @dchen1107
k/enhancements) update PR(s): KEP-5517: Introduce DRA for Native Resources KEP #5755k/k) update PR(s): Introduce support of DRA for Node Allocatable Resources kubernetes#136725k/website) update PR(s): Docs update for KEP-5517: DRA for Node Allocatable Resources website#54598Please keep this description up to date. This will help the Enhancement Team to track the evolution of the enhancement efficiently.