Draft of the DRA support - #771

Closed

bart0sh wants to merge 110 commits into

containers:mainfrom

bart0sh:DRA

Commits on Aug 26, 2026

docs/dra: add initial design draft for PCT-DRA integration
bart0sh
committed
b4b9156
Add CLAUDE.md with project info
bart0sh
committed
9a4335b
docs/dra: add DRA ecosystem landscape reference

bart0sh
and
claude
committed
e79f81b
docs/dra: add PR #536 prior-art analysis

bart0sh
and
claude
committed
52c9fd9
docs/dra: rewrite design doc, expand scope to full cpuClass

bart0sh
and
claude
committed
cda1fb7
docs/dra: add v1 implementation plan

bart0sh
and
claude
committed
495dbf7
docs/dra: make references clickable on GitHub

bart0sh
and
claude
committed
da3a80f
CLAUDE.md: fix broken pkg/resmgr/policy/topology-aware/ link

bart0sh
and
claude
committed
31b4f48
docs/dra: rework pr-536-analysis takeaway as reuse tiers

bart0sh
and
claude
committed
11fbafe
docs/dra: use tags.cncf.io/container-device-interface for CDI writes

bart0sh
and
claude
committed
5351aa7
docs: add DRA v1 step 1 implementation plan

bart0sh
and
claude
committed
e453fd2
pkg/kubernetes/client: add package skeleton with rest-config helpers

bart0sh
and
claude
committed
e54cf77
pkg/kubernetes/client: add New constructor, config options, methods

bart0sh
and
claude
committed
8a99220
pkg/kubernetes/client: add content-type options

bart0sh
and
claude
committed
042cdca
pkg/kubernetes/watch: move verbatim from pkg/agent/watch

bart0sh
and
claude
committed
9a8aabb
pkg/agent: rewire to use pkg/kubernetes/client

bart0sh
and
claude
committed
f99ed51
docs: mark DRA step 1 complete, move plan to completed/

bart0sh
and
claude
committed
2fc99c1
docs: add DRA v1 step 2 implementation plan

bart0sh
and
claude
committed
7a9efc8
pkg/resmgr/dra: introduce DRA plugin package skeleton

bart0sh
and
claude
committed
a0d4d3f
build: add *.test to .gitignore

bart0sh
and
claude
committed
120fbd4
docs: mark DRA step 2 complete, move plan to completed/

bart0sh
and
claude
committed
aff0b08
docs: add DRA v1 step 3 implementation plan

bart0sh
and
claude
committed
06d8405
pkg: add cpuClass.dra.publish config field + ValidateCPUClassesForDRA

bart0sh
and
claude
committed
d59ef42
docs: mark DRA step 3 complete, sync plan.md + design.md

bart0sh
and
claude
committed
bff7b50
build: ignore extracted sysfs testdata fixture directories

bart0sh
and
claude
committed
3f07a67
docs: add DRA v1 step 4 implementation plan

bart0sh
and
claude
committed
35b212e
pkg/resmgr/cpuclass/internal/pct: add PickHpCpus, ReleaseHpCpus, Punits

bart0sh
and
claude
committed
7982db5
docs: mark DRA step 4 complete, move plan to completed/

bart0sh
and
claude
committed
364cc39
docs: add DRA v1 step 5 implementation plan

bart0sh
and
claude
committed
3ad9c80
feat: bump k8s.io/api and related modules to v0.36.3
bart0sh
committed
0ae8bc0
feat: export pct.Allocator.IsHPClass
bart0sh
committed
56f0450
feat: implement DRA device builder for cpuClass (Tasks 3-5)
bart0sh
committed
883c91b
docs: update DRA plan.md and design.md for step 5 completion
bart0sh
committed
6d5ea9b
fix: address code review findings
bart0sh
committed
67f1ba2
docs: move completed plan 20260820-dra-step5-cpuclass-dra-devices.md to completed/
bart0sh
committed
4bc8cf9
feat: add k8s.io/dynamic-resource-allocation dependency (DRA step 6 task 1)
bart0sh
committed
1c79d58
feat: add logr.Logger bridge (logging.go) for DRA kubelet plugin
bart0sh
committed
8ed4f79
feat: add DRAPlugin interface stubs (PrepareResourceClaims, UnprepareResourceClaims, HandleError)
bart0sh
committed
de8bc4c
feat: flesh out Deps, Plugin struct, and implement New with validation
bart0sh
committed
202551c
feat: implement Plugin.Start, Stop, PublishResources (DRA step 6, task 5)
bart0sh
committed
b7687e4
feat: add integration test for DRA plugin PublishResources against fake clientset
bart0sh
committed
524e591
docs: add feature-gate probe deferral note to plan.md
bart0sh
committed
33488af
docs: update plan.md and design.md for DRA step 6 completion
bart0sh
committed
7fb6133
fix: address code review findings
bart0sh
committed
15ad916
docs: move completed plan 20260821-dra-step6-plugin-wire.md to completed/
bart0sh
committed
e5b1a9b
docs: add DRA step 7 implementation plan (Prepare/Unprepare + CDI writer)

bart0sh
and
claude
committed
3b7b8c1
docs: resolve DRA open decision 1 — class-derived attribute freshness (Option B)

bart0sh
and
claude
committed
e4d3970
docs: update DRA plan.md for resolved decision 8 (Option B freshness check)

bart0sh
and
claude
committed
9270e09
feat: add pct.AccountHpCpus for DRA restart reconciliation
bart0sh
committed
0bfb193
feat: add ClaimAllocator pass-throughs on cpuclass.Handler
bart0sh
committed
be489a8
feat: filter non-HP devices from DRADevices (Task 3)
bart0sh
committed
f68f5e1
feat: define state types, interfaces, and extend Deps (DRA Task 4)
bart0sh
committed
0cb9dd0
docs: add LiveClaimClasses() to step 7 plan for Option B Reconfigure check

bart0sh
and
claude
committed
c2289df
feat: add CDI dependency and implement cdiWriter (DRA step 7 task 5)
bart0sh
committed
8621faf
feat: add claim state persistence (Task 6)
bart0sh
committed
9b80efb
feat: implement PrepareResourceClaims (DRA step 7)
bart0sh
committed
8cd5500
feat: implement UnprepareResourceClaims (DRA step 7, task 8)
bart0sh
committed
64afa0b
feat: task 9 — LiveClaimClasses, RestoreClaimsLocked, Start reconciliation
bart0sh
committed
35fdcc4
chore: task 10 — dependency hygiene + lint
bart0sh
committed
22293fb
chore: verify acceptance criteria task 11
bart0sh
committed
a0a4be9
docs: update plan.md and design.md for DRA step 7 completion
bart0sh
committed
2d6b6be
docs: move completed plan 20260821-dra-step7-prepare-unprepare-cdi.md to completed/
bart0sh
committed
16110c3
docs: add DRA step 8 implementation plan (topology-aware wire-up)

bart0sh
and
claude
committed
ed59531
docs: flesh out DRA plan.md Step 10 e2e test scenario

bart0sh
and
claude
committed
7f5611b
feat: add DRA config CR extension to topology-aware policy (Step 8 Task 1)
bart0sh
committed
2e83027
feat: add Backend.Stop() and WithLock/KubeClientFn/NodeName wiring for DRA
bart0sh
committed
597496c
feat: add cache.Container.GetCDIDeviceNames accessor
bart0sh
committed
5908837
feat: add DRA driver-name constant, policy-backed adapter, and Plugin.LiveClaimsLocked (Step 8 Task 4)
bart0sh
committed
d545bb1
feat: add tree-wide DRA claim CPU accounting to pool supply
bart0sh
committed
31de5d8
feat: DRA claim identification and allocateClaim/releaseClaim with eviction
bart0sh
committed
c790538
feat: wire DRA claim allocation into AllocateResources/ReleaseResources (Task 7)
bart0sh
committed
367022c
feat: implement DRA claim pool-supply re-apply after Start/Reconfigure (Task 8)
bart0sh
committed
e6b9e2c
feat: wire DRA plugin lifecycle into topology-aware policy (Setup/Start/Stop)
bart0sh
committed
b6a56bf
feat: add Reconfigure refusal + PostReconfigure post-unlock seam for DRA
bart0sh
committed
9fb8d0d
test: verify DRA step 8 acceptance criteria (task 11)
bart0sh
committed
0348262
docs: land DRA step 8 in plan.md/design.md (task 12)
bart0sh
committed
840a527
fix: address code review findings
bart0sh
committed
bb6348d
fix: address code review findings (iteration 2)

bart0sh
and
claude
committed
2320e85
fix: hold resmgr write lock around reapplyDRAClaims in Start()

bart0sh
and
claude
committed
47f6c6b
refactor: address code smell findings (attribute-key constants, stale comments)

bart0sh
and
claude
committed
5b3e3db
fix: apply per-alloc cpuClass for multi-class DRA claims

bart0sh
and
claude
committed
9835596
docs: move completed plan 20260823-dra-step8-topology-aware-wire-up.md to completed/
bart0sh
committed
16fa1cc
docs: add DRA step 9 (Helm chart additions) implementation plan
bart0sh
committed
fed4492
feat: add config.dra toggle to topology-aware Helm chart
bart0sh
committed
c5d9a3b
feat: add DRA RBAC rules to topology-aware clusterrole
bart0sh
committed
18b61c9
feat: add DRA host mounts to topology-aware daemonset
bart0sh
committed
2922b7d
feat: add base DeviceClass template for DRA-enabled topology-aware chart
bart0sh
committed
680f67e
test: verify DRA step 9 acceptance criteria (task 5)
bart0sh
committed
7199124
docs: land DRA step 9 in plan.md/design.md (task 6)
bart0sh
committed
162245b
fix: address code review findings
bart0sh
committed
511c9f5
fix: address code smells review findings
bart0sh
committed
0548489
docs: move completed plan 20260825-dra-step9-helm-chart-additions.md to completed/
bart0sh
committed
1317b44
docs: add DRA step 10 e2e test implementation plan

bart0sh
and
claude
committed
7e1ba14
feat: provision DRAConsumableCapacity/DRANodeAllocatableResources gates in e2e VM
bart0sh
committed
4635615
feat: add dra.enabled passthrough to shared helm-config template
bart0sh
committed
efc5fcc
feat: add DRA feature-gate probe and skip path to test20-dra
bart0sh
committed
3f7103f
feat: add ResourceClaim and pod manifests to test20-dra
bart0sh
committed
a1a501a
feat: add CLOS/env/allocatable assertions and cleanup to test20-dra
bart0sh
committed
648b2e7
fix: assert KEP-5517 allocation via pod.status.nodeAllocatableResourceClaimStatuses
bart0sh
committed
e83ab09
test: verify DRA e2e acceptance criteria (task 6)
bart0sh
committed
217d95e
docs: land DRA step 10 e2e test in plan.md
bart0sh
committed
f1ce320
fix: address code review findings
bart0sh
committed
cb4dc7a
style: address code smells review findings
bart0sh
committed
5fcb6c7
docs: move completed plan 20260825-dra-step10-e2e-test.md to completed/
bart0sh
committed
1774fca
go mod tidy
bart0sh
committed
4e8e8ac

Commits on Aug 28, 2026

e2e: add k8s_log_verbosity var for apiserver/scheduler/controller-manager/kubelet debug logging

bart0sh
and
claude
committed
3af636f
fix: support Kubernetes 1.37's NodeAllocatableResources field rename

bart0sh
and
claude
committed
38f6959
docs: record the NodeAllocatableResources field rename and 1.37 floor bump

bart0sh
and
claude
committed
656b3c6
fix: update DRA e2e test for 1.37's NodeAllocatableResourceClaimStatus.Mapping

bart0sh
and
claude
committed
e025e92
docs: record the second (pod-status-side) KEP-5517 tombstone found on 1.37

bart0sh
and
claude
committed
c5844b4

Uh oh!

Commits

Commits on Aug 26, 2026

Commits on Aug 28, 2026