Conversation
📝 WalkthroughSummary by CodeRabbit
WalkthroughDRA now supports configurable Kubernetes-label accelerator discovery with a default label. Shared Kubernetes helpers handle node selectors, canonical targets, domain assignments, and topology maps. InfiniBand providers use these helpers. Validation, examples, tests, and documentation reflect the configuration. ChangesShared Kubernetes discovery integration
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Client
participant DRAProvider
participant KubernetesAPI
participant DiscoverKubernetesDomains
participant AcceleratorDiscoverer
Client->>DRAProvider: Request topology
DRAProvider->>KubernetesAPI: List nodes using node selector
DRAProvider->>DiscoverKubernetesDomains: Discover domains from nodes and instances
DiscoverKubernetesDomains->>AcceleratorDiscoverer: Discover assignments from targets
AcceleratorDiscoverer-->>DRAProvider: Return accelerator domain map
DRAProvider-->>Client: Return topology or gateway error
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Greptile SummaryThe PR consolidates Kubernetes accelerator discovery and identity mapping across the DRA and InfiniBand providers while allowing DRA’s accelerator label to be configured.
Confidence Score: 5/5The PR appears safe to merge because no blocking failure remains within the eligible follow-up-review scope. No blocking failure remains. Important Files Changed
Sequence DiagramsequenceDiagram
participant Engine
participant Provider as DRA / InfiniBand-K8s
participant Kubernetes
participant Resolver as Shared accelerator resolver
Engine->>Provider: GenerateTopologyConfig(instances)
Provider->>Kubernetes: List nodes(nodeSelector)
Kubernetes-->>Provider: Selected Node objects
Provider->>Resolver: DiscoverKubernetesDomains(nodes, instances)
Resolver->>Resolver: Resolve annotated region/instance identities
Resolver->>Resolver: Read configured accelerator label
Resolver-->>Provider: Canonical DomainMap
Provider-->>Engine: topology.Graph
Reviews (5): Last reviewed commit: "refactor(providers): share Kubernetes ac..." | Re-trigger Greptile |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## main #462 +/- ##
==========================================
+ Coverage 72.15% 77.12% +4.96%
==========================================
Files 89 99 +10
Lines 5689 7200 +1511
==========================================
+ Hits 4105 5553 +1448
+ Misses 1382 1358 -24
- Partials 202 289 +87 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@charts/topograph/tests/validation_test.yaml`:
- Around line 125-134: Add positive validation cases in the chart tests for an
omitted provider.params.accelerator preserving the nvidia.com/gpu.clique default
and for source kubernetes-label with kubernetesLabel.key passing validation. Run
the Helm chart tests and review any snapshot changes before committing.
In `@docs/providers/dra.md`:
- Around line 64-66: Rewrite the Helm configuration sentence near the
`provider.params` discussion to explicitly state that the chart manages the
Topograph configuration and topology request payload, so neither needs to be
supplied separately; avoid the ambiguous “they” pronoun.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Enterprise
Run ID: a5eb2176-ddac-41fa-831c-df550514361e
📒 Files selected for processing (23)
CHANGELOG.mdcharts/topograph/templates/_validation.tplcharts/topograph/tests/validation_test.yamlcharts/topograph/values.yamldemos/dra-slinky/values.dra-slinky.kwok.yamldocs/api.mddocs/overview.mddocs/providers/dra.mddocs/reference/node-labels.mdinternal/k8s/utils.gointernal/k8s/utils_test.gopkg/accelerator/accelerator.gopkg/accelerator/accelerator_test.gopkg/accelerator/kubernetes.gopkg/providers/dra/provider.gopkg/providers/dra/provider_test.gopkg/providers/infiniband/bm.gopkg/providers/infiniband/bm_test.gopkg/providers/infiniband/k8s.gopkg/providers/infiniband/provider_bm.gopkg/providers/infiniband/provider_k8s.gopkg/providers/infiniband/provider_k8s_test.gotests/ci/values.slinky-dra-block.yaml
💤 Files with no reviewable changes (2)
- pkg/providers/infiniband/bm.go
- pkg/providers/infiniband/bm_test.go
📜 Review details
⏰ Context from checks skipped due to timeout. (5)
- GitHub Check: oci-sim / slinky
- GitHub Check: test
- GitHub Check: check
- GitHub Check: k8s / aws-sim
- GitHub Check: Greptile Review
🧰 Additional context used
📓 Path-based instructions (11)
CHANGELOG.md
📄 CodeRabbit inference engine (AGENTS.md)
Record applicable user-facing features, fixes, breaking changes, or Helm migrations under
[Unreleased].
Files:
CHANGELOG.md
**/*
📄 CodeRabbit inference engine (AGENTS.md)
**/*: Do not publicly disclose suspected security vulnerabilities; report them privately through NVIDIA PSIRT using the channels documented inSECURITY.md.
Every commit must include aSigned-off-by:trailer for DCO compliance.
Use Conventional Commits with an allowed type such asfeat,fix,docs,chore,refactor,test,build, orci.
Before pushing, runmake qualify; all CI checks, including Go build/test/lint, Codecov, and DCO, must be green before merge.
Changes to provider, engine, chart, API, configuration schema, labels, annotations, or user-facing behavior must include the corresponding documentation updates described in the documentation impact table.
Files:
CHANGELOG.mdcharts/topograph/tests/validation_test.yamldocs/api.mdcharts/topograph/values.yamldocs/reference/node-labels.mddemos/dra-slinky/values.dra-slinky.kwok.yamlcharts/topograph/templates/_validation.tplpkg/providers/infiniband/k8s.gointernal/k8s/utils_test.gotests/ci/values.slinky-dra-block.yamlpkg/providers/infiniband/provider_bm.godocs/overview.mdpkg/providers/infiniband/provider_k8s_test.gopkg/accelerator/accelerator.gopkg/providers/dra/provider_test.gopkg/accelerator/accelerator_test.gopkg/accelerator/kubernetes.gopkg/providers/infiniband/provider_k8s.godocs/providers/dra.mdpkg/providers/dra/provider.gointernal/k8s/utils.go
charts/topograph/**/*.yaml
📄 CodeRabbit inference engine (AGENTS.md)
Do not enable both
ingress.enabledandgatewayAPI.enabledin one Helm release; the routing resources are mutually exclusive.
Files:
charts/topograph/tests/validation_test.yamlcharts/topograph/values.yaml
charts/topograph/tests/**/*.{yaml,yml}
📄 CodeRabbit inference engine (AGENTS.md)
Run Helm chart tests when changing
charts/topograph/; review snapshot changes before committing updates.
Files:
charts/topograph/tests/validation_test.yaml
charts/topograph/**
⚙️ CodeRabbit configuration file
charts/topograph/**: - Check RBAC least privilege and Kubernetes API compatibility.
- ingress.enabled and gatewayAPI.enabled must remain mutually exclusive.
- HTTPRoute must contain only portable Gateway API v1 fields.
- Flag changes where values, schema, templates, NOTES, tests,
snapshots, documentation, or changelog become inconsistent.
Files:
charts/topograph/tests/validation_test.yamlcharts/topograph/values.yamlcharts/topograph/templates/_validation.tpl
**/*.go
📄 CodeRabbit inference engine (AGENTS.md)
**/*.go: Rungo fmt ./...; Go formatting is authoritative and code should not be hand-formatted.
Every new Go file must include the NVIDIA copyright header followed by the matching Apache 2.0 boilerplate.
Files:
pkg/providers/infiniband/k8s.gointernal/k8s/utils_test.gopkg/providers/infiniband/provider_bm.gopkg/providers/infiniband/provider_k8s_test.gopkg/accelerator/accelerator.gopkg/providers/dra/provider_test.gopkg/accelerator/accelerator_test.gopkg/accelerator/kubernetes.gopkg/providers/infiniband/provider_k8s.gopkg/providers/dra/provider.gointernal/k8s/utils.go
pkg/providers/**/*.go
📄 CodeRabbit inference engine (AGENTS.md)
pkg/providers/**/*.go: Providers must return a*topology.Graphand*httperr.Error; plainerroris not acceptable at the provider interface boundary.
Providers discover topology and return the canonicaltopology.Graph; they must not emit scheduler-specific output.
When usingClusterTopology, populate fabric tiers closest-first, preserve optional accelerator domain and sub-domain fields, and callToGraph.
Files:
pkg/providers/infiniband/k8s.gopkg/providers/infiniband/provider_bm.gopkg/providers/infiniband/provider_k8s_test.gopkg/providers/dra/provider_test.gopkg/providers/infiniband/provider_k8s.gopkg/providers/dra/provider.go
{cmd,pkg,internal}/**/*.go
⚙️ CodeRabbit configuration file
{cmd,pkg,internal}/**/*.go: Focus on correctness, robustness, and failure paths:
- Check error handling and propagation, including whether callers
receive enough context to diagnose failures.- Verify context propagation, cancellation, timeouts, resource cleanup,
and goroutine lifecycle.- Look for races, deadlocks, unsafe shared state, leaks, and partial
updates that can leave state inconsistent.- Check nil, empty, malformed, duplicate, boundary, and partial inputs.
- Verify deterministic and idempotent behavior where operations may be
retried or repeated.- Preserve public contracts and the architectural boundaries documented
in AGENTS.md.- Flag behavior changes that lack meaningful regression coverage.
Files:
pkg/providers/infiniband/k8s.gointernal/k8s/utils_test.gopkg/providers/infiniband/provider_bm.gopkg/providers/infiniband/provider_k8s_test.gopkg/accelerator/accelerator.gopkg/providers/dra/provider_test.gopkg/accelerator/accelerator_test.gopkg/accelerator/kubernetes.gopkg/providers/infiniband/provider_k8s.gopkg/providers/dra/provider.gointernal/k8s/utils.go
**/*_test.go
📄 CodeRabbit inference engine (AGENTS.md)
New or changed public behavior and new code paths should be covered by tests.
Files:
internal/k8s/utils_test.gopkg/providers/infiniband/provider_k8s_test.gopkg/providers/dra/provider_test.gopkg/accelerator/accelerator_test.go
⚙️ CodeRabbit configuration file
**/*_test.go: Flag missing negative, cancellation, concurrency, malformed-input,
boundary, and regression cases. Check that assertions would fail if
the behavior under test were broken.
Files:
internal/k8s/utils_test.gopkg/providers/infiniband/provider_k8s_test.gopkg/providers/dra/provider_test.gopkg/accelerator/accelerator_test.go
docs/providers/**/*.md
📄 CodeRabbit inference engine (AGENTS.md)
Provider additions or changes require the corresponding provider documentation, including prerequisites, credentials, parameters, operation, and verification.
Files:
docs/providers/dra.md
pkg/providers/*/provider.go
📄 CodeRabbit inference engine (AGENTS.md)
A provider must expose
func NamedLoader() (string, providers.Loader)and be registered inpkg/registry/registry.go.
Files:
pkg/providers/dra/provider.go
🪛 LanguageTool
docs/providers/dra.md
[grammar] ~64-~64: Ensure spelling is correct
Context: ...provider.params`. The chart manages the Topograph configuration and topology request payl...
(QB_NEW_EN_ORTHOGRAPHY_ERROR_IDS_1)
🔇 Additional comments (20)
internal/k8s/utils.go (1)
24-44: LGTM!internal/k8s/utils_test.go (1)
13-13: LGTM!Also applies to: 85-120
pkg/accelerator/accelerator.go (1)
53-70: LGTM!Also applies to: 184-213, 227-235
pkg/accelerator/kubernetes.go (1)
13-20: LGTM!Also applies to: 22-84
pkg/accelerator/accelerator_test.go (1)
14-15: LGTM!Also applies to: 213-273
pkg/providers/dra/provider.go (1)
19-19: LGTM!Also applies to: 31-117
pkg/providers/dra/provider_test.go (1)
12-18: LGTM!Also applies to: 20-100
pkg/providers/infiniband/k8s.go (1)
51-51: LGTM!pkg/providers/infiniband/provider_bm.go (1)
46-51: LGTM!pkg/providers/infiniband/provider_k8s.go (1)
30-35: LGTM!Also applies to: 37-92
pkg/providers/infiniband/provider_k8s_test.go (1)
9-25: LGTM!CHANGELOG.md (1)
11-11: LGTM!charts/topograph/templates/_validation.tpl (1)
32-32: LGTM!Also applies to: 52-52, 64-67
charts/topograph/values.yaml (1)
11-19: LGTM!demos/dra-slinky/values.dra-slinky.kwok.yaml (1)
4-7: LGTM!docs/api.md (1)
75-77: LGTM!docs/overview.md (1)
43-43: LGTM!Also applies to: 69-69
docs/providers/dra.md (1)
3-5: LGTM!Also applies to: 29-32, 36-58, 62-63, 67-83, 108-114
docs/reference/node-labels.md (1)
65-65: LGTM!tests/ci/values.slinky-dra-block.yaml (1)
6-9: LGTM!
| - it: rejects a DRA label accelerator source without a key | ||
| set: | ||
| provider: | ||
| name: dra | ||
| params: | ||
| accelerator: | ||
| source: kubernetes-label | ||
| asserts: | ||
| - failedTemplate: | ||
| errorMessage: "provider.params.accelerator.kubernetesLabel.key must be set for source kubernetes-label" |
There was a problem hiding this comment.
🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win
Add positive DRA validation coverage and run the chart tests.
These tests cover only rejection paths. Add or verify cases for omitted provider.params.accelerator, which must preserve the nvidia.com/gpu.clique default, and explicit source: kubernetes-label with kubernetesLabel.key, which must pass.
As per coding guidelines, “Run Helm chart tests when changing charts/topograph/; review snapshot changes before committing updates.”
Also applies to: 136-145
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@charts/topograph/tests/validation_test.yaml` around lines 125 - 134, Add
positive validation cases in the chart tests for an omitted
provider.params.accelerator preserving the nvidia.com/gpu.clique default and for
source kubernetes-label with kubernetesLabel.key passing validation. Run the
Helm chart tests and review any snapshot changes before committing.
Source: Coding guidelines
|
🌿 Preview your docs: https://nvidia-preview-pull-request-462.docs.buildwithfern.com/topograph |
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@pkg/providers/dra/provider_test.go`:
- Around line 36-40: Extend the unsupported-source table in the provider test to
include accelerator.SourceNvidiaSMI, asserting the same `dra provider supports
only accelerator source "kubernetes-label"` error as the existing SourceNone
case. Keep the test focused on rejecting non-label accelerator sources.
In `@pkg/providers/dra/provider.go`:
- Around line 24-28: The exported dra.DomainLabel API removal must be documented
or kept source-compatible. In pkg/providers/dra/provider.go lines 24-28, retain
DomainLabel as a deprecated alias to defaultDomainLabel if compatibility is
required; in CHANGELOG.md line 74, under [Unreleased], explicitly document the
removal of dra.DomainLabel and state that DRA defaults to nvidia.com/gpu.clique.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Enterprise
Run ID: 3cd779d9-0876-44d7-b087-5ecd448f24c7
📒 Files selected for processing (5)
CHANGELOG.mdpkg/providers/dra/provider.gopkg/providers/dra/provider_test.gopkg/topology/graph_test.gopkg/topology/topology.go
💤 Files with no reviewable changes (1)
- pkg/topology/topology.go
📜 Review details
⏰ Context from checks skipped due to timeout. (5)
- GitHub Check: Greptile Review
- GitHub Check: check
- GitHub Check: govulncheck
- GitHub Check: build
- GitHub Check: Fern Check
🧰 Additional context used
📓 Path-based instructions (8)
**/*.go
📄 CodeRabbit inference engine (AGENTS.md)
**/*.go: Rungo fmt ./...; Go formatting is authoritative and code should not be hand-formatted.
Every new Go file must include the NVIDIA copyright header followed by the matching Apache 2.0 boilerplate.
Files:
pkg/topology/graph_test.gopkg/providers/dra/provider_test.gopkg/providers/dra/provider.go
pkg/topology/**/*.go
📄 CodeRabbit inference engine (AGENTS.md)
Do not change
Graph, theVertextree, or topology constants casually; changes require discussion because every provider and engine depends on them.
Files:
pkg/topology/graph_test.go
**/*_test.go
📄 CodeRabbit inference engine (AGENTS.md)
New or changed public behavior and new code paths should be covered by tests.
Files:
pkg/topology/graph_test.gopkg/providers/dra/provider_test.go
⚙️ CodeRabbit configuration file
**/*_test.go: Flag missing negative, cancellation, concurrency, malformed-input,
boundary, and regression cases. Check that assertions would fail if
the behavior under test were broken.
Files:
pkg/topology/graph_test.gopkg/providers/dra/provider_test.go
**/*
📄 CodeRabbit inference engine (AGENTS.md)
**/*: Do not publicly disclose suspected security vulnerabilities; report them privately through NVIDIA PSIRT using the channels documented inSECURITY.md.
Every commit must include aSigned-off-by:trailer for DCO compliance.
Use Conventional Commits with an allowed type such asfeat,fix,docs,chore,refactor,test,build, orci.
Before pushing, runmake qualify; all CI checks, including Go build/test/lint, Codecov, and DCO, must be green before merge.
Changes to provider, engine, chart, API, configuration schema, labels, annotations, or user-facing behavior must include the corresponding documentation updates described in the documentation impact table.
Files:
pkg/topology/graph_test.goCHANGELOG.mdpkg/providers/dra/provider_test.gopkg/providers/dra/provider.go
{cmd,pkg,internal}/**/*.go
⚙️ CodeRabbit configuration file
{cmd,pkg,internal}/**/*.go: Focus on correctness, robustness, and failure paths:
- Check error handling and propagation, including whether callers
receive enough context to diagnose failures.- Verify context propagation, cancellation, timeouts, resource cleanup,
and goroutine lifecycle.- Look for races, deadlocks, unsafe shared state, leaks, and partial
updates that can leave state inconsistent.- Check nil, empty, malformed, duplicate, boundary, and partial inputs.
- Verify deterministic and idempotent behavior where operations may be
retried or repeated.- Preserve public contracts and the architectural boundaries documented
in AGENTS.md.- Flag behavior changes that lack meaningful regression coverage.
Files:
pkg/topology/graph_test.gopkg/providers/dra/provider_test.gopkg/providers/dra/provider.go
CHANGELOG.md
📄 CodeRabbit inference engine (AGENTS.md)
Record applicable user-facing features, fixes, breaking changes, or Helm migrations under
[Unreleased].
Files:
CHANGELOG.md
pkg/providers/**/*.go
📄 CodeRabbit inference engine (AGENTS.md)
pkg/providers/**/*.go: Providers must return a*topology.Graphand*httperr.Error; plainerroris not acceptable at the provider interface boundary.
Providers discover topology and return the canonicaltopology.Graph; they must not emit scheduler-specific output.
When usingClusterTopology, populate fabric tiers closest-first, preserve optional accelerator domain and sub-domain fields, and callToGraph.
Files:
pkg/providers/dra/provider_test.gopkg/providers/dra/provider.go
pkg/providers/*/provider.go
📄 CodeRabbit inference engine (AGENTS.md)
A provider must expose
func NamedLoader() (string, providers.Loader)and be registered inpkg/registry/registry.go.
Files:
pkg/providers/dra/provider.go
🔇 Additional comments (3)
pkg/providers/dra/provider.go (1)
19-22: LGTM!Also applies to: 30-117
pkg/providers/dra/provider_test.go (1)
12-35: LGTM!Also applies to: 43-100
pkg/topology/graph_test.go (1)
237-253: LGTM!
b888067 to
a46b4ef
Compare
Signed-off-by: Dmitry Shmulevich <17212177+dmitsh@users.noreply.github.com>
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/providers/dra.md`:
- Around line 64-66: Update the DRA provider documentation to describe
provider.params.accelerator as optional only as a complete object, while stating
that when present it must include both accelerator.source and
accelerator.kubernetesLabel.key. Remove wording that implies the source label or
nodeSelector fields are independently optional, and keep the existing guidance
that the chart manages the related Topograph configuration and topology request
payload.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Enterprise
Run ID: 7f3c81ed-50c7-4f4e-acf9-9b296c10f7f5
📒 Files selected for processing (1)
docs/providers/dra.md
📜 Review details
⏰ Context from checks skipped due to timeout. (9)
- GitHub Check: Greptile Review
- GitHub Check: k8s / aws-sim
- GitHub Check: k8s / gcp-sim
- GitHub Check: check
- GitHub Check: k8s / test
- GitHub Check: build
- GitHub Check: test
- GitHub Check: oci-sim / slinky
- GitHub Check: govulncheck
🧰 Additional context used
📓 Path-based instructions (2)
docs/providers/**/*.md
📄 CodeRabbit inference engine (AGENTS.md)
Provider additions or changes require the corresponding provider documentation, including prerequisites, credentials, parameters, operation, and verification.
Files:
docs/providers/dra.md
**/*
📄 CodeRabbit inference engine (AGENTS.md)
**/*: Do not publicly disclose suspected security vulnerabilities; report them privately through NVIDIA PSIRT using the channels documented inSECURITY.md.
Every commit must include aSigned-off-by:trailer for DCO compliance.
Use Conventional Commits with an allowed type such asfeat,fix,docs,chore,refactor,test,build, orci.
Before pushing, runmake qualify; all CI checks, including Go build/test/lint, Codecov, and DCO, must be green before merge.
Changes to provider, engine, chart, API, configuration schema, labels, annotations, or user-facing behavior must include the corresponding documentation updates described in the documentation impact table.
Files:
docs/providers/dra.md
🧠 Learnings (6)
📓 Common learnings
Learnt from: CR
Repo: NVIDIA/topograph PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-08-08T10:57:01.397Z
Learning: Applies to pkg/engines/k8s/**/*.go : Do not invent Kubernetes label keys in provider code; use canonical graph values and the configured closest-first `fabricLabels` and fixed accelerator sub-domain key.
📚 Learning: 2026-08-08T10:57:01.397Z
Learnt from: CR
Repo: NVIDIA/topograph PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-08-08T10:57:01.397Z
Learning: Applies to **/* : Changes to provider, engine, chart, API, configuration schema, labels, annotations, or user-facing behavior must include the corresponding documentation updates described in the documentation impact table.
Applied to files:
docs/providers/dra.md
📚 Learning: 2026-08-08T10:57:01.397Z
Learnt from: CR
Repo: NVIDIA/topograph PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-08-08T10:57:01.397Z
Learning: Applies to pkg/engines/k8s/**/*.go : Do not invent Kubernetes label keys in provider code; use canonical graph values and the configured closest-first `fabricLabels` and fixed accelerator sub-domain key.
Applied to files:
docs/providers/dra.md
📚 Learning: 2026-08-08T10:57:01.397Z
Learnt from: CR
Repo: NVIDIA/topograph PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-08-08T10:57:01.397Z
Learning: Applies to docs/providers/**/*.md : Provider additions or changes require the corresponding provider documentation, including prerequisites, credentials, parameters, operation, and verification.
Applied to files:
docs/providers/dra.md
📚 Learning: 2026-08-08T10:57:01.397Z
Learnt from: CR
Repo: NVIDIA/topograph PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-08-08T10:57:01.397Z
Learning: Applies to pkg/providers/**/*.go : Providers discover topology and return the canonical `topology.Graph`; they must not emit scheduler-specific output.
Applied to files:
docs/providers/dra.md
📚 Learning: 2026-08-08T10:57:01.397Z
Learnt from: CR
Repo: NVIDIA/topograph PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-08-08T10:57:01.397Z
Learning: Applies to pkg/providers/**/*.go : When using `ClusterTopology`, populate fabric tiers closest-first, preserve optional accelerator domain and sub-domain fields, and call `ToGraph`.
Applied to files:
docs/providers/dra.md
🪛 LanguageTool
docs/providers/dra.md
[grammar] ~64-~64: Ensure spelling is correct
Context: ...provider.params. The chart manages the Topograph configuration and topology request payl...
(QB_NEW_EN_ORTHOGRAPHY_ERROR_IDS_1)
🔇 Additional comments (1)
docs/providers/dra.md (1)
3-5: LGTM!Also applies to: 29-32, 36-44, 56-58, 80-82, 108-112
| source label and `nodeSelector` under `provider.params`. The chart manages the | ||
| Topograph configuration and topology request payload, so you do not need to | ||
| supply either one separately. |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick win
Describe the optional configuration accurately.
Line 64 calls this an “optional source label.” provider.params.accelerator is optional, but DRA requires both accelerator.source and accelerator.kubernetesLabel.key when that object is present. Users can otherwise treat these fields as independently optional and submit a configuration that Helm rejects.
Proposed wording
-Set the optional source label and `nodeSelector` under `provider.params`.
+Set the optional accelerator configuration and `nodeSelector` under `provider.params`.Cross-file evidence: charts/topograph/templates/_validation.tpl:43-67 requires the accelerator source and Kubernetes label key for this DRA configuration.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| source label and `nodeSelector` under `provider.params`. The chart manages the | |
| Topograph configuration and topology request payload, so you do not need to | |
| supply either one separately. | |
| Set the optional accelerator configuration and `nodeSelector` under `provider.params`. The chart manages the | |
| Topograph configuration and topology request payload, so you do not need to | |
| supply either one separately. |
🧰 Tools
🪛 LanguageTool
[grammar] ~64-~64: Ensure spelling is correct
Context: ...provider.params. The chart manages the Topograph configuration and topology request payl...
(QB_NEW_EN_ORTHOGRAPHY_ERROR_IDS_1)
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/providers/dra.md` around lines 64 - 66, Update the DRA provider
documentation to describe provider.params.accelerator as optional only as a
complete object, while stating that when present it must include both
accelerator.source and accelerator.kubernetesLabel.key. Remove wording that
implies the source label or nodeSelector fields are independently optional, and
keep the existing guidance that the chart manages the related Topograph
configuration and topology request payload.
1 participant
Reuse node selector parsing, Kubernetes target resolution, accelerator
assignment conversion, and base node annotations across the DRA and
InfiniBand providers.
Allow DRA to use the shared kubernetes-label accelerator configuration
while preserving nvidia.com/gpu.clique as its legacy default.
closes #461