feat(provider/lldp): add LLDP leaf topology discovery

[dmitsh]

Add bare-metal and Kubernetes LLDP providers that group compute
nodes by their directly connected switch chassis ID.

Support interface filtering, reject ambiguous multi-switch
attachments, and integrate LLDP collection with the node-data-broker,
Helm chart, documentation, and Slurm update tooling.

closes #442

[greptile-apps]

Greptile Summary

The PR adds bare-metal and Kubernetes LLDP providers, LLDP collection through the node-data-broker, interface filtering, topology construction, Helm integration, documentation, and Slurm tooling support.

• Registers lldp-bm and lldp-k8s providers and maps chassis IDs into closest-first leaf topology.
• Adds keyed LLDP JSON decoding, interface selection, and multi-switch ambiguity handling.
• Adds broker annotation collection and deletion semantics for stale LLDP data.
• Adds container, Helm, documentation, and script integration.

Confidence Score: 4/5

The PR is not yet safe to merge because the keyed-chassis parser fails its new regression case and several LLDP discovery failures can still preserve stale topology state.

The parser classifies a chassis keyed as id as a normalized record, dropping its nested identifier and causing the added regression assertion to fail. Separately, command, parse, and ambiguity errors still abort the broker before deleting an existing LLDP chassis annotation, leaving nodes grouped under outdated switches.

Files Needing Attention: pkg/providers/lldp/common.go, pkg/providers/lldp/provider_k8s.go, cmd/node-data-broker/main.go

Important Files Changed

Filename	Overview
pkg/providers/lldp/common.go	Adds LLDP JSON decoding, filtering, normalization, and switch-ID generation, but the record discriminator mishandles the keyed-chassis regression case.
pkg/providers/lldp/provider_k8s.go	Adds Kubernetes annotation-backed topology and broker collection, while some discovery failures still preserve stale chassis state.
pkg/providers/lldp/provider_bm.go	Adds pdsh-based bare-metal LLDP collection and leaf topology generation.
cmd/node-data-broker/main.go	Wires LLDP collection into the broker and supports empty-value annotation deletion.
charts/topograph/templates/nodeDataBroker/daemonset.yaml	Forwards configured LLDP interfaces to the node-data-broker.
pkg/registry/registry.go	Registers both LLDP provider loaders.

Sequence Diagram

sequenceDiagram
    participant LLDPD as Host lldpd
    participant Broker as node-data-broker
    participant K8s as Kubernetes API
    participant Provider as lldp-k8s provider
    participant Graph as Topology graph
    Broker->>LLDPD: lldpctl -f json
    LLDPD-->>Broker: keyed interface/chassis JSON
    Broker->>Broker: parse and select chassis
    Broker->>K8s: update node annotations
    Provider->>K8s: list selected nodes
    K8s-->>Provider: instance and chassis annotations
    Provider->>Graph: build leaf-switch tiers

Loading

_{Reviews (6): Last reviewed commit: "feat(provider/lldp): add LLDP leaf topol..." | Re-trigger Greptile}

[codecov]

Codecov Report

❌ Patch coverage is 68.93491% with 105 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.35%. Comparing base (8468abe) to head (6ad1aa8).
⚠️ Report is 41 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/providers/lldp/common.go	81.67%	18 Missing and 17 partials ⚠️
pkg/providers/lldp/provider_bm.go	48.52%	24 Missing and 11 partials ⚠️
pkg/providers/lldp/provider_k8s.go	54.16%	24 Missing and 9 partials ⚠️
cmd/node-data-broker/main.go	71.42%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #443      +/-   ##
==========================================
+ Coverage   72.15%   75.35%   +3.19%     
==========================================
  Files          89       98       +9     
  Lines        5689     6994    +1305     
==========================================
+ Hits         4105     5270    +1165     
- Misses       1382     1421      +39     
- Partials      202      303     +101     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

🌿 Preview your docs: https://nvidia-preview-pull-request-443.docs.buildwithfern.com/topograph

[greptile-apps]

Keyed chassis ID is discarded

When a keyed chassis object uses id as its chassis name, isChassisRecord classifies the outer object as a normalized record. Decoding then leaves the nested chassis identifier empty, so parseNeighbors discards the neighbor and the new TestParseNeighborsHandlesNamesMatchingRecordFields regression case fails.

Knowledge Base Used: Providers registry