[sylvesterkaczmarek]

Changelog title

Manual nodes persist across restarts

Changelog body

• Manually added nodes are restored after the broker or manual-node worker restarts.

Bumps

• services: patch
• nvpair-cluster-manager: none
• nvpair-engine-manager: none
• nvpair-errors: none
• nvpair-job-scheduler: none
• nvpair-manual-nodes: none
• nvpair-node-info: none
• nvpair-node-scanner: none
• nvpair-node-settings: none
• nvpair-proxy: none
• nvpair-tui: none
• nvpair-ui-broker: patch
• nvpair-workload-manager: none

Description

Closes #58.

Move manual-node durability into nvpair-ui-broker so headless TUI/service nodes keep manual peers across broker and worker restarts. Successful add/remove operations update the existing configs/manual-nodes.json, and a restarted manual-nodes worker is repopulated automatically.

This version is rebased and adapted to the current develop architecture after the TUI overhaul. The obsolete Electron-owned store/replay path is removed so the broker is the single persistence owner.

Scope

• persist successful manual-node adds/removes in the broker
• replay durable manual nodes into replacement workers
• resolve stable discovered node IDs back to manual aliases for removal
• remove Electron-owned persistence/replay logic
• keep the existing JSON-RPC methods and on-disk file location
• declare a patch bump for nvpair-ui-broker through release intent; do not edit services/versions.json manually

Validation

• cd services/nvpair-ui-broker && go test ./... — passed
• cd desktop && npm run typecheck — passed
• cd desktop && npm run lint — passed
• cd desktop && npm run test:unit -- modular/modular-supervisor-readiness.test.ts — 4 passed
• cd desktop && npm run test:unit — 39 files, 231 tests passed
• cd desktop && npm run dead-code:check — passed with zero findings
• cd desktop && npm run service-contracts:check — passed with no unignored drift
• node scripts/spdx-headers.mjs — 1027 checked, 0 missing
• git diff --check — passed

Risk

The JSON-RPC surface and persisted file schema are unchanged. Existing manual-nodes.json entries remain readable. The ownership change removes the duplicate Electron persistence path and makes broker/TUI-only deployments durable as well.

Checklist

• I have read the Contributing Guidelines.
• Every commit is signed off, certifying the Developer Certificate of Origin.
• New or existing tests cover the change.
• Relevant documentation is updated.
• I checked the diff, changed filenames, and commit messages for credentials, private data, internal URLs, internal issue identifiers, and generated artifacts.
• I recorded the validation commands and results above.
• I declared version bumps in the release-intent block above. services/versions.json is written by automation and is not edited in this PR.

[sherief-nv]

This makes sense. Moving MR against develop branch.

[Noah-Tervalon-Nvidia]

I'm working on a massive overhaul to the TUI that we will want to bring in before we make other TUI changes. As soon as that comes in I'd love to get this updated and then will verify it against the new codebase.

[sylvesterkaczmarek]

Thanks Noah, and thanks Sherief for moving this onto develop. Makes sense. I’ll rebase and adapt the persistence change once the TUI overhaul lands so it fits the new structure cleanly.

[Noah-Tervalon-Nvidia]

@sylvesterkaczmarek can you have a look at the major overhaul I made to the TUI - #117. Would love to get eyes there, I think we will still need this fix.

[sylvesterkaczmarek]

I reviewed the manual-node/persistence path in #117. The new node-first TUI is compatible with this PR and does not replace it: the manual-nodes worker is still in-memory only, the broker still documents restart loss, and the desktop still owns replay. The TUI keeps the existing node/add, node/remove and nodes/list contract and correctly preserves the manual alias separately from the stable node key. So I agree this fix is still needed. I’ll adapt/rebase #60 once #117 lands rather than racing the overhaul.

[sylvesterkaczmarek]

@Noah-Tervalon-Nvidia @ckelseynv I have now done the adaptation we deferred until the TUI overhaul landed. This PR is rebased onto current develop, no longer conflicts, and keeps manual-node persistence in nvpair-ui-broker while removing the old Electron-owned persistence/replay path. It also follows the current release-intent/versioning rules, with no direct services/versions.json edit. Validation on the current head 6281a4c: full broker Go tests pass; desktop typecheck and lint pass; all 231 desktop unit tests pass; dead-code, service-contract, SPDX, and git diff checks pass. The commit is signed and GitHub Verified. Could you re-review the current head when convenient?

[Noah-Tervalon-Nvidia]

@sylvesterkaczmarek we were figuring out getting in llama.cpp and the updated TUI along with a handful of other small fixes. Sorry for the slow response here. This looks good to me, I'd love to merge it in on Wednesday next week.