doca-upgrade

작성자: nvidia

사용자가 DOCA 업그레이드 또는 다운그레이드를 고려할 때 이 스킬을 사용하세요 — 호스트를 최신 DOCA 릴리스로 이동하거나, BlueField BFB를 새로고침하거나, 버전을 올리는…

npx skills add https://github.com/nvidia/skills --skill doca-upgrade

DOCA upgrade

Where to start: This skill is the bundle's single source of truth for the discipline that wraps a DOCA upgrade or downgrade. Open TASKS.md when the user wants to do something with an upgrade (detect the gap, ask for confirmation, walk the guided upgrade, verify it, or diagnose a failed one); open CAPABILITIES.md when the question is what does an upgrade even cover (the host apt-upgrade vs BFB reflash vs container-tag bump vs downgrade/rollback modes, the never-auto rule, the confirmation gate, the sunset-awareness concern, the error taxonomy, and the upgrade-specific safety overlay).

For a planned move, the single most important rule this skill exists to enforce is detect → report → ASK → only-then guided upgrade. Never upgrade automatically. The skill detects the installed DOCA env and what newer release is available, reports the gap clearly, then STOPS and asks the user for explicit confirmation. Only on an explicit "yes" does it walk a guided upgrade. A failed or partial move instead enters the recovery ladder without presenting a new target move. In both paths every hardware / firmware / reboot step is delegated to doca-hardware-safety, never redefined here. Version detection itself is owned by doca-version; this skill routes there and does not restate the detection chain.

Example questions this skill answers well

The CLASSES of upgrade questions this skill is built to answer, each with one worked example. The agent should treat the class as the load-bearing piece — the worked example is a single instance.

  • "Is there a newer DOCA than what I have, and how far behind am I?" — worked example: "pkg-config --modversion doca-common says 3.1.0; what's current and what does moving cost me?". Answered by the detect-and-report-the-gap workflow in TASKS.md ## configure, which routes the detection to doca-version and the what is current lookup to doca-public-knowledge-map.
  • "Should I just run apt upgrade to move to the next DOCA release?" — worked example: "I want the newest Flow features; can I upgrade in place right now?". Answered by the never-auto rule and the confirmation gate in CAPABILITIES.md ## Safety policy plus the ask-before-acting step in TASKS.md ## run — the agent reports the gap and STOPS for explicit confirmation before any upgrade command runs.
  • "My upgrade left the host on a different release than the BlueField — what happened?" — worked example: "after the upgrade pkg-config says 3.3.0 but bfver still says 3.1.0". Answered by the host/BFB-skew row in CAPABILITIES.md ## Error taxonomy and the failed-upgrade diagnosis in TASKS.md ## debug, which re-routes the detection to doca-version ## test.
  • "The upgrade aborted half-way and now dpkg is wedged — how do I recover?" — worked example: "apt upgrade was interrupted; dpkg reports an interrupted transaction". Answered by the partial-upgrade and aborted-transaction rows in CAPABILITIES.md ## Error taxonomy plus the recovery ladder in TASKS.md ## debug.
  • "The upgrade needs me to reflash the BFB / flip BlueField mode / reboot — is that safe?" — worked example: "the target release requires a new BFB; can I reflash now?". Answered by routing every hardware / firmware / reboot step to doca-hardware-safety via TASKS.md ## run; this skill never redefines the reflash, mode-flip, or cold-power-cycle discipline.
  • "My installed component is being sunset — should I keep building on it or move?" — worked example: "I rely on DOCA App Shield / the DOCA Flow Inspector service; are they on a deprecation track?". Answered by the sunset / deprecation awareness concern in CAPABILITIES.md ## Capabilities and modes, which routes the user to the public release notes via doca-public-knowledge-map to confirm lifecycle status rather than recommending continued investment in a component that may be sunsetting.

When to load this skill

Load this skill whenever an upgrade or downgrade of DOCA is the load-bearing concern — the user is asking whether a newer release exists, whether to move to it, how to move safely, or how to recover a move that went wrong. The decision must be made before the agent composes its first sentence; the activation checklist below is mirrored here so the activation rule is at hand whenever this skill is consulted. For a planned upgrade or downgrade, it reports the gap and then STOPS for explicit user confirmation. For a failed or partial upgrade, it enters TASKS.md ## debug first to capture and stabilize the current state; it never uses another upgrade command as the first recovery action, retries the move, or issues a new upgrade command without explicit confirmation.

Agent activation checklist — load this skill at the START of the answer when any cell below is true

Trigger classConcrete prompt-side signals (any one fires the overlay)
Upgrade / downgrade intent"is there a newer DOCA", "should I move to the next release", "upgrade me to the latest", "roll me back to the previous DOCA", "downgrade to 3.1 for a bug repro"
Accidental / drifted upgrade"apt upgrade pulled a newer DOCA point release and broke my build", "my host and BlueField are on different releases now", *"the source channel is latest and I didn't mean to move" *
Failed / partial upgrade"the upgrade aborted half-way", "dpkg is wedged after an interrupted transaction", "upgrade finished but the four-way match no longer holds"
Container tag bumpthe user wants to move an NGC DOCA container deployment from one tag to a newer one, or asks whether bumping the tag is an upgrade
Sunset / deprecation awarenessthe user asks whether the component they depend on is deprecated, being sunset, or worth continued investment

First classify the request. A planned move (including a requested rollback) follows detect → report → ASK. A failed, partial, or accidentally drifted move is already a diagnosis/recovery request: load this skill, capture the current four-source state, and enter TASKS.md ## debug; do not ask for permission to perform an upgrade that has already failed, and do not issue another upgrade command as the first recovery action.

For a planned move, the agent MUST:

  1. Route planned-upgrade version detection to doca-version ## configure. For failed or partial upgrades, start with doca-version ## debug and TASKS.md ## debug. Do NOT restate the four-source detection chain here; this skill consumes the detected state and four-way match status.
  2. Report the gap and STOP for explicit confirmation. Present the installed release, the available target release, and what moving costs (per CAPABILITIES.md ## Version compatibility), then ask the user to confirm. Do NOT proceed to any upgrade command until the user explicitly says yes.
  3. Route every hardware / firmware / reboot step to doca-hardware-safety. This skill names when such a step is part of the upgrade; that skill names how it is applied safely.

For a failed/partial move, the agent MUST instead classify the captured state against CAPABILITIES.md ## Error taxonomy, delegate the single documented corrective action, and re-enter the recovery ladder. Any unsupported installed → target jump fails closed: do not improvise intermediate releases or commands; require an upgrade path explicitly documented by the Compatibility Policy or release notes.

Do not load this skill for first-time install (use doca-setup), for the version-detection chain in isolation (use doca-version), or for general orientation (use doca-public-knowledge-map).

What this skill provides

This is a thin loader. The body keeps only the orientation needed to pick the right next file. The substantive upgrade material lives in two companion files:

  • CAPABILITIES.md — the upgrade surface: the upgrade-mode taxonomy (host apt upgrade vs BFB reflash vs container-tag bump vs downgrade/rollback), the never-auto rule and the confirmation gate, the sunset / deprecation-awareness concern, the version-compatibility overlay that redirects to doca-version, the error taxonomy (partial upgrade, apt-source drift, host/BFB skew, aborted transaction / dpkg interrupted), the observability surface (which commands prove the upgrade did or did not happen), and the upgrade-specific safety overlay on top of doca-hardware-safety.
  • TASKS.md — step-by-step workflows for the in-scope verbs: configure (detect + discover target + apt-source precheck), build and modify (routing stubs — the version-pin change is owned elsewhere), run (the confirmation-gated guided upgrade), test (post-upgrade four-way verification), debug (diagnose a failed / partial upgrade), plus a Deferred task verbs block.

Loading order

  1. Read this SKILL.md first to confirm the user's question is in scope (an upgrade / downgrade is being contemplated or recovered).
  2. For the upgrade-mode taxonomy, the never-auto rule, the sunset-awareness concern, the error taxonomy, observability, and the safety overlay, see CAPABILITIES.md.
  3. For step-by-step workflows — detect-and-report, the confirmation-gated guided upgrade, post-upgrade verification, and failed-upgrade diagnosis — see TASKS.md.

Related skills

  • doca-version — owns the four-source detection chain and the four-way match rule. This skill's detect step routes there and does not redefine version detection; the gap report consumes the detected version.
  • doca-setup — owns first-time install, install verification, and the apt-source path. This skill's apt-source consistency precheck cross-links there before any apt-shaped upgrade is contemplated.
  • doca-hardware-safety — owns every hardware / firmware / reboot step (mlxconfig, BFB reflash, BlueField mode flip, cold power cycle). This skill's ## Safety policy overlays that meta-policy and adds only the upgrade-specific surface (confirmation gate, rollback-first, maintenance window).
  • doca-public-knowledge-map — the routing table to public DOCA docs, including the release notes and the Compatibility Policy. This skill routes what is the current release and is this component sunsetting there; it does not duplicate the routing.
  • doca-debug — the cross-cutting debug ladder. A failed upgrade whose residual symptom lives at a software layer hands off there once the upgrade state is known.
  • doca-programming-guide — the build pattern a consumer rebuilds against after an upgrade. This skill surfaces that a rebuild may be needed; the build mechanics live there.
  • doca-structured-tools-contract — the JSON schemas the agent prefers when present; the upgrade detect step reuses the same doca-env one-shot the version skill uses, when the host has it.

nvidia의 다른 스킬

compileiq-debug
nvidia
무언가 잘못되었을 때 사용: Search()가 멈추거나, 모든 평가가 INVALID_SCORE를 반환하거나, 점수가 개선되지 않거나, 모든 설정이 동일한 숫자를 반환하거나, ptxas 오류 등이 발생할 때
create-github-pr
nvidia
gh CLI를 사용하여 GitHub 풀 리퀘스트를 생성합니다. 사용자가 새 PR을 만들거나, 코드 리뷰를 제출하거나, 풀 리퀘스트를 열고자 할 때 사용합니다. 트리거 키워드 -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
다른 열린 이슈들을 스캔하여 주어진 PR이 함께 수정하거나 실수로 망가뜨릴 수 있는 이슈를 찾습니다. 인접 수정 기회와 모순 위험을 file:line…과 함께 출력합니다.
fhir-basics
nvidia
에이전트에게 FHIR R4 API의 작동 방식, 사용 가능한 리소스, 검색 매개변수를 사용한 쿼리 방법, 모든 응답 형식을 올바르게 파싱하는 방법을 가르칩니다…
compileiq-validate-result
nvidia
검색이 완료된 후, 속도 향상을 청구하거나 ACF를 발송하기 전에 사용합니다. dump_results CSV를 로드하고, 상위 K개 후보(단일 목표)를 추출합니다…
changelog-audit
nvidia
릴리스 전에 Warp CHANGELOG.md를 감사합니다: 누락된 항목 복구, 사용자 영향별 정렬, 항목 언어 다듬기, 줄 바꿈, (릴리스 브랜치 모드) 비교 업데이트…
maintain-dynamic-plugins
nvidia
NeMo Relay 동적 플러그인 로더, 매니페스트, Rust 네이티브 SDK, gRPC 워커 프로토콜, Python 워커 SDK, 문서, 테스트 및 릴리스 워크플로 커버리지를 유지 관리합니다.
dgx-diagnose
nvidia
일반적인 DGX Station GB300 문제 진단 — CUDA 충돌, 잘못된 GPU 타겟팅, vLLM/SGLang 컨테이너 버그, MIG 상태 문제, NVLink/Fabric Manager 오류,…