toki
98a976a746
- Update vllm-provider-serving-validation.md: unlock state, finalize decisions - Add SDD directory structure for vLLM provider serving validation - Add split task directories: 01_spark_container, 02+01_config_contract, 03+02_openai_vllm_smoke, 04+03_field_smoke - Record decisions on adapter approach, base URL/port, model alias, auth, streaming
48 lines
2.6 KiB
Text
48 lines
2.6 KiB
Text
# Complete - m-vllm-provider-serving-validation/01_spark_container
|
|
|
|
## 완료 일시
|
|
|
|
2026-06-17
|
|
|
|
## 요약
|
|
|
|
DGX Spark vLLM container 기동 검증을 1회 리뷰 루프로 완료했다. 최종 판정은 PASS다.
|
|
|
|
## 루프 이력
|
|
|
|
| Plan | Review | Verdict | 메모 |
|
|
|------|--------|---------|------|
|
|
| `plan_cloud_G07_0.log` | `code_review_cloud_G07_0.log` | PASS | vLLM image digest, launch command/cache/volume, `/v1/models` served model evidence를 확인했다. |
|
|
|
|
## 구현/정리 내용
|
|
|
|
- DGX Spark field host에서 `vllm/vllm-openai:nightly-aarch64` image digest와 ARM64/aarch64 기준을 확인했다.
|
|
- `iop-vllm-qwen36` container launch command, model cache/volume, served model `nvidia/Qwen3.6-35B-A3B-NVFP4` evidence를 기록했다.
|
|
- host port 충돌을 피하기 위해 `8001:8000`으로 노출한 사유와 검증 명령 대체 근거를 남겼다.
|
|
- 리뷰 중 task artifact의 private SSH target/host IP 표기를 `<DGX_SPARK_SSH>`와 `<DGX_SPARK_HOST>`로 정리했다.
|
|
|
|
## 최종 검증
|
|
|
|
- `DGX_SPARK_SSH="<DGX_SPARK_SSH from agent-test/local/rules.md>" ssh "$DGX_SPARK_SSH" 'set -eu; docker image inspect vllm/vllm-openai:nightly-aarch64 --format "{{json .RepoDigests}}"; docker ps --filter name=iop-vllm-qwen36 --format "{{.Names}} {{.Image}} {{.Status}}"; curl -fsS http://127.0.0.1:8001/v1/models'` - PASS; image digest `sha256:1aef087aa5159bcb8f8cb91301ecdf2f6daa6aa41420706afe30b6a9a7001858`, running container `iop-vllm-qwen36`, served model `nvidia/Qwen3.6-35B-A3B-NVFP4` 응답을 확인했다.
|
|
|
|
## Roadmap Completion
|
|
|
|
- Milestone: `agent-roadmap/phase/inference-provider-extension/milestones/vllm-provider-serving-validation.md`
|
|
- Completed task ids:
|
|
- `spark-container`: PASS; evidence=`agent-task/archive/2026/06/m-vllm-provider-serving-validation/01_spark_container/plan_cloud_G07_0.log`, `agent-task/archive/2026/06/m-vllm-provider-serving-validation/01_spark_container/code_review_cloud_G07_0.log`; verification=`DGX Spark vLLM docker image inspect + docker ps + /v1/models field command`
|
|
- Not completed task ids: 없음
|
|
|
|
## Spec Completion
|
|
|
|
- SDD: `agent-roadmap/sdd/inference-provider-extension/vllm-provider-serving-validation/SDD.md`
|
|
- Completed scenario ids:
|
|
- `S01`: PASS; task=`spark-container`; evidence=`agent-task/archive/2026/06/m-vllm-provider-serving-validation/01_spark_container/plan_cloud_G07_0.log`, `agent-task/archive/2026/06/m-vllm-provider-serving-validation/01_spark_container/code_review_cloud_G07_0.log`; verification=`DGX Spark vLLM docker image inspect + docker ps + /v1/models field command`
|
|
- Not completed scenario ids: 없음
|
|
|
|
## 잔여 Nit
|
|
|
|
- 없음
|
|
|
|
## 후속 작업
|
|
|
|
- 없음
|