dsh-scnet 0.6.1 → 0.6.2

package/README.md

@@ -4,11 +4,18 @@
 
 DSH-SCNet (`dsh-scnet` on npm) is a community-maintained DSH bundle for operating Supercomputing Network (SCNet). It packages the source [`scnet-hpc`](https://github.com/lql341/scnet-hpc) Skill, profile-aware utilities, and bounded tools for SSH, Slurm, OpenAPI files, regional resources, and Notebook discovery.
 
-Current release: **0.6.1**
+Current release: **0.6.2**
 
 > This is an independent community project. It is compatible with DeepSeek Harness but is not an official DeepSeek product and does not imply endorsement, partnership, or authorization by DeepSeek.
 
-## What's new in 0.6.1
+## What's new in 0.6.2
+
+- Completed jobs now resolve through the fast filtered history-list endpoint rather than timing
+  out on the empty history-detail response.
+- Historical field aliases such as `workdir` are normalized for log-path discovery.
+- OpenAPI operations continue without persistent caching when the cache directory is read-only.
+
+## 0.6.1 highlights
 
 - Job details now fall back to the history endpoint when realtime records expire, with complete
   normalized terminal states.
@@ -113,7 +120,7 @@ synchronizes the skill and version into `dsh-scnet` and `codex-scnet-hpc`. Publi
 `dsh-scnet` npm package only after that downstream change is reviewed.
 
 For a synchronized release, run the package validation on `main`, then trigger the repository's
-`Publish DSH-SCNet` workflow with the exact value from `package.json` (currently `0.6.1`). The
+`Publish DSH-SCNet` workflow with the exact value from `package.json` (currently `0.6.2`). The
 workflow verifies the package/Skill version match, runs validation, creates the tarball, publishes
 stable releases with the `latest` tag, and creates the matching Git tag. Do not bump only
 `package.json`: `skills/scnet-hpc/VERSION`, the package version, and both README release markers

package/README_CN.md

@@ -4,10 +4,16 @@
 
 <p align="center">在 DeepSeek Harness 里操作超算互联网（SCNet）：配置 SSH、管理 Slurm 作业、查询 OpenAPI 区域和文件、检查 Notebook 资源与实例，并保留海光 DCU/DTK 诊断能力。</p>
 
-当前版本：**0.6.1**
+当前版本：**0.6.2**
 
 > 本项目是独立维护的社区项目，与 DeepSeek Harness 兼容，但不是 DeepSeek 官方产品，也不表示 DeepSeek 对本项目提供背书、合作或授权。
 
+## 0.6.2 更新
+
+- 已结束作业改用快速的历史列表过滤查询，不再卡在返回空数据的历史详情接口。
+- 兼容历史记录中的 `workdir` 等字段变体，可继续推导日志路径。
+- 缓存目录只读时自动降级为无持久缓存，OpenAPI 主功能不再被缓存故障阻断。
+
 ## 0.6.1 更新
 
 - 作业实时记录过期时自动回退历史接口，并补全终态映射。
@@ -126,7 +132,7 @@ git 源、本地目录安装以及验证步骤见[安装说明](./docs/installat
 才能发布对应版本的 `dsh-scnet` npm 包。
 
 下游同步合并到 `main` 后，先运行本地校验，再手动触发仓库的 `Publish DSH-SCNet` workflow，
-输入 `package.json` 中的精确版本（当前为 `0.6.1`）。workflow 会校验 package 与 Skill
+输入 `package.json` 中的精确版本（当前为 `0.6.2`）。workflow 会校验 package 与 Skill
 版本一致，执行验证、生成 tarball、以稳定版 `latest` 标签发布，并创建对应 Git tag。
 不要只修改 `package.json`；`skills/scnet-hpc/VERSION`、package 版本和两份 README 的版本
 标记必须保持一致。

package/docs/testing.zh-CN.md

@@ -61,22 +61,22 @@ DEEPSEEK_API_KEY: sk-你的key
 | 8 | 用 scnet_status 查看当前配置 | 返回脱敏配置 | 只读 |
 | 9 | 用 scnet_openapi_regions 查看授权区域 | 返回区域且不含 token | 只读 |
 | 10 | 用 scnet_job_queues 查看昆山队列 | 返回实时队列和空闲资源 | 只读 |
-| 10a | 用 scnet_job_list 列出当前或历史作业 | 返回紧凑作业记录 | 只读 |
-| 10b | 用 scnet_limits 查询资源限制 | 返回用户和调度器限制 | 只读 |
-| 10a | 用 scnet_submit_job（dry_run=true）预览一次提交 | 返回将要发送的请求，不真正提交 | 只读（dry_run） |
-| 10b | 用 scnet_submit_job 提交一个 5 秒小作业 | 返回 job_id、作业目录与日志路径 | 高（会消耗配额） |
-| 10c | 用 scnet_job_show 查该 job_id | 返回状态与 stdout/stderr 路径 | 只读 |
-| 10d | 用 scnet_job_logs 读该作业日志 | 返回日志内容（同时验证 path 与 job_id+work_dir 两种模式结果一致） | 只读 |
-| 10e | 用 scnet_job_cancel 取消一个作业 | 建议先用 dry_run=true 预览 | 高（变更） |
-| 11 | 用 scnet_file_list 查看默认目录 | 返回共享存储文件 | 只读 |
-| 11a | 用 scnet_file_transfer 传输一个文件 | 默认不覆盖目标 | 有副作用 |
-| 12 | 用 scnet_notebook_regions 查看区域 | 返回 Notebook 可用区域 | 只读 |
-| 13 | 用 scnet_notebook_resources 查看昆山资源 | 返回 CPU/GPU/DCU 资源 | 只读 |
-| 14 | 用 scnet_notebook_list 查看实例 | 返回脱敏实例列表 | 只读 |
+| 11 | 用 scnet_job_list 列出当前或历史作业 | 返回紧凑作业记录 | 只读 |
+| 12 | 用 scnet_limits 查询资源限制 | 返回用户和调度器限制 | 只读 |
+| 13 | 用 scnet_submit_job（dry_run=true）预览一次提交 | 返回将要发送的请求，不真正提交 | 只读（dry_run） |
+| 14 | 用 scnet_submit_job 提交一个 5 秒小作业 | 返回 job_id、作业目录与日志路径 | 高（会消耗配额） |
+| 15 | 用 scnet_job_show 查该 job_id | 返回状态与 stdout/stderr 路径 | 只读 |
+| 16 | 用 scnet_job_logs 读该作业日志 | 返回日志内容（同时验证 path 与 job_id+work_dir 两种模式结果一致） | 只读 |
+| 17 | 用 scnet_job_cancel 取消一个作业 | 建议先用 dry_run=true 预览 | 高（变更） |
+| 18 | 用 scnet_file_list 查看默认目录 | 返回共享存储文件 | 只读 |
+| 19 | 用 scnet_file_transfer 传输一个文件 | 默认不覆盖目标 | 有副作用 |
+| 20 | 用 scnet_notebook_regions 查看区域 | 返回 Notebook 可用区域 | 只读 |
+| 21 | 用 scnet_notebook_resources 查看昆山资源 | 返回 CPU/GPU/DCU 资源 | 只读 |
+| 22 | 用 scnet_notebook_list 查看实例 | 返回脱敏实例列表 | 只读 |
 
 第 4 至 7 步只有在你有私钥和目标集群时才测，测前确认副作用。
 
-第 10b、10e 步是变更操作，会真实占用或释放调度器资源：先用 `dry_run=true` 预览，确认返回里的
+第 14、17 步是变更操作，会真实占用或释放调度器资源：先用 `dry_run=true` 预览，确认返回里的
 目标区域/集群，并且**不要在提交超时后盲目重试**（平台没有幂等键），先查作业状态。
 
 Notebook 创建、启动、停止和释放没有作为 DSH deterministic tool 暴露。此类操作通过

package/index.mjs

@@ -197,6 +197,9 @@ function translateScnetError(text) {
       '若只想看日志，请直接给日志文件的 path；若刚提交不久，可稍后重试。'
     )
   }
+  if (/realtime and history endpoints returned no job record/i.test(value)) {
+    return '平台的实时和最近 30 天历史记录中都没有找到这个作业，请核对 job_id 和区域。'
+  }
   return value
 }
 
@@ -577,7 +580,7 @@ export function apply(ctx) {
     defineTool({
       name: 'scnet_job_show',
       description:
-        '只读查询一个作业的状态与资源（state / exit_code / work_dir / stdout / stderr）。作业完成较久后平台可能不再返回记录，此时用 scnet_job_logs 的 path 模式读日志即可。',
+        '只读查询一个作业的状态与资源（state / exit_code / work_dir / stdout / stderr）。自动先查实时记录，再回退最近 30 天历史记录；用户只需提供 job_id。',
       parameters: {
         job_id: { type: 'string', required: true, description: '作业号' },
         backend: { type: 'string', description: 'openapi（默认）或 ssh' },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-scnet",
-  "version": "0.6.1",
+  "version": "0.6.2",
   "description": "Community DeepSeek Harness bundle for SCNet SSH and OpenAPI jobs, files, Notebook resources, Slurm diagnostics, and Hygon DCU/DTK guidance.",
   "type": "module",
   "main": "./index.mjs",

package/skills/scnet-hpc/VERSION

@@ -1 +1 @@
-0.6.1
+0.6.2

package/skills/scnet-hpc/references/openapi.md

@@ -125,6 +125,13 @@ python3 scripts/scnet.py --backend openapi --region <id> \
   jobs --scope history --days 30 --limit 20
 ```
 
+`job <job-id>` checks the realtime endpoint first and automatically falls back to the filtered
+history-list endpoint. The caller does not need to provide `acctTime`.
+
+Region tokens are cached under `${XDG_CACHE_HOME:-~/.cache}/scnet-hpc` when that location is
+writable. Cache locking or writes are best-effort; an unavailable cache never blocks an OpenAPI
+operation.
+
 Uploads larger than 8 MiB automatically use the documented burst/merge flow. Override the
 threshold and chunk size with `--chunk-size <bytes>`; use `--chunk-size 0` to force ordinary
 multipart upload. The current implementation keeps one chunk in memory at a time.

package/skills/scnet-hpc/scripts/scnet_backends/openapi.py

@@ -38,6 +38,7 @@ STATUS_MAP = {
     "statRQ": "REQUEUED",
 }
 WALLTIME_RE = re.compile(r"^(?:[0-9]+-)?[0-9]{1,3}:[0-9]{2}:[0-9]{2}$")
+SCNET_TIMEZONE = timezone(timedelta(hours=8))
 
 
 def _bounded_int(options: Mapping[str, Any], name: str, minimum: int) -> int:
@@ -62,19 +63,23 @@ def normalize_job(item: Mapping[str, Any]) -> dict[str, Any]:
         "name": item.get("jobName") or item.get("name"),
         "state": STATUS_MAP.get(str(raw_state), raw_state),
         "raw_state": raw_state,
-        "queue": item.get("queue"),
-        "user": item.get("user"),
-        "nodes": item.get("nodeNumReq") or item.get("nodeUsed"),
-        "cpus": item.get("procNumReq") or item.get("procNumUsed"),
+        "queue": item.get("queue") or item.get("queueName"),
+        "user": item.get("user") or item.get("userName"),
+        "nodes": item.get("nodeNumReq") or item.get("nodeUsed") or item.get("nodeNum"),
+        "cpus": item.get("procNumReq") or item.get("procNumUsed") or item.get("nproc"),
         "gpus": item.get("gpuNumReq") or item.get("gpuNumUsed"),
         "dcus": item.get("dcuNumReq") or item.get("dcuNumUsed"),
-        "elapsed": item.get("jobRunTime"),
-        "submitted_at": item.get("jobSubmitTime"),
+        "elapsed": item.get("jobRunTime") or item.get("jobWalltimeUsed"),
+        "submitted_at": (
+            item.get("jobSubmitTime")
+            or item.get("jobQueueTime")
+            or item.get("queueTime")
+        ),
         "started_at": item.get("jobStartTime"),
-        "ended_at": item.get("jobEndTime"),
+        "ended_at": item.get("jobEndTime") or item.get("acctTime"),
         "exit_code": item.get("exitCode"),
         "reason": item.get("reason"),
-        "work_dir": item.get("workDir"),
+        "work_dir": item.get("workDir") or item.get("workdir"),
         "stdout": item.get("outputPath"),
         "stderr": item.get("errorPath"),
     }
@@ -371,16 +376,23 @@ class OpenAPIBackend(Backend):
                     result["raw"] = candidate
                 return result
 
-        history = self._history_job(
+        history_items, _, history_raw = self._history_jobs(
             hpc_url,
             token,
             scheduler_id,
-            job_id,
-            options,
+            job_id=job_id,
+            days=30,
+            limit=1,
         )
-        if not isinstance(history, dict) or not (
-            history.get("jobId") or history.get("job_id")
-        ):
+        history = next(
+            (
+                item
+                for item in history_items
+                if str(item.get("jobId") or item.get("job_id") or "") == job_id
+            ),
+            None,
+        )
+        if history is None:
             detail = "realtime and history endpoints returned no job record"
             if realtime_error:
                 detail += f" ({realtime_error})"
@@ -388,7 +400,10 @@ class OpenAPIBackend(Backend):
         result = normalize_job(history)
         result["source"] = "history"
         if options.get("raw"):
-            result["raw"] = history
+            result["raw"] = {
+                "item": history,
+                "response": history_raw,
+            }
         return result
 
     @staticmethod
@@ -405,25 +420,53 @@ class OpenAPIBackend(Backend):
             )
         )
 
-    def _history_job(
+    @staticmethod
+    def _job_items(data: Any) -> tuple[list[dict[str, Any]], int | None]:
+        if isinstance(data, dict):
+            items = data.get("items") or data.get("list") or data.get("rows") or []
+            total = data.get("total")
+        else:
+            items = data or []
+            total = None
+        if not isinstance(items, list):
+            raise BackendError("job list endpoint returned an unexpected data shape")
+        normalized_items = [item for item in items if isinstance(item, dict)]
+        return normalized_items, total
+
+    def _history_jobs(
         self,
         hpc_url: str,
         token: str,
         scheduler_id: str,
-        job_id: str,
-        options: Mapping[str, Any],
-    ) -> Any:
+        *,
+        job_id: str | None = None,
+        days: int = 30,
+        limit: int = 20,
+    ) -> tuple[list[dict[str, Any]], int | None, Any]:
         if not scheduler_id:
             raise BackendError("history lookup requires a scheduler id")
-        path = f"/openapi/v2/historyjobs/{quote(scheduler_id)}/{quote(job_id)}"
-        acct_time = str(options.get("acct_time") or "").strip()
-        if acct_time:
-            path += "?" + urlencode({"acctTime": acct_time})
-        return self._json_request(
+        end = datetime.now(SCNET_TIMEZONE)
+        start = end - timedelta(days=days)
+        params = {
+            "strClusterNameList": scheduler_id,
+            "timeType": "CUSTOM",
+            "isQueryByQueueTime": "false",
+            "start": 0,
+            "limit": limit,
+            "startTime": start.strftime("%Y-%m-%d %H:%M:%S"),
+            "endTime": end.strftime("%Y-%m-%d %H:%M:%S"),
+        }
+        if job_id:
+            params["jobId"] = job_id
+        data = self._json_request(
             "GET",
-            service_endpoint(hpc_url, "hpc", path),
+            service_endpoint(hpc_url, "hpc", "/openapi/v2/historyjobs")
+            + "?"
+            + urlencode(params),
             token=token,
         )
+        items, total = self._job_items(data)
+        return items, total, data
 
     def op_jobs(self, options: Mapping[str, Any]) -> Any:
         hpc_url, token, scheduler_id, _, _ = self._hpc_context(options)
@@ -440,35 +483,23 @@ class OpenAPIBackend(Backend):
                 "start": 0,
                 "limit": limit,
             }
+            data = self._json_request(
+                "GET",
+                service_endpoint(hpc_url, "hpc", path) + "?" + urlencode(params),
+                token=token,
+            )
+            items, total = self._job_items(data)
         else:
             days = _bounded_int(options, "days", 1)
             if days > 90:
                 raise BackendError("days must be <= 90")
-            end = datetime.now(timezone.utc)
-            start = end - timedelta(days=days)
-            path = "/openapi/v2/historyjobs"
-            params = {
-                "strClusterNameList": scheduler_id,
-                "timeType": "CUSTOM",
-                "isQueryByQueueTime": "false",
-                "start": 0,
-                "limit": limit,
-                "startTime": start.strftime("%Y-%m-%d %H:%M:%S"),
-                "endTime": end.strftime("%Y-%m-%d %H:%M:%S"),
-            }
-        data = self._json_request(
-            "GET",
-            service_endpoint(hpc_url, "hpc", path) + "?" + urlencode(params),
-            token=token,
-        )
-        if isinstance(data, dict):
-            items = data.get("items") or data.get("list") or data.get("rows") or []
-            total = data.get("total")
-        else:
-            items = data or []
-            total = None
-        if not isinstance(items, list):
-            raise BackendError("job list endpoint returned an unexpected data shape")
+            items, total, data = self._history_jobs(
+                hpc_url,
+                token,
+                scheduler_id,
+                days=days,
+                limit=limit,
+            )
         result = [normalize_job(item) for item in items if isinstance(item, dict)]
         payload = {
             "scope": scope,

package/skills/scnet-hpc/scripts/scnet_sdk/client.py

@@ -86,28 +86,39 @@ class SCNetClient:
     @classmethod
     def _write_token_cache(
         cls, fingerprint: str, regions: list[dict[str, Any]]
-    ) -> None:
+    ) -> bool:
         path, _ = cls._token_cache_paths()
-        path.parent.mkdir(parents=True, mode=0o700, exist_ok=True)
         temporary = path.with_name(f".{path.name}.{os.getpid()}.tmp")
-        temporary.write_text(
-            json.dumps(
-                {
-                    "created_at": time.time(),
-                    "fingerprint": fingerprint,
-                    "regions": regions,
-                },
-                ensure_ascii=False,
-            ),
-            encoding="utf-8",
-        )
-        os.chmod(temporary, 0o600)
-        os.replace(temporary, path)
+        try:
+            path.parent.mkdir(parents=True, mode=0o700, exist_ok=True)
+            temporary.write_text(
+                json.dumps(
+                    {
+                        "created_at": time.time(),
+                        "fingerprint": fingerprint,
+                        "regions": regions,
+                    },
+                    ensure_ascii=False,
+                ),
+                encoding="utf-8",
+            )
+            os.chmod(temporary, 0o600)
+            os.replace(temporary, path)
+            return True
+        except OSError:
+            try:
+                temporary.unlink()
+            except OSError:
+                pass
+            return False
 
     @classmethod
     def _acquire_token_cache_lock(cls) -> Path | None:
         _, lock = cls._token_cache_paths()
-        lock.parent.mkdir(parents=True, mode=0o700, exist_ok=True)
+        try:
+            lock.parent.mkdir(parents=True, mode=0o700, exist_ok=True)
+        except OSError:
+            return None
         for _ in range(50):
             try:
                 fd = os.open(lock, os.O_CREAT | os.O_EXCL | os.O_WRONLY, 0o600)
@@ -121,6 +132,8 @@ class SCNetClient:
                 except OSError:
                     pass
                 time.sleep(0.2)
+            except OSError:
+                return None
         return None
 
     @staticmethod
@@ -128,7 +141,7 @@ class SCNetClient:
         if lock is not None:
             try:
                 lock.unlink()
-            except FileNotFoundError:
+            except OSError:
                 pass
 
     def request(
@@ -257,7 +270,8 @@ class SCNetClient:
             self._regions_cache = [
                 item for item in data if isinstance(item, dict)
             ]
-            self._write_token_cache(fingerprint, self._regions_cache)
+            if lock is not None:
+                self._write_token_cache(fingerprint, self._regions_cache)
             return self._regions_cache
         finally:
             self._release_token_cache_lock(lock)