fix(collection): correct run status during jobs and add retry-failed

Fix pending mislabeled as fail due to timezone and queue stale checks, show full error details, sync live progress counts, and add retry-failed endpoint for failed NEs only.

Co-authored-by: Cursor <cursoragent@cursor.com>

web/src/i18n/en.ts

@@ -61,6 +61,7 @@ const en = {
     started: "Collection job started: {{id}}",
     paused: "Job paused",
     restarted: "Collection restarted",
+    retryFailedDone: "Retrying failed devices",
     deleted: "Job deleted",
     nothingToRetry: "No failed devices to retry",
     confirmDelete: "Delete this collection job and its log files?",
@@ -80,6 +81,7 @@ const en = {
       collapse: "Hide",
       pause: "Pause",
       restart: "Restart",
+      retryFailed: "Retry failed",
       downloadResults: "Download results",
       delete: "Delete",
       autoRefresh: "Auto-refresh every 2s while jobs are running",

web/src/i18n/zh.ts

@@ -61,6 +61,7 @@ const zh = {
     started: "采集任务已启动：{{id}}",
     paused: "任务已暂停",
     restarted: "已重新开始采集",
+    retryFailedDone: "已开始重采失败网元",
     deleted: "任务已删除",
     nothingToRetry: "没有可重试的失败网元",
     confirmDelete: "确定删除该采集任务？相关日志文件将一并删除。",
@@ -80,6 +81,7 @@ const zh = {
       collapse: "收起",
       pause: "暂停",
       restart: "重新开始",
+      retryFailed: "重采失败网元",
       downloadResults: "下载结果",
       delete: "删除",
       autoRefresh: "任务进行中，每 2 秒自动刷新",

web/src/index.css

@@ -957,6 +957,21 @@ pre {
   background: #fff;
 }
 
+.collect-run-message {
+  max-width: 420px;
+  max-height: 160px;
+  overflow: auto;
+  white-space: pre-wrap;
+  word-break: break-word;
+  font-size: 12px;
+  line-height: 1.45;
+  color: #334155;
+  padding: 6px 8px;
+  background: #f8fafc;
+  border: 1px solid #e2e8f0;
+  border-radius: 6px;
+}
+
 @media (max-width: 1200px) {
   .cards {
     grid-template-columns: 1fr;

web/src/pages/CollectPage.tsx

@@ -9,6 +9,7 @@ import {
   fetchNeCollections,
   pauseCollectionJob,
   restartCollectionJob,
+  retryFailedCollectionJob,
   collectionJobDownloadUrl,
   collectionRunDownloadUrl,
 } from "../services/api";
@@ -104,6 +105,16 @@ export function CollectPage() {
     onError: (err) => showError(String(err)),
   });
 
+  const retryFailedMutation = useMutation({
+    mutationFn: retryFailedCollectionJob,
+    onSuccess: async (job) => {
+      showOk(t("collect.retryFailedDone"));
+      setExpandedJobId(job.id);
+      await invalidateJobs(job.id);
+    },
+    onError: (err) => showError(String(err)),
+  });
+
   const deleteMutation = useMutation({
     mutationFn: deleteCollectionJob,
     onSuccess: async (_, jobId) => {
@@ -294,10 +305,16 @@ export function CollectPage() {
                 onToggle={() => setExpandedJobId(expandedJobId === job.id ? "" : job.id)}
                 onPause={() => pauseMutation.mutate(job.id)}
                 onRestart={() => restartMutation.mutate(job.id)}
+                onRetryFailed={() => retryFailedMutation.mutate(job.id)}
                 onDelete={() => {
                   if (window.confirm(t("collect.confirmDelete"))) deleteMutation.mutate(job.id);
                 }}
-                actionPending={pauseMutation.isPending || restartMutation.isPending || deleteMutation.isPending}
+                actionPending={
+                  pauseMutation.isPending ||
+                  restartMutation.isPending ||
+                  retryFailedMutation.isPending ||
+                  deleteMutation.isPending
+                }
               />
             ))}
           </tbody>
@@ -325,6 +342,7 @@ function JobRow({
   onToggle,
   onPause,
   onRestart,
+  onRetryFailed,
   onDelete,
   actionPending,
 }: {
@@ -334,12 +352,14 @@ function JobRow({
   onToggle: () => void;
   onPause: () => void;
   onRestart: () => void;
+  onRetryFailed: () => void;
   onDelete: () => void;
   actionPending: boolean;
 }) {
   const { t } = useI18n();
   const canPause = job.status === "running";
   const canRestart = job.status !== "running";
+  const canRetryFailed = job.status !== "running" && job.fail_count > 0;
   const canDelete = job.status !== "running";
   const hasResults = (job.output_count ?? 0) > 0;
 
@@ -370,6 +390,11 @@ function JobRow({
               {t("collect.jobs.restart")}
             </button>
           ) : null}
+          {canRetryFailed ? (
+            <button type="button" className="link-btn" disabled={actionPending} onClick={onRetryFailed}>
+              {t("collect.jobs.retryFailed")}
+            </button>
+          ) : null}
           <button
             type="button"
             className="link-btn"
@@ -388,7 +413,14 @@ function JobRow({
       {expanded ? (
         <tr>
           <td colSpan={6}>
-            <JobRunsPanel jobId={job.id} jobStatus={job.status} commands={detail?.job.commands ?? ""} />
+            <JobRunsPanel
+              jobId={job.id}
+              jobStatus={job.status}
+              failCount={job.fail_count}
+              commands={detail?.job.commands ?? ""}
+              onRetryFailed={onRetryFailed}
+              retryPending={actionPending}
+            />
           </td>
         </tr>
       ) : null}
@@ -402,11 +434,17 @@ const RUN_STATUS_OPTIONS = ["pending", "running", "success", "fail", "cancelled"
 function JobRunsPanel({
   jobId,
   jobStatus,
+  failCount,
   commands,
+  onRetryFailed,
+  retryPending,
 }: {
   jobId: string;
   jobStatus: string;
+  failCount: number;
   commands: string;
+  onRetryFailed: () => void;
+  retryPending: boolean;
 }) {
   const { t } = useI18n();
   const [runPage, setRunPage] = useState(1);
@@ -478,6 +516,11 @@ function JobRunsPanel({
             {t("common.clearFilters")}
           </button>
         ) : null}
+        {jobStatus !== "running" && failCount > 0 ? (
+          <button type="button" className="link-btn" disabled={retryPending} onClick={onRetryFailed}>
+            {t("collect.jobs.retryFailed")}
+          </button>
+        ) : null}
       </div>
       {runsQuery.isLoading ? <p>{t("common.refreshing")}</p> : null}
       {!runsQuery.isLoading && runs.length === 0 ? <p>{t("common.empty")}</p> : null}
@@ -498,7 +541,15 @@ function JobRunsPanel({
                 <td>{run.ne_name}</td>
                 <td>{run.ne_ip}</td>
                 <td>{run.status}</td>
-                <td title={run.message}>{run.message || t("common.empty")}</td>
+                <td>
+                  {run.message ? (
+                    <div className="collect-run-message" title={run.message}>
+                      {run.message}
+                    </div>
+                  ) : (
+                    t("common.empty")
+                  )}
+                </td>
                 <td>
                   {run.has_output ? (
                     <a className="link-btn" href={collectionRunDownloadUrl(run.id)} target="_blank" rel="noreferrer">

web/src/services/api.ts

@@ -186,6 +186,9 @@ export const pauseCollectionJob = (jobId: string) =>
 export const restartCollectionJob = (jobId: string) =>
   apiPost<CollectionJobItem>(`/v1/ne-collections/${jobId}/restart`, {});
 
+export const retryFailedCollectionJob = (jobId: string) =>
+  apiPost<CollectionJobItem>(`/v1/ne-collections/${jobId}/retry-failed`, {});
+
 export const deleteCollectionJob = (jobId: string) =>
   apiDelete<{ ok: boolean }>(`/v1/ne-collections/${jobId}`);