pytorch · clee2000 · Sep 10, 2025
diff --git a/torchci/clickhouse_queries/commit_jobs_query/params.json b/torchci/clickhouse_queries/commit_jobs_query/params.json
@@ -2,13 +2,15 @@
   "params": {
     "repo": "String",
     "sha": "String",
-    "workflowId": "Int64"
+    "workflowId": "Int64",
+    "runAttempt": "Int64"
   },
   "tests": [
     {
       "repo": "pytorch/pytorch",
       "sha": "85df746892d9b0e87e7a5dfa78ef81a84aec6de0",
-      "workflow_id": 0
+      "workflow_id": 0,
+      "run_attempt": 0
     }
   ]
 }
diff --git a/torchci/clickhouse_queries/commit_jobs_query/query.sql b/torchci/clickhouse_queries/commit_jobs_query/query.sql
@@ -36,7 +36,8 @@ WITH job AS (
         job.torchci_classification_kg.'line_num' as line_num,
         job.torchci_classification_kg.'context' as context,
         job.runner_name AS runner_name,
-        workflow.head_commit. 'author'.'email' AS authorEmail
+        workflow.head_commit. 'author'.'email' AS authorEmail,
+        job.run_attempt AS run_attempt
     FROM
         workflow_job job final
         INNER JOIN workflow_run workflow final ON workflow.id = job.run_id
@@ -51,6 +52,10 @@ WITH job AS (
             {workflowId: Int64} = 0
             OR workflow.id = {workflowId: Int64} -- If a specific workflow ID is provided, filter by it
         )
+        AND (
+            {runAttempt: Int64} = 0
+            OR job.run_attempt = {runAttempt: Int64} -- If a specific run attempt
+        )
         AND job.id in (select id from materialized_views.workflow_job_by_head_sha where head_sha = {sha: String})
         AND workflow.repository. 'full_name' = {repo: String } --         UNION
         AND workflow.name != 'Upload test stats while running' -- Continuously running cron job that cancels itself to avoid running concurrently
@@ -82,7 +87,8 @@ WITH job AS (
         0 as line_num,
         [ ] as context,
         '' AS runner_name,
-        workflow.head_commit.author.email AS authorEmail
+        workflow.head_commit.author.email AS authorEmail,
+        workflow.run_attempt as run_attempt
     FROM
         workflow_run workflow final
     WHERE
@@ -94,6 +100,10 @@ WITH job AS (
             {workflowId: Int64} = 0
             OR workflow.id = {workflowId: Int64} -- If a specific workflow ID is provided, filter by it
         )
+        AND (
+            {runAttempt: Int64} = 0
+            OR workflow.run_attempt = {runAttempt: Int64} -- If a specific run attempt is provided, filter by it
+        )
         AND workflow.repository.full_name = {repo: String }
         AND workflow.name != 'Upload test stats while running' -- Continuously running cron job that cancels itself to avoid running concurrently
 )
@@ -118,6 +128,7 @@ SELECT
     runner_name AS runnerName,
     authorEmail,
     time,
+    run_attempt AS runAttempt
 FROM
     job
 ORDER BY

diff --git a/torchci/components/commit/CommitStatus.tsx b/torchci/components/commit/CommitStatus.tsx
@@ -7,6 +7,7 @@ import {
 import { CommitData, IssueData, JobData } from "lib/types";
 import useScrollTo from "lib/useScrollTo";
 import _ from "lodash";
+import { WorkflowRunInfo } from "pages/api/[repoOwner]/[repoName]/commit/[sha]";
 import { useState } from "react";
 import { linkIt, UrlComponent, urlRegex } from "react-linkify-it";
 import { getConclusionSeverityForSorting } from "../../lib/JobClassifierUtil";
@@ -59,7 +60,7 @@ function WorkflowsContainer({
 }: {
   jobs: JobData[];
   unstableIssues: IssueData[];
-  workflowIdsByName: Record<string, number[]>;
+  workflowIdsByName: Record<string, [WorkflowRunInfo]>;
   repoFullName: string;
 }) {
   useScrollTo();
@@ -117,7 +118,7 @@ export default function CommitStatus({
   repoName: string;
   commit: CommitData;
   jobs: JobData[];
-  workflowIdsByName: Record<string, number[]>;
+  workflowIdsByName: Record<string, [WorkflowRunInfo]>;
   isCommitPage: boolean;
   unstableIssues: IssueData[];
 }) {

diff --git a/torchci/components/commit/WorkflowBox.tsx b/torchci/components/commit/WorkflowBox.tsx
@@ -14,7 +14,10 @@ import {
   ListUtilizationMetadataInfoAPIResponse,
   UtilizationMetadataInfo,
 } from "lib/utilization/types";
-import { CommitApiResponse } from "pages/api/[repoOwner]/[repoName]/commit/[sha]";
+import {
+  CommitApiResponse,
+  WorkflowRunInfo,
+} from "pages/api/[repoOwner]/[repoName]/commit/[sha]";
 import React, { useEffect, useState } from "react";
 import { FaInfoCircle } from "react-icons/fa";
 import useSWR from "swr";
@@ -154,18 +157,18 @@ export default function WorkflowBox({
   jobs: JobData[];
   unstableIssues: IssueData[];
   wide: boolean;
-  allWorkflowIds: number[];
+  allWorkflowIds: [WorkflowRunInfo];
   setWide: any;
   repoFullName: string;
 }) {
   const [selectedWorkflowId, setSelectedWorkflowId] = useState<
-    string | undefined
+    WorkflowRunInfo | undefined
   >(undefined);
-  const workflowId = selectedWorkflowId || jobs[0].workflowId;
+  const workflowId = selectedWorkflowId?.id || jobs[0].workflowId;
 
   const { data: jobsFromSelectedWorkflowId } = useSWR<CommitApiResponse>(
     selectedWorkflowId &&
-      `/api/${repoFullName}/commit/${jobs[0].sha}?workflowId=${selectedWorkflowId}`,
+      `/api/${repoFullName}/commit/${jobs[0].sha}?workflowId=${selectedWorkflowId.id}&runAttempt=${selectedWorkflowId.attempt}`,
     fetcher
   );
 
@@ -180,7 +183,7 @@ export default function WorkflowBox({
 
   const anchorName = encodeURIComponent(workflowName.toLowerCase());
 
-  const { utilMetadataList } = useUtilMetadata(workflowId);
+  const { utilMetadataList } = useUtilMetadata(workflowId?.toString());
   const groupUtilMetadataList = groupMetadataByJobId(utilMetadataList);
 
   const { artifacts, error } = useArtifacts(jobs.map((job) => job.workflowId));
@@ -219,20 +222,30 @@ export default function WorkflowBox({
           <Typography fontWeight="bold" paddingBottom={2}>
             Job Status
           </Typography>
-        </Stack>
+        </Stack>{" "}
         <Stack direction="column" spacing={1}>
           <Stack direction="row" spacing={1}>
             <select
-              value={selectedWorkflowId}
+              value={
+                selectedWorkflowId
+                  ? `${selectedWorkflowId?.id} ${selectedWorkflowId?.attempt}`
+                  : ""
+              }
               onChange={(e) => {
-                setSelectedWorkflowId(e.target.value);
+                const split = e.target.value.split(" ");
+                setSelectedWorkflowId({
+                  id: parseInt(split[0]),
+                  attempt: parseInt(split[1]),
+                });
               }}
-              style={{ width: "100%" }}
             >
               <option value={""}>Select Workflow ID</option>
               {allWorkflowIds.sort().map((id) => (
-                <option key={id} value={id}>
-                  {id}
+                <option
+                  key={`${id.id} ${id.attempt}`}
+                  value={`${id.id} ${id.attempt}`}
+                >
+                  {id.id} (Attempt {id.attempt})
                 </option>
               ))}
             </select>
@@ -275,7 +288,15 @@ export default function WorkflowBox({
         </Stack>
       </Stack>
       {wide && (
-        <TestInfo workflowId={workflowId!} runAttempt={"1"} jobs={jobs} />
+        <TestInfo
+          workflowId={workflowId!.toString()}
+          runAttempt={
+            selectedWorkflowId?.attempt?.toString() ||
+            jobs[0].runAttempt?.toString() ||
+            "1"
+          }
+          jobs={jobs}
+        />
       )}
       <>
         {jobs.sort(sortJobsByConclusion).map((job) => (

diff --git a/torchci/lib/fetchCommit.ts b/torchci/lib/fetchCommit.ts
@@ -1,6 +1,9 @@
 import _ from "lodash";
 import { Octokit } from "octokit";
-import { CommitApiResponse } from "pages/api/[repoOwner]/[repoName]/commit/[sha]";
+import {
+  CommitApiResponse,
+  WorkflowRunInfo,
+} from "pages/api/[repoOwner]/[repoName]/commit/[sha]";
 import { queryClickhouseSaved } from "./clickhouse";
 import { commitDataFromResponse, getOctokit } from "./github";
 import { removeCancelledJobAfterRetry } from "./jobUtils";
@@ -10,12 +13,14 @@ async function fetchDatabaseInfo(
   owner: string,
   repo: string,
   sha: string,
-  workflowId: number
+  workflowId: number,
+  runAttempt: number
 ) {
   const response = await queryClickhouseSaved("commit_jobs_query", {
     repo: `${owner}/${repo}`,
     sha: sha,
     workflowId,
+    runAttempt,
   });
 
   for (const row of response) {
@@ -32,16 +37,29 @@ async function fetchDatabaseInfo(
  * @param jobs
  * @returns
  */
-function getWorkflowIdsByName(jobs: JobData[]): Record<string, number[]> {
+function getWorkflowIdsByName(
+  jobs: JobData[]
+): Record<string, [WorkflowRunInfo]> {
   return _(jobs)
     .groupBy((job) => job.workflowName)
     .map((jobs, key) => {
-      const workflowIds = _(jobs)
-        .map((job) => job.workflowId)
-        .filter((id) => id !== null && id !== undefined)
-        .uniq()
+      const idAndAttempts = _(jobs)
+        .map((job) => {
+          return {
+            id: job.workflowId,
+            attempt: job.runAttempt,
+          };
+        })
+        .filter(
+          (id) =>
+            id.id !== null &&
+            id.id !== undefined &&
+            id.attempt !== null &&
+            id.attempt !== undefined
+        )
+        .uniqBy((id) => `${id.id}-${id.attempt}`)
         .value();
-      return [key, workflowIds];
+      return [key, idAndAttempts];
     })
     .fromPairs()
     .value();
@@ -60,14 +78,15 @@ export default async function fetchCommit(
   owner: string,
   repo: string,
   sha: string,
-  workflowId: number = 0
+  workflowId: number = 0,
+  runAttempt: number = 0
 ): Promise<CommitApiResponse> {
   // Retrieve commit data from GitHub
   const octokit = await getOctokit(owner, repo);
 
   const [githubResponse, response] = await Promise.all([
     octokit.rest.repos.getCommit({ owner, repo, ref: sha }),
-    await fetchDatabaseInfo(owner, repo, sha, workflowId),
+    await fetchDatabaseInfo(owner, repo, sha, workflowId, runAttempt),
   ]);
 
   let jobs = response as any[];

diff --git a/torchci/lib/types.ts b/torchci/lib/types.ts
@@ -30,6 +30,7 @@ export interface JobData extends BasicJobData {
   repo?: string;
   failureAnnotation?: string;
   failedPreviousRun?: boolean;
+  runAttempt?: number;
 }
 
 // Used by Dr.CI

diff --git a/torchci/pages/api/[repoOwner]/[repoName]/commit/[sha].ts b/torchci/pages/api/[repoOwner]/[repoName]/commit/[sha].ts
@@ -2,25 +2,32 @@ import fetchCommit from "lib/fetchCommit";
 import { CommitData, JobData } from "lib/types";
 import type { NextApiRequest, NextApiResponse } from "next";
 
+export type WorkflowRunInfo = {
+  id: number;
+  attempt: number;
+};
+
 export type CommitApiResponse = {
   commit: CommitData;
   jobs: JobData[];
-  workflowIdsByName: Record<string, number[]>;
+  workflowIdsByName: Record<string, [WorkflowRunInfo]>;
 };
 
 export default async function handler(
   req: NextApiRequest,
   res: NextApiResponse<CommitApiResponse>
 ) {
   const workflowId = parseInt(req.query.workflowId as string, 10) || 0;
+  const runAttempt = parseInt(req.query.runAttempt as string, 10) || 0;
   res
     .status(200)
     .json(
       await fetchCommit(
         req.query.repoOwner as string,
         req.query.repoName as string,
         req.query.sha as string,
-        workflowId
+        workflowId,
+        runAttempt
       )
     );
 }