[CI] Reuse old whl: track why failed to use the old whl (#155860)

As in title
Any other things I should track?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/155860
Approved by: https://github.com/malfet

.github/actions/reuse-old-whl/action.yml

@@ -13,6 +13,12 @@ inputs:
   github-token:
     description: GitHub token
     required: true
+  job-id:
+    description: Job ID
+    required: true
+  job-name:
+    description: Job name
+    required: true
 
 outputs:
   reuse:
@@ -30,8 +36,11 @@ runs:
       continue-on-error: true
       env:
         GITHUB_TOKEN: ${{ inputs.github-token }}
+        JOB_ID: ${{ inputs.job-id }}
+        JOB_NAME: ${{ inputs.job-name }}
       run: |
         set -x
+        python3 -m pip install boto3==1.35.42
         python3 ${GITHUB_ACTION_PATH}/reuse_old_whl.py \
           --build-environment "${{ inputs.build-environment }}" \
           --run-id "${{ inputs.run-id }}" \

.github/actions/reuse-old-whl/reuse_old_whl.py

@@ -1,6 +1,7 @@
 import argparse
 import os
 import subprocess
+import sys
 from functools import lru_cache
 from pathlib import Path
 from typing import Any, cast, Optional, Union
@@ -8,6 +9,14 @@ from typing import Any, cast, Optional, Union
 import requests
 
 
+REPO_ROOT = Path(__file__).resolve().parent.parent.parent.parent
+sys.path.insert(0, str(REPO_ROOT))
+from tools.stats.upload_metrics import emit_metric
+
+
+sys.path.remove(str(REPO_ROOT))  # Clean up sys.path after import
+
+
 FORCE_REBUILD_LABEL = "ci-force-rebuild"
 
 
@@ -308,7 +317,7 @@ def parse_args() -> argparse.Namespace:
     return parser.parse_args()
 
 
-def can_reuse_whl(args: argparse.Namespace) -> bool:
+def can_reuse_whl(args: argparse.Namespace) -> tuple[bool, str]:
     if args.github_ref and any(
         args.github_ref.startswith(x)
         for x in [
@@ -318,37 +327,50 @@ def can_reuse_whl(args: argparse.Namespace) -> bool:
         ]
     ):
         print("Release branch, rebuild whl")
-        return False
+        return (False, "Release branch")
 
     if not check_changed_files(get_merge_base()):
         print("Cannot use old whl due to the changed files, rebuild whl")
-        return False
+        return (False, "Changed files not allowed")
 
     if check_labels_for_pr():
         print(f"Found {FORCE_REBUILD_LABEL} label on PR, rebuild whl")
-        return False
+        return (False, "Found FORCE_REBUILD_LABEL on PR")
 
     if check_issue_open():
         print("Issue #153759 is open, rebuild whl")
-        return False
+        return (False, "Issue #153759 is open")
 
     workflow_id = get_workflow_id(args.run_id)
     if workflow_id is None:
         print("No workflow ID found, rebuild whl")
-        return False
+        return (False, "No workflow ID found")
 
     if not find_old_whl(workflow_id, args.build_environment, get_merge_base()):
         print("No old whl found, rebuild whl")
+        return (False, "No old whl found")
         # TODO: go backwards from merge base to find more runs
-        return False
 
-    return True
+    return (True, "Found old whl")
 
 
 if __name__ == "__main__":
     args = parse_args()
 
-    if can_reuse_whl(args):
+    reuse_whl, reason = can_reuse_whl(args)
+
+    if reuse_whl:
         print("Reusing old whl")
         unzip_artifact_and_replace_files()
         set_output()
+
+    emit_metric(
+        "reuse_old_whl",
+        {
+            "reuse_whl": reuse_whl,
+            "reason": reason,
+            "build_environment": args.build_environment,
+            "merge_base": get_merge_base(),
+            "head_sha": get_head_sha(),
+        },
+    )

.github/workflows/_linux-build.yml

@@ -158,6 +158,13 @@ jobs:
           role-session-name: gha-linux-build
           aws-region: us-east-1
 
+      - name: Get workflow job id
+        id: get-job-id
+        uses: ./.github/actions/get-workflow-job-id
+        if: always()
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Check if can use old whl build
         id: use-old-whl
         uses: ./.github/actions/reuse-old-whl
@@ -166,6 +173,8 @@ jobs:
           build-environment: ${{ inputs.build-environment }}
           run-id: ${{ github.run_id }}
           github-token: ${{ secrets.GITHUB_TOKEN }}
+          job-id: ${{ steps.get-job-id.outputs.job-id }}
+          job-name: ${{ steps.get-job-id.outputs.job-name }}
 
       - name: Calculate docker image
         id: calculate-docker-image
@@ -194,13 +203,6 @@ jobs:
         id: parse-ref
         run: .github/scripts/parse_ref.py
 
-      - name: Get workflow job id
-        id: get-job-id
-        uses: ./.github/actions/get-workflow-job-id
-        if: always()
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-
       # Apply the filter logic to the build step too if the test-config label is already there
       - name: Select all requested test configurations (if the test matrix is available)
         id: filter