[AUTOREVERT] Adds circuit breaker with issue in pytorch/pytorch 'ci: disable-autorevert' (#7219)

This adds a circuit breaker to pytorch autorevert. This can be
accomplished by opening an issue on pytorch/pytorch with a tag `ci:
disable-autorevert`.

The decision to NOT gatekeep who are the users that can disable
autorevert is based on `ci: sev`. Currently any user that can open a
issue in PyTorch can add a `ci: sev` issue that is merge blocking. If we
believe this is a safe action and does not require getekeeping, surely
it is OK to disable autorevert with the same level of access.

Author: jeanschmidt

aws/lambda/pytorch-auto-revert/pytorch_auto_revert/__main__.py

@@ -8,6 +8,7 @@
 
 from dotenv import load_dotenv
 
+from .autorevert_circuit_breaker import check_autorevert_disabled
 from .clickhouse_client_helper import CHCliFactory
 from .github_client_helper import GHClientFactory
 from .testers.autorevert_v2 import autorevert_v2
@@ -234,13 +235,22 @@ def main(*args, **kwargs) -> None:
         )
 
     if opts.subcommand is None:
+        repo_name = os.environ.get("REPO_FULL_NAME", DEFAULT_REPO_FULL_NAME)
+
+        if check_autorevert_disabled(repo_name):
+            logging.error(
+                "Autorevert is disabled via circuit breaker (ci: disable-autorevert issue found). "
+                "Exiting successfully."
+            )
+            return
+
         autorevert_v2(
             os.environ.get("WORKFLOWS", ",".join(DEFAULT_WORKFLOWS)).split(","),
             hours=int(os.environ.get("HOURS", DEFAULT_HOURS)),
             notify_issue_number=int(
                 os.environ.get("NOTIFY_ISSUE_NUMBER", DEFAULT_COMMENT_ISSUE_NUMBER)
             ),
-            repo_full_name=os.environ.get("REPO_FULL_NAME", DEFAULT_REPO_FULL_NAME),
+            repo_full_name=repo_name,
             restart_action=(RestartAction.LOG if opts.dry_run else RestartAction.RUN),
             revert_action=(
                 RevertAction.LOG if opts.dry_run else RevertAction.RUN_NOTIFY

aws/lambda/pytorch-auto-revert/pytorch_auto_revert/autorevert_circuit_breaker.py

@@ -0,0 +1,56 @@
+import logging
+
+from .github_client_helper import GHClientFactory
+
+
+logger = logging.getLogger(__name__)
+
+
+def check_autorevert_disabled(repo_full_name: str = "pytorch/pytorch") -> bool:
+    """
+    Check if autorevert is disabled by looking for open issues with 'ci: disable-autorevert' label.
+
+    Args:
+        repo_full_name: Repository name in format 'owner/repo'
+
+    Returns:
+        True if autorevert should be disabled (circuit breaker active), False otherwise
+    """
+    try:
+        gh_client = GHClientFactory().client
+        repo = gh_client.get_repo(repo_full_name)
+
+        should_disable = False
+
+        # Search for open issues with the specific label
+        disable_issues = repo.get_issues(
+            state="open", labels=["ci: disable-autorevert"]
+        )
+
+        for issue in disable_issues:
+            logger.info(
+                f"Found open issue #{issue.number} with 'ci: disable-autorevert' label "
+                f"created by user {issue.user.login}. "
+                f"Autorevert circuit breaker is ACTIVE."
+            )
+            should_disable = True
+
+        sev_issues = repo.get_issues(state="open", labels=["ci: sev"])
+        for issue in sev_issues:
+            logger.info(
+                f"Found open issue #{issue.number} with 'ci: sev' label "
+                f"created by user {issue.user.login}. "
+                f"Autorevert circuit breaker is ACTIVE."
+            )
+            should_disable = True
+
+        if should_disable:
+            return True
+
+        logger.debug("No open issues with 'ci: disable-autorevert' label found.")
+        return False
+
+    except Exception as e:
+        logger.error(f"Error checking autorevert circuit breaker: {e}")
+        # On error, default to allowing autorevert to continue
+        return False