Merge pull request #66440 from ClickHouse/ci_not_block_on_few_test_failures

CI: Do not block CI on few number of test failures

tests/ci/ci_config.py

@@ -13,6 +13,9 @@ class CI:
     each config item in the below dicts should be an instance of JobConfig class or inherited from it
     """
 
+    MAX_TOTAL_FAILURES_BEFORE_BLOCKING_CI = 5
+    MAX_TOTAL_FAILURES_PER_JOB_BEFORE_BLOCKING_CI = 2
+
     # reimport types to CI class so that they visible as CI.* and mypy is happy
     # pylint:disable=useless-import-alias,reimported,import-outside-toplevel
     from ci_definitions import BuildConfig as BuildConfig

tests/ci/ci_utils.py

@@ -1,8 +1,9 @@
 import os
+import re
 import subprocess
 from contextlib import contextmanager
 from pathlib import Path
-from typing import Any, Iterator, List, Union
+from typing import Any, Iterator, List, Union, Optional
 
 
 class WithIter(type):
@@ -83,3 +84,15 @@ class Shell:
             check=False,
         )
         return result.returncode == 0
+
+
+class Utils:
+    @staticmethod
+    def get_failed_tests_number(description: str) -> Optional[int]:
+        description = description.lower()
+
+        pattern = r"fail:\s*(\d+)\s*(?=,|$)"
+        match = re.search(pattern, description)
+        if match:
+            return int(match.group(1))
+        return None

tests/ci/merge_pr.py

@@ -26,6 +26,8 @@ from pr_info import PRInfo
 from report import SUCCESS, FAILURE
 from env_helper import GITHUB_UPSTREAM_REPOSITORY, GITHUB_REPOSITORY
 from synchronizer_utils import SYNC_BRANCH_PREFIX
+from ci_config import CI
+from ci_utils import Utils
 
 # The team name for accepted approvals
 TEAM_NAME = getenv("GITHUB_TEAM_NAME", "core")
@@ -251,23 +253,77 @@ def main():
         # set mergeable check status and exit
         commit = get_commit(gh, args.pr_info.sha)
         statuses = get_commit_filtered_statuses(commit)
-        state = trigger_mergeable_check(
-            commit,
-            statuses,
-            workflow_failed=(args.wf_status != "success"),
-        )
 
-        # Process upstream StatusNames.SYNC
-        pr_info = PRInfo()
-        if (
-            pr_info.head_ref.startswith(f"{SYNC_BRANCH_PREFIX}/pr/")
-            and GITHUB_REPOSITORY != GITHUB_UPSTREAM_REPOSITORY
-        ):
-            print("Updating upstream statuses")
-            update_upstream_sync_status(pr_info, state)
+        max_failed_tests_per_job = 0
+        job_name_with_max_failures = None
+        total_failed_tests = 0
+        failed_to_get_info = False
+        has_failed_statuses = False
+        for status in statuses:
+            if not CI.is_required(status.context):
+                continue
+            if status.state == FAILURE:
+                has_failed_statuses = True
+                failed_cnt = Utils.get_failed_tests_number(status.description)
+                if failed_cnt is None:
+                    failed_to_get_info = True
+                else:
+                    if failed_cnt > max_failed_tests_per_job:
+                        job_name_with_max_failures = status.context
+                        max_failed_tests_per_job = failed_cnt
+                    total_failed_tests += failed_cnt
+            elif status.state != SUCCESS and status.context not in (
+                CI.StatusNames.SYNC,
+                CI.StatusNames.PR_CHECK,
+            ):
+                # do not block CI on failures in (CI.StatusNames.SYNC, CI.StatusNames.PR_CHECK)
+                has_failed_statuses = True
+                print(
+                    f"Unexpected status for [{status.context}]: [{status.state}] - block further testing"
+                )
+                failed_to_get_info = True
 
-        if args.wf_status != "success":
-            # exit with 1 to rerun on workflow failed job restart
+        can_continue = True
+        if total_failed_tests > CI.MAX_TOTAL_FAILURES_BEFORE_BLOCKING_CI:
+            print(
+                f"Required check has [{total_failed_tests}] failed - block further testing"
+            )
+            can_continue = False
+        if max_failed_tests_per_job > CI.MAX_TOTAL_FAILURES_PER_JOB_BEFORE_BLOCKING_CI:
+            print(
+                f"Job [{job_name_with_max_failures}] has [{max_failed_tests_per_job}] failures - block further testing"
+            )
+            can_continue = False
+        if failed_to_get_info:
+            print("Unexpected commit status state - block further testing")
+            can_continue = False
+        if args.wf_status != SUCCESS and not has_failed_statuses:
+            # workflow failed but reason is unknown as no failed statuses present
+            can_continue = False
+            print(
+                "WARNING: Either the runner is faulty or the operating status is unknown. The first is self-healing, the second requires investigation."
+            )
+
+        if args.wf_status == SUCCESS or has_failed_statuses:
+            # do not set mergeable check status if args.wf_status == failure, apparently it has died runners and is to be restarted
+            state = trigger_mergeable_check(
+                commit,
+                statuses,
+            )
+            # Process upstream StatusNames.SYNC
+            pr_info = PRInfo()
+            if (
+                pr_info.head_ref.startswith(f"{SYNC_BRANCH_PREFIX}/pr/")
+                and GITHUB_REPOSITORY != GITHUB_UPSTREAM_REPOSITORY
+            ):
+                print("Updating upstream statuses")
+                update_upstream_sync_status(pr_info, state)
+        else:
+            print(
+                "Workflow failed but no failed statuses found (died runner?) - cannot set Mergeable Check status"
+            )
+
+        if not can_continue:
             sys.exit(1)
         sys.exit(0)