From 027584656edb4a4bc6fa0836b59c22d1ebccc2f8 Mon Sep 17 00:00:00 2001
From: Eman Elsabban <emanelsabban@yelp.com>
Date: Thu, 16 Nov 2023 13:00:02 -0800
Subject: [PATCH] A quick script to check if a job took longer than x time to
 run and update python in precommit

---
 .pre-commit-config.yaml                |  7 ++-
 requirements-dev.txt                   |  6 +-
 tron/bin/get_jobs_exceeding_runtime.py | 86 ++++++++++++++++++++++++++
 3 files changed, 93 insertions(+), 6 deletions(-)
 create mode 100644 tron/bin/get_jobs_exceeding_runtime.py

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 66128a600..fe9c40164 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,6 @@
 ---
 default_language_version:
-    python: python3.6
+    python: python3.8
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
     rev: v2.5.0
@@ -21,6 +21,7 @@ repos:
     hooks:
       - id: flake8
         exclude: ^docs/source/conf.py$
+        language_version: python3.8
   - repo: https://github.com/asottile/reorder_python_imports
     rev: v1.9.0
     hooks:
@@ -41,8 +42,8 @@ repos:
         language: script
         files: ^tests/.*\.py$
   - repo: http://github.com/psf/black
-    rev: 19.10b0
+    rev: 23.3.0
     hooks:
       - id: black
-        language_version: python3.6
+        language_version: python3.8
         args: [--target-version, py36]
diff --git a/requirements-dev.txt b/requirements-dev.txt
index c87f4b21c..a9a007c25 100644
--- a/requirements-dev.txt
+++ b/requirements-dev.txt
@@ -4,7 +4,7 @@ asynctest==0.12.0
 cfgv==2.0.1
 entrypoints==0.3
 flake8==3.7.9
-identify==1.4.9
+identify==2.4.4
 importlib-resources==1.0.2
 iniconfig==1.1.1
 isort==4.3.18
@@ -15,7 +15,7 @@ mypy-extensions==0.4.3
 nodeenv==1.3.3
 packaging==19.2
 pluggy==0.13.0
-pre-commit==1.21.0
+pre-commit==2.9.2
 py==1.10.0
 pycodestyle==2.5.0
 pyflakes==2.1.1
@@ -26,4 +26,4 @@ pytest-asyncio==0.14.0
 requirements-tools==1.2.1
 toml==0.10.2
 typed-ast==1.4.0
-virtualenv==16.7.5
+virtualenv==20.0.8
diff --git a/tron/bin/get_jobs_exceeding_runtime.py b/tron/bin/get_jobs_exceeding_runtime.py
new file mode 100644
index 000000000..786f7d3c2
--- /dev/null
+++ b/tron/bin/get_jobs_exceeding_runtime.py
@@ -0,0 +1,86 @@
+#!/usr/bin/env python3.8
+import argparse
+import logging
+import sys
+from typing import Optional
+
+import pytimeparse
+
+from tron.commands import cmd_utils
+from tron.commands.client import Client
+
+
+log = logging.getLogger("check_exceeding_time")
+
+STATES_TO_CHECK = {"queued", "scheduled", "cancelled", "skipped"}
+
+
+def parse_args() -> argparse.Namespace:
+    parser = cmd_utils.build_option_parser()
+    parser.add_argument(
+        "--job",
+        default=None,
+        help="Check if a particular job exceeded a time to run. If unset checks all jobs",
+    )
+    parser.add_argument(
+        "--time",
+        help="This is used to specify the time that if any job exceeds will show. Defaults to 5 hours",
+        type=int,
+        dest="time_limit",
+        default=18000,
+    )
+    args = parser.parse_args()
+    return args
+
+
+def check_if_time_exceeded(job_runs, job_expected_runtime) -> list:
+    result = []
+    for job_run in job_runs:
+        if job_run.get("state", "unknown") not in STATES_TO_CHECK:
+            if is_job_run_exceeding_expected_runtime(
+                job_run,
+                job_expected_runtime,
+            ):
+                result.append(job_run["id"])
+    return result
+
+
+def is_job_run_exceeding_expected_runtime(job_run, job_expected_runtime) -> bool:
+    if job_expected_runtime is not None:
+        duration_seconds = pytimeparse.parse(job_run.get("duration", ""))
+        return duration_seconds and duration_seconds > job_expected_runtime
+    return False
+
+
+def check_job_time(job, time_limit) -> list:
+    job_runs = job.get("runs", [])
+    return check_if_time_exceeded(job_runs, time_limit)
+
+
+def main() -> Optional[int]:
+    args = parse_args()
+    cmd_utils.setup_logging(args)
+    cmd_utils.load_config(args)
+    client = Client(args.server, args.cluster_name)
+    results = []
+
+    if args.job is None:
+        jobs = client.jobs(include_job_runs=True)
+        for job in jobs:
+            job_url = client.get_url(job["name"])
+            job = client.job_runs(job_url)
+            results.extend(check_job_time(job=job, time_limit=args.time_limit))
+    else:
+        job_url = client.get_url(args.job)
+        job = client.job_runs(job_url)
+        results.extend(check_job_time(job=job, time_limit=args.time_limit))
+
+    if not results:
+        print("All jobs ran within the time limit")
+    else:
+        print(f"These are the runs that took longer than {args.time_limit} to run: {sorted(results)}")
+    return
+
+
+if __name__ == "__main__":
+    sys.exit(main())