chore: improve image worker logging

This commit is contained in:
QiuSW
2026-07-09 11:57:06 +08:00
parent d79bd6fae9
commit 20a73440af
5 changed files with 70 additions and 2 deletions
@@ -4,6 +4,7 @@ import uuid
from django.conf import settings
from django.core.management.base import BaseCommand
from apps.api.models import ImageGenerationTask
from apps.api.image_tasks import reap_stale_image_tasks, run_one_image_task
@@ -56,7 +57,7 @@ class Command(BaseCommand):
task = run_one_image_task(worker_id=worker_id)
if task is not None:
self.stdout.write(f"task={task.task_id} status={task.status}")
self.stdout.write(format_task_log_line(task, started_at=now))
if once:
return
continue
@@ -64,3 +65,18 @@ class Command(BaseCommand):
if once:
return
time.sleep(sleep_seconds)
def format_task_log_line(task: ImageGenerationTask, *, started_at: float) -> str:
duration_ms = max(0, int((time.monotonic() - started_at) * 1000))
alias = str((task.request_payload or {}).get("model") or "")
fields = {
"event": "image_task_processed",
"task_id": str(task.task_id),
"status": task.status,
"alias": alias,
"duration_ms": str(duration_ms),
}
if task.status in {ImageGenerationTask.Status.FAILED, ImageGenerationTask.Status.EXPIRED}:
fields["error_code"] = task.error_code or "upstream_error"
return " ".join(f"{key}={value}" for key, value in fields.items())
+30
View File
@@ -1,5 +1,6 @@
import uuid
import base64
import io
import json
import tempfile
from datetime import timedelta
@@ -13,6 +14,7 @@ from django.conf import settings
from django.contrib import admin
from django.contrib.auth import get_user_model
from django.core.cache import cache
from django.core.management import call_command
from django.test import TestCase, override_settings
from django.urls import path
from django.utils import timezone
@@ -1585,6 +1587,34 @@ class GenerateApiTests(TestCase):
self.assertEqual(poll.data["status"], ImageGenerationTask.Status.FAILED)
self.assertEqual(poll.data["error"]["code"], "upstream_timeout")
def test_run_image_tasks_logs_failed_task_alias_error_and_duration(self):
self.provider.image_error = requests.Timeout("image upstream deadline exceeded")
response = self.post_with_provider(
"/api/v1/generate/image/tasks",
{"prompt": "生成图片", "model": self.image_alias, "resolution": "1K"},
)
out = io.StringIO()
with patch("apps.api.generation.get_provider", return_value=self.provider):
call_command(
"run_image_tasks",
"--once",
"--worker-id",
"worker-log",
stdout=out,
)
task = ImageGenerationTask.objects.get(task_id=response.data["task_id"])
output = out.getvalue()
self.assertEqual(task.status, ImageGenerationTask.Status.FAILED)
self.assertIn("event=image_task_processed", output)
self.assertIn(f"task_id={task.task_id}", output)
self.assertIn(f"alias={self.image_alias}", output)
self.assertIn("status=failed", output)
self.assertIn("error_code=upstream_timeout", output)
self.assertRegex(output, r"duration_ms=\d+")
self.assertNotIn("生成图片", output)
def test_async_image_reaper_fails_stale_running_task_and_refunds(self):
response = self.post_with_provider(
"/api/v1/generate/image/tasks",